A Goal Programming Approach to

Resource Allocation in
Acute Care Hospitals

John Thomas Blake

A thesis subrnitted in conformity with the requirements

for the degree of Doctor of Phiiosophy
Graduate Department of Mechanical and Industrial Engineering
University of Toronto

Q Copyright by John Thomas Blake, 1997

1+1

National tibrary
of a n a d a

Bibliothque nationale
du Canada

Acquisitions and

Bibliographie Services

Acquisitions et
services bibliographiques

395 Wellington Street

Ottawa ON K1A O N 4

395, nie Wellington

Ottawa ON K1A ON4

Canada

Canada

The author has granted a nonexclusive licence allowing the

National Library of Canada to
reproduce, loan, distribute or seil
copies of this thesis in rnicroform,
paper or electronic formats.

L'auteur a accord une licence non

exclusive permettant la
Bibliothque nationale du Canada de
reproduire, prter, distribuer ou
vendre des copies de cette thse sous
la forme de microfiche/nlm, de
reproduction sur papier ou sur format
lectronique.

The author retains ownership of the

copyright in this thesis. Neither the
thesis nor substantial extracts fkom it
may be printed or otherwise
reproduced without the author's
permission.

L'auteur conserve la proprit du

droit d'auteur qui protge cette thse.
Ni la thse ni des extraits substantiels
de celle-ci ne doivent tre imprims
ou autrement reproduits sans son
autorisation.

Abstract
A Goal Programming Approach to
Resource AUocation in
Acute Care Hospitals
John Thomas Blake
Doaor of Philosophy
Graduate Department of Mechanical and Industrial Engineering
University of Toronto
1997

In an era of decreasing fnding, it is vital that acute care hospitals make the most efficient
and effective use of theu resources. This thesis describes the use of lexicographie goal

programming to allocate resources to physicians at Mount Sinai Hospital in Toronto.

The hospital is modelled as a single firm with two management hierarchies: physicians and
administrators. The hospital's administration and its medical staff are assumed to compete for
control in "a rarefied atmosphere of intrigue and deceit." To jointly satisQ the objectives of these
two groups, a case mix selection model, based on a lexicographicai goal programming approach,
is developed. The algorithm selects a case mix for physicians in such a way that:
1) the hospital generates sufficient revenue to offset the h e d costs of
production;

2) physicians enjoy a stable incorne;

3) disturbance to physicians' preferred mix and volume of cases is rninirnized.

An dtemate view of resource allocation is provided by a second goal programming model

that translates case mix recommendations into a set of commensurate changes to physician
practice. This model identifies practice changes necessary to ensure each physicians cm produce
a preferred mix of cases within the confines of the resources ailocated by the case mix model.

The models are tested against a scenario in which fundimg from the provincial governrnent

is reduced by L 8% over a three year period. Results indicate it is possible to jointiy satisfy the
econornic desires of both hospitais and physicians, through case mix changes alone, at funding
reduction levels of 1 1% or les. M e r an 18% reduction, a joint strategy of pnce reduction and
case mi.change is necessary. An extension is described that allows investigation of price/voIume
tradeoffs.
Mode1 results also suggest the availability of operating room time is a critical determinant
of both hospital profit and physician revenue; wherever possible, administrators should avoid cost
cutting strategies that reduce this resource.
Several policy insights can be der-ived from mode1 results. Tests under both prospective
payrnent and global budgeting suggest hospital funding methodology has little impact on resource
allocation. Nevenheless, to limit the absolute value of fnding reductions institutions should
attempt to achieve strict prospective payment break even in the period immediately prior to a
change in fnding methodoiogy.

Acknowledgements

The responsibility for the content of this paper, with ail of its errors and omissions, Lies
solely with me, the author. The credit for its completion, however, truly belongs to the
individuals who have supported, inspired, or spurred me to action over the past four years. As 1
look back over the course of my studies, 1 reaiize that this paper says more about the faith people
have s h o w in my abilities, rather than my abilities themselves.
1 would like to th&

first of dl, my wife, editor, copy corrector, and chief booster,

Angela Blake. Angela has encouraged me to continue when 1 encountered difficulties. She has
rerninded me to pause, when I've needed perspective. She has been both the voice of reason and
the voice of wonder for me.
I would also like to thank, on a personal and a professional level, my supervisor, Professor

Michael W. Carter. It has been a privilege to work with Mike; he is both one of the brightest and
most approachable persons 1 have ever met. 1 have greatiy valued his input, his suppon, and his
friendship.
My thanks goes out to a11 the members of my research cornmittee. In particular. 1 would
like to thank Dr. Judith Shamian, the Vice-President of Nursing at Mount Sinai, who has time and
time again, made the resources of the hospitd available to me in support of this research. This
work would not have been possible without her support.
Thanks are also due to Ms. Cathy Davis, the Director of Heaith Records at Mount Sinai
Hospital, and Mr. Michael Stewart, the Director of Decision Support at Mount Sinai, for
providing the majority of the data used in this dissertation. 1 am also grateful to Mr. Levy Pineda
of the Ontario Ministry of Health, for patiently and expertly guiding rny request for physician

biiling information through the Ministry's Freedom of Information commission. Mr. Tony Kim, at
the Ontario Case Costing Project, kindly provided the length of stay and average case costing
information used as benchmarks in this work; for his timely response to my pleas for additional
data 1 am etemally grateful.
1 would also iike to thank ail of the staffat Mount Sinai who took the time to meet with

me to discuss, review, read or critique my dissertation. Dr. Lynn Nagle, in particular, kept me

rooted in reality. By asking, "But, so what?", Lynn remuided me that while the aesthetics of
modelling may be interesting, practical results are critical. My thanks are also extended to Dr.
Robert Bell and Dr. Zane Cohen who provided me with much of my insight into physician
preferences.
Finally, 1 would like to thank my parents, Paul and Barbara Blake, for their unwavering
support throughout the years and their encouragement to begin this wondefil joumey of
discovery.

Table of Contents
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

..

II

TableofModeIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv
Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.0Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1TragicChoices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I
1.2 Mount Sinai Hospital . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.0Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 A Review of Hospital Funding in Ontario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3 -0 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1 Refining the Objective Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 Goal Programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4.0 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

4.1 Literature Review Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.0AigorithmOveMew . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1 The Volume Mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
S . 1.1 Volume Mode1 Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.2 Profitability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1- 3 Global Budget / Revenue Cap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.4 Physician Billings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.5 Implied Bounds on Physical Resources . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.6 Case Mix Targets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.7 Operating Roorn Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.8 Inpatient Beds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 The Cost Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.1 The Cost Mode1 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.2 Cost Mode1 Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.3 Available Beds and Length of Stay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.4 Operating Room Availability and Case Time . . . . . . . . . . . . . . . . . . . . . .
5.2.5 Profitability and Direct Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.6 Tota! Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3 A Numerical Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

42
44
51
54
55
58
59
62

67
68
69

70
82
83
85
87
94
95

6.0 Refining the Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

6.1 Refining the Model's Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

6.2 Determining Physician Billings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.3 Applying Appropriate Benchmarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.3.1 LOS Benchmarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.3 .2 Case Cost Benchmark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.3 - 3 Operating Room Time Benchmark . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.4 Operating Room Setup and Tumaround Time . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.5 Non-Detenninistic Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6.6 Team Level Data Aggregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.7 Fractional Case Mix Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.8 Cornpetition and Cooperation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.9 Mode1 Revisions to Run Under a Global Budget . . . . . . . . . . . . . . . . . . . . . . . . . . 135

7.0Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.1 Vdidating the Mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2 Theoretical Vaiidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.2.1 Major Modelling Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.3 Data Validity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.3.1 Mode1 Sensitivity to Changes in Physician Billing Estimates . . . . . . . . . .
7.3.2 Mode1 Sensitivity to Y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.3.3 Data Validity Sumrnary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.4 Logical ValidityA4odel Venfication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.5 Predictive Validity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.5.1 Establishing Face Validity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.5.1.1 Reasonableness of Mode1 Assumptions . . . . . . . . . . . . . . . . .
7.5.1.2 Additional Mode1 Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.5.1.3 Specific Recomrnendations Regarding Case Mx . . . . . . . . .
7.5.1.4 Specific Recommendations Regarding Practice Change . . . . . .
7.5.1.5 Model Acceptability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.5-2 Establishing Predictive Validity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7.6 Validation Surnrnary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.0Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.1 Decision Making Scenario .Phased Budget Reductions . . . . . . . . . . . . . . . . . . . .
8.2.1 Phased Budget Reductions .5% Reduction . . . . . . . . . . . . . . . . . . . . . .
8.2.1.1 Institutional Profit - 5% Reduction . . . . . . . . . . . . . . . . . . . . .
8.2.1.2 Aggregate Physician Billings - 5% Reduction . . . . . . . . . . . . .
8.2.1-3 Specific Physician Billings 5% Reduction Level . . . . . . . . . . . .
8.2.1.4 Case M i x Recommendations - 5% Reduction Level . . . . . . . . .
8.2.1.5 Case Mix Recomrnendations - 5% Reduction . . . . . . . . . . . . .
8.2.1.6 Practice Parameter Recornmendations - 5% Reduction . . . . . .
8.2.2 The 11% Budget Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.2.2.1 Institutional Profit - 11% Reduction . . . . . . . . . . . . . . . . . . . .
8.2.2.2 Aggregate Physician Billings - 11% Reduction . . . . . . . . . . . .
8 .2.2.3 Specific Physician Billings - 11% Reduction . . . . . . . . . . . . . .
8.2.2.4 Case M x Recornmendations - 11% Budget Reduction . . . . . .
vii

137
139
142
143
145
148
157
165

166
167
168
170
171
173
174
175
181

188
190

190
194
194

195
203
206
211
312

214
214
215
216
219

8.2.2.5 Case Mix Recommendations - 11% Reduction . . . . . . . . . . . . 223

8.2.2.6 Practice Parameter Recornmendations .1 1% Reduction . . . . . 225
8.2.3 The 18% Budget Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
8.2.3.1 Institutional Profit - 18% Reduction . . . . . . . . . . . . . . . . . . . . 227
8.2.3.2 Aggregate Physician Biiings - L 8% Reduction . . . . . . . . . . . . 228
8.2.3 -3 SpecSc Physician Billigs - 18% Reduction . . . . . . . . . . . . . . 229
8.2.3.4 Case Mk Recommendations - 18% Budget Reduction . . . . . . 233
8.2.3.5 Case Mix Recomendations - 18% Reduction . . . . . . . . . . . . 237
8.2.3-6 Practice Parameter Recommendations - 18% Reduction . . . . . 239
8.3 Solution Qudity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
8.4 Practical Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
8.4.1 Decreasing Acceptable Physich Income . . . . . . . . . . . . . . . . . . . . . . . . 247
8.4.2 Lowering the Fked Costs of Production . . . . . . . . . . . . . . . . . . . . . . . . 249
8.4.3 Irnproving Ctinical Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
8.5 Using the Mode1 to Determine Mked Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . 257

9.0 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272

9.1 Policy Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
9.2 Institutional Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
9.3 Mode1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
9.4FutureResearch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
9.5Summa1-y . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
282

Appendix A: Example Volume Mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287

Appendix B: Example Cost Mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
Appendix C: Input Database Table Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
Appendix D: Benchmark Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
Appmdix E: Default Model Weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
Appendix F: Background Information and Scenario Presented to Tearn Leaders . . . . . . . . . . 305
Appendix G: Mode1 Recommendations .5% Budget Reduction . . . . . . . . . . . . . . . . . . . . . . . 325
Appendk H: Mode1 Recommendations .1 1% Budget Reduction . . . . . . . . . . . . . . . . . . . . . . 336
Appendix 1: Mode1 Recornmendations .18% Budget Reduction . . . . . . . . . . . . . . . . . . . . . . 347
Appendix J: Model Recommendations Mixed Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358

List of Figures
Figure 5.1 A conceptuai o v e ~ e w
of the resource allocation aigorithm. . . . . . . . . . . . . . . . . . . 44
Figure 5.2 Volume mode1 profitability constrauit space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Figure 5.3 Penalty function defied in P, by constraints 4. 4% and 4b . . . . . . . . . . . . . . . . . . . . 65
Figure 6.1 Product value analysis - Surgicai Services. Mount Sinai Hospital . . . . . . . . . . . . . . 105
Figure 6.2 Regression plot. inpatient p hysician billings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Figure 6.3 Regression plot. outpatient physician billings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Figure 6.4 Frequency plot ofcase-to-case intervals for inpatient surgery at Mount Sinai. . . . . 122
Figure 6.5 Frequency plot of case-to-case intervals for outpatient surgery at Mount Sinai. . . . 123
Figure 7.1 A Hypothetical Relationship Between Nursing Workioad & LOS . . . . . . . . . . . . . . 158
Figure 7.2 Glouberman and Mintzberg Four Faces Mode1. . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Figure 8.1 Hypothetical solution space for a hospital facing budget reductions. assurning
case rnix proportions must be maintained. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
Figure 8.2 Hypothetical solution space for hospital facing budget reductions assurning
case mix can be varied. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
Figure 8.3 The impact of case mk constraints on physician billings. Results shown here
are based on the assumption of high OR time availability. . . . . . . . . . . . . . . . . . . . . 204
Figure 8.4 The impact of case mix constraints on physician billings. Results shown here
are based on the assumption of low OR time availability. . . . . . . . . . . . . . . . . . . . . . . 205
Figure 8.5 Total change to case mix versus hard bounds on case rnix. Results presented
assume high bed and high OR availability at a 5% reduction in institutional
hnding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
Figure 8.6 Mode1 case mix recomrnendations - 5% reduction . . . . . . . . . . . . . . . . . . . . . . . . . 210
Figure 8.7 Physician billings venus case bounds - 1 1% budget reduction. Operating room
time availability assumed to be high. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Figure 8.8 Physician billings versus case bounds - 11% budget reduction. Operating room
time availability assumed to be low. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
Figure 8.9 Total case mix change versus maximum case bounds. 1 1% budget reduction .
Operating room availability high. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Figure 8.10 Total case mix change versus maximum case bounds. 11% budget reduction.
Operating room availability low. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Figure 8.1 1 Distribution of model case mix recommendations - 11% budget reduction.
Operating room time availability assumed hi*; maximum case deviation *60% of
preferred . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
Figure 8.12 Physician billings versus case bounds - 18% budget reduction . Operating
room time avaiiability assumed to be high. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Figure 8.13 Physician billings versus case bounds - 18% budget reduction . Operating
room time availability assumed to be low. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Figure 8.14 Total case mix change versus maximum case bounds. 18% budget reduction .
Operating room availability assumed to be high. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
Figure 8.15 Total case rnix change versus maximum case bounds, 18% budget reduction .
Operating room availability assumed to be low. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
Figure 8.16 Recommended change in volume for CMG 000 (Other Inpatient) and D28
(Endoscopy). 1 8% budget reduction, OR availabiiity assumed low . . . . . . . . . . . . . . . 235

Figure 8.17 Distribution of case rnix recommendations .18% budget reduction .

Operating room tirne availability assumed hi&; maximum case deviation 100%
ofprefemed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Figure 8.18 The ratio of model recornmendations to an across the board (Am)reduction
in total bed days assigned to ciinical tearns. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Figure 8.19 The ratio of model recommendations to an across the board (ATB) reduction
in total direct cost dollars assigned to clinical teams . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
Figure 8.20 The impact of hard bounds on physician billings. . . . . . . . . . . . . . . . . . . . . . . . . . 748
Figure 8.21 The impact of k e d cost on physician bilings after an 18% budget reduction.
Inpatient bed availability is high (42,963). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
Figure 8.22 The impact of fixed cost on physician billings after an 18% budget reduction.
Inpatient bed availability is Iow (35,229). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
Figure 8.23 The impact of clinical efficiency improvement on physician billings after an
18% budget reduction . Inpatient bed availability is assumed to be low (35,229) . . . . . 254
Figure 8.24 The effect of improvements in Oncology team clinical efficiency. AU other
teams are assumed to have improved clinicai efficiency by 8%. . . . . . . . . . . . . . . . . . . 356
Figure 8.25 Theoretical production possibility curve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Figure 8.26 The impact of step size on price/volurne savings . . . . . . . . . . . . . . . . . . . . . . . . . 261

List of Tables
Table 5.3.1 : Example Physician Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Table 5.3-2:Exarnple Physician Sumrnary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Table 5.3-3 Exarnple Penalty Weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Table 5.3.4 Volume Mode1 Case Mix Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . -98
Table 5.3.5 Resource Allocation - Volume Mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Table 5.3.6 Example Satisfied / Dissatisfied Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Table 5.3.7 Updated Resource Availability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Table 5.3.8 Example Direct Nursing Cost Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Table 5-3-9Recommended Physician Profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 03
Table 5.3.10 Resource Allocation - Cost Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I O 3
Table 6.1 Product Value Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 0 6
Table 6.2 CMGs Commonly Missing Billing Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1 1
Table 6.3 Regression Sumrnary Inpatient Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 12
Table 6.4 ANOVA - Regression Analysis for Inpatient Data . . . . . . . . . . . . . . . . . . . . . . . . . 1 12
Table 6.5 Regression Model for Inpatient Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 12
Table 6.6 Regression Summary Outpatient Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 14
Table 6.7 ANOVA - Regression Analysis for Outpatient Data . . . . . . . . . . . . . . . . . . . . . . . . 1 14
Table 6.8 Regression Mode1 for Inpatient Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 14
Table 6.9 Case-to-Case Analysis Inpatient Surgery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Table 6.1O Case-to-Case Analysis Outpatient Surgery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Table 6.1 1 Resource Profile Surgical Council Mount Sinai Hospital 1994195 . . . . . . . . . . . 125
Table 6.12 Exarnple Scenario Results: Non-commensurate Recomrnendations Between
Volume and Cost Mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Table 6.13 Example Scenario Results: Cornmensurate Recornrnendations Between Volume
andCostModel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Table 6.14 Goal Prograrnming Penalties: Cornpetition vs . Cooperation. . . . . . . . . . . . . . . . 134
Table 7. la Model Sensitivity to Changes in Inpatient Billing Regression Coefficient P (Direct
Cost)Under Global Funding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Table 7.lb Model Sensitivity to Changes in Inpatient Billing Regression Coefficient P (Direct
Cost) Under Rate-Based Funding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 1
Table 7.2a Model Sensitivity to Changes in Outpatient Billing Regression Coefficient P,
(Direct Cost)Under Global Funding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Table 7.2b Model Sensitivity to Changes in Outpatient Billing Regression Coefficient P,
(Direct Cost) Under Rate-Based Funding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Table 7.3a Model Sensitivity to Changes in Outpatient Billing Regression Coefficient P1
(Potential Revenue) Under Global Funding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
Table 7.3b Model Sensitivity to Changes in Outpatient Billing Regression Coefficient
(Potential Revenue) Under Rate-B ased Funding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
Table 7.4a Model Sensitivity to Changes in Outpatient Biiling Regression y Intercept a Under
GlobalFunding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Table 7.4b Model Sensitivity to Changes in Outpatient B i h g Regression y Intercept a Under
Rate-Based Funding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Table 7 S a Cost Model Sensitivity to Changes in V - Bed Days* . . . . . . . . . . . . . . . . . . . . . . 160
Table 7.Sb Cost Mode1 Sensitivity to Changes in P - Direct Case Cost* . . . . . . . . . . . . . . . . 160

PL

Table 7.6 Cost Model Sensitivity to Changes in 'P .Specific Recommendations. . . . . . . . . . . 162
Table 7.7 Volume Mode1 Error Cornparison .Cornputer Generated vs. Manual. . . . . . . . . . . 166
Table 7.8 Cost Mode1 Error Cornparison .Computer Generated vs . Manual. . . . . . . . . . . . . 167
Table 7.9 Case Bounds Identified . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Table 7.10 Annualized Resource Consumption Profile .Fiscal 1994/95 . . . . . . . . . . . . . . . . . 185
Table 7.1 1 Model Profiles by Clinical Team. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
Table 8.1 Resource Availabilities .Phased Budget Reductions . . . . . . . . . . . . . . . . . . . . . . . . 192
Table 8.2 Penalty Weights - Economic Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Table 8.3 Penalty Weights Physician Practice Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 193
Table 8.4 Institutional Profit at the 5% Budget Reduction Level . . . . . . . . . . . . . . . . . . . . . . 194
Table 8.5 Total Physician Income at the 5% Budget Reduction Level . . . . . . . . . . . . . . . . . . 195
Table 8.6 Total Physician Income at the 5% Budget Reduction Level (Global Budget) . . . . . 197
Table 8.7 Example reduction profile under global budget . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
Table 8.8 Example reduction profile under rate-based funding . . . . . . . . . . . . . . . . . . . . . . . 198
Table 8.9 DiEerence in Model Recomrnendations Global vs . Rate-Based Funding . . . . . . . . . 200
Table 8.10 Physician Income at the 5% Budget Reduction Level . . . . . . . . . . . . . . . . . . . . . . 204
Table 8. I 1 Total Case Mix Changes at the 5% Reduction Level - Beds High, ORS High . . . . 207
Table 8.12 Total Case M x Changes at the 5% Reduction Level - Beds High, ORS Low . . . . 207
Table 8.13 Volume Model Recommendation Summary .5% Reduction . . . . . . . . . . . . . . . . . 210
Table 8.14 Recommended Changes to Case Mix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
TabIe 8.15 Cost Mode1 LOS Distribution .5% Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
Table 8.16 Cost Mode1 OR Tirne Distribution .5% Reduction . . . . . . . . . . . . . . . . . . . . . . . . 213
Table 8.17 Cost Mode1 Direct Cost Distribution - 5% Reduction . . . . . . . . . . . . . . . . . . . . . . 213
Table 8.18 Institutional Profit at the 1 1% Budget Reduction Level . . . . . . . . . . . . . . . . . . . . 215
Table 8.19 Aggregate Physician Billing - 1 1% Budget Reduction Level . . . . . . . . . . . . . . . . . 216
Table 8.20 Physician Income at the 11% Budget Reduction Level . . . . . . . . . . . . . . . . . . . . . 216
Table 8.2 1 Total Case M x Changes at the 11% Reduction Level .OR Availability High . . . . 220
Table 8.22 Total Case Mix Changes at the 1 1% Reduction Level .OR Availability Low . . . . 220
Table 8.23 Volume Mode1 Recommendation Summary .1 1% Reduction . . . . . . . . . . . . . . . . 222
Table 8.24 Recomrnended Changes to Case Mix .11% Reduction . . . . . . . . . . . . . . . . . . . . . 224
Table 8.25 Cost Model LOS Distribution .1 1% Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Table 8.26 Cost Mode1 OR Time Distribution .11% Reduction . . . . . . . . . . . . . . . . . . . . . . . 226
Table 8.27 Cost Model Direct Cost Distribution - 11% Reduction . . . . . . . . . . . . . . . . . . . . . 226
Table 8.28 Institutional Profit at the 18% Budget Reduction Level . . . . . . . . . . . . . . . . . . . . 228
Table 8.29 Aggregate Physician Billing - 18% Budget Reduction Level . . . . . . . . . . . . . . . . . 229
Table 8.30 Physician Incorne at the 18% Budget Reduction Level . . . . . . . . . . . . . . . . . . . . . 229
Table 8.3 1 Total Case Mix Changes at the 18% Reduction Level - OR Availability High . . . . 233
Table 8.32 Total Case Mix Changes at the 18 Reduction Level - OR Availability Low . . . . . . 234
Table 8.33 Volume Model Recornmendation Distribution .18% Reduction . . . . . . . . . . . . . . 237
Table 8.34 Recommended Changes to Case Mix - 18% Reduction . . . . . . . . . . . . . . . . . . . . . 238
Table 8.35 Cost Mode1 LOS Distribution .18% Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Table 8.36 Cost Mode1 OR Time Distribution - 18% Reduction . . . . . . . . . . . . . . . . . . . . . . . 239
Table 8.37 Cost Mode1 Direct Cost Distribution .18% Reduction . . . . . . . . . . . . . . . . . . . . . 240
Table 8.38 Physician Billing Cornparison: Mode1 vs . Across the Board Cuts . . . . . . . . . . . . . 242
Table 8.39 Institutional Profit Cornparison: Model vs . Across the Board Cuts . . . . . . . . . . . . 242
Table 8.40 LOS Cornparison: Model vs. Across the Board . . . . . . . . . . . . . . . . . . . . . . . . . . 243

xii

Table 8.41 Direct Cost Cornparison: Mode1 vs . Across the Board . . . . . . . . . . . . . . . . . . . . . 243
Table 8.42 Physician Billings versus Hard Bounds on Billings . . . . . . . . . . . . . . . . . . . . . . . . 248
Table 8.43 Physician Biliings versus Fixed Costs - Inpatient Beds High (42,963) . . . . . . . . . . 250
Table 8.44 Physician Billings versus Fixed Costs - Inpatient Beds Low (35.229) . . . . . . . . . . 250
Table 8.45 Physician Billings versus Clinical Efficiency Improvements . . . . . . . . . . . . . . . . . . 254
Table 8.46 Physician Billings versus Oncology Clinical EtFciency Improvement . . . . . . . . . . . 256
Table 8.47 The impact ofs on price& volume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Table 8.48 Team Billing Sumrnary. Iterative Model, Step Size 60% . . . . . . . . . . . . . . . . . . . 262
Table 8.49 Tearn Bed Day Summary, Iterative Model. Step Size 60% . . . . . . . . . . . . . . . . . . 263
Table 8.50 Team OR Time Surnmary, Iterative Model. Step Size 60% . . . . . . . . . . . . . . . . . 264
Table 8.51 Team Direct Cost Summary. Iterative Model. Step Size 60% . . . . . . . . . . . . . . . 264
Table 8.52 Marginal billings per direct cost douar - cornparison with mode1 results . . . . . . . . 265
Table 8.53 Marginal billings per direct cost dollar .comparison with model results d e r
minimum case bounds are eliminated. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
Table 8.54 T e m Case Mix Change Sumrnary, Iterative Model, Step Size 60% . . . . . . . . . . . 267
Table 8.55 Recomrnended Changes to Case Mix - Iterative Solution . . . . . . . . . . . . . . . . . . . 268
Table 8.56 LOS Recomrnendation Distribution - Iterative Solution . . . . . . . . . . . . . . . . . . . . 269
Table 8.57 Case Time Recomrnendation Distribution - Iterative Solution . . . . . . . . . . . . . . . . 270
Table 8.58 Direct Cost Recommendation Distribution - Iterative SoIution . . . . . . . . . . . . . . . 271
Table E.1 Penalty Weights Employed in Production Runs . . . . . . . . . . . . . . . . . . . . . . . . . . 305
Table F. 1 Validation Case Mix Recommendations Presented to Team Leaders . . . . . . . . . . . . 310
Table G. 1 Mode1 Recommendation Summary - 5% Budget Reduction . . . . . . . . . . . . . . . . . . 327
Table G.2 Detailed Mode1 Recornmendations - 5% Budget Reduction . . . . . . . . . . . . . . . . . . 328
Table H. 1 Mode1 Recomrnendation S u r n m q - 1 1% Budget Reduction . . . . . . . . . . . . . . . . . 338
Table H.2 Detailed Mode1 Recommendations - 11% Budget Reduction . . . . . . . . . . . . . . . . . 339
Table 1.1 Mode1 Recornmendation Sumrnary - 18% Budget Reduction . . . . . . . . . . . . . . . . . 349
Table L2 Detailed Mode1 Recomrnendations - 18% Budget Reduction . . . . . . . . . . . . . . . . . . 350
Table J . 1 Mixed Strategy Recornrnendation Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
Table J.2 Mixed Strategy Detailed Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361

...

Xlll

Table of Models
Strategic planning under global budgeting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Strategic planning under rate based funding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
A linear prograrnming problem in canonicai form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Lexicographie goal prograrn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Fieldstein's linear case rnix planning mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Baiigh and Laughhunn tinear economic mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Dowling's linear production mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Morey and Ditiman mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
The Schwartz and Lenard prima1 mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1
The Schwartz and Lenard dual problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
The Hughes and Soliman mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
The Robbins and Tuntiwongpiboon mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 4
The m a i and Pecenka LP mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
The Rifai and Pecenka GP mode1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 3 6
ThePanikmodel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
VolurneModel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
CostModel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

xiv

Glossary
Across-the-Board Cut (ATB): A linear reduction in case volume, length of stay,
operating room case time, or direct cost, equdy assigned to al1 physicians or
clinical teams in a hospital in response to a change in provincial funding.
Attribute: Descriptors of objective reality. Attributes may be objective, or subjective, but
are perceived by decision makers as characteristics of objects in the physical world.
Although attributes c m o t be separated from the decision rnaker's values and
mode1 of reality, they can be identified and measured independently of decision
makers' needs or desires (Zeleny, 1982). See objectives, goals.
Case MU Group (CMG): Case mix grouping is a method of classiQing inpatient cases

into groups having homogeneous resource requirements. There were 492 defined
CMGs in 1994/95. CMGs are analogous to the Arnerican Diagnosis Related
Group @RG).

Costs, Direct: Direct costs are costs that cm be physically traced to a particular object or
segment of production. The Ontario Case Costing Project defines direct costs as
expenses for salaries, supplies, equipment depreciation, and other outlays seen in
the budget of functional, or absorbing, cost centres. (A functional cost centre is an
organizational unit providing care directly to patients, such as Nursing,
Laboratory, Diagnostic Imaging, Clinicai Nutrition and Pharmacy).
Costs, Fixed: Fixed costs are costs that remain constant, regardless of changes in the level
of activity undertaken by an organization.
Costs, Indirect: Indirect costs cannot be traced to any particular units of production and
must therefore be allocated, according to a defined nile, to units of production.
Overhead, for exampie, would be an indirect cost of a product or seMce line.
Overhead is not directly identifiable with any particular product or service line, but
rather is incurred as a consequence of general, overail operating activities
( G h s o n et. al, 1996). The Ontario Case Costing Project defines an indirect cost
as a cost applied against a transient, or supporting, cost centre such as
Housekeeping or Personnel.
Costs, Variable: Variable costs are cost that Vary in direct proportion to changes in the
level of activity undertaken by an organization.

Day Procedure Group (DPG): A day procedure group is analogous to a CMG for
outpatient procedures.
Goals: Goals are specific values or levels, defined in tems of attributes and objectives,
determined a priori by decision makers (Zeleny, 1982). See attributes, objectives.

Goal Programming (GP): A solution technique that rninimizes the set of deviations nom
specified, multiple goals, which are considered simultaneously but are weighted
according to their relative importance (Zeleny, 1982).
Lexicographieal Goal Programming (LGP):A particular fom of goal programming
that minimizes deviations for an ordered set of objectives. The procedure first
determines the alternatives that minimize the deviation of the single most important
objective fiom its correspondhg goal value. From those alternatives the
procedure chooses the alternatives that minirnize the deviation of the second most
important objective frorn its goal. The process continues until al1 objectives have
been considered. This method of lexicographical screening is also known as
preemptive goal programming (Zeleny, 1982).

Linear Programming: A solution technique that selects the optimal quantity and
combination of variables to rnaximize (rninimize) the utilization of a set of
resources.
Objectives: Objectives relate attributes. objective descriptors of reality, to decision maker
needs and desires. An objective represents directions of improvement or
preferences for attributes or groups of attributes. See attributes, goals.
Resource Intensity Weight (RIW):1s a methodology for ranking the relative
consumption of resources by patients within a particular CMG. By ranking all
CMGs relative to a base case, it is possible to compare the output of different
institutions, with divergent case mixes, via cornmensurate units.

1.0 Introduction
Hospitals, like many public institutions in Canada today, face unprecedented economic
pressures. M e r a period of feverish growth throughout the late 1980's in which Canada's real

GNP increased by 3-93% per annum, economic growth stagnated. In the penod from 1990
through 1993, Canada's gross domestic product grew by only 0.3 1 percent (Health Canada 1994).
The recession of the early 1990's coupled with increased concem over public debt has lead to
policies of fiscal restraint at both the provincial and federal levels of govemment.

In Ontario, where 32.5% of provincial spending is devoted to health, a great deai of effon
has been expended to control health care costs. The subject of much debate among hedth policy
analysts (for example see Cuyler 1989), health care spending in Ontario has been stabilized
through dictates placing strict limits on the funds transferred to hospitals and physicians. While in
the period from 1987 to 199 1, real hospital expenditures in Ontario, when measured on a per
capita basis grew by 1.9% per annum, expenditures in the period from 1991 to 1993 fell by 0.2
percent per annurn (Health Canada 1994). As a result of restricted funding, hospitds in Ontario

now face a number of difficult planning and operational issues. These difficulties are likely to
increase as the province withdraws 18% of total hospital funding over the next three years.
1.1 Tragic Choices

Calabresi and Bobbit (1978) define heaith care as a 'tragic choice'. Tragic choices are
choices necessitated by scarcities of critical goods (health services, military manpower, etc.) that
create a confiict for society. Although scarcity can be avoided for some critical goods, it cannot
be avoided for dl. Society must, therefore, decide how scarce resources should be allocated and
distributed. These decisions, by their very nature, guarantee Life, heaith and happiness to some
and death, suffiering, or hardship to others. The implicit limitation on the value of life and the

separation of those who will Live from those who will die is abhorrent to societies founded on
egaiitarian, humanistic principles and yet a praaicai necessity.
Tragic choices consist of a first-order decision that &cesthe amount of a scarce resource
that will be produced and a second-order decision that determines who will benefit from the use
of a good or avoidance of a bad. First-order detenninations define global capacity, whether
existentially imposed, as in a condition of absolute natural scarcity, or chosen from relative
priorities within the larger context of societal interests. Second-order determinations allocate the
available resources defined by the first-order decision. Unless imposed by naturai scarcity, al1
first-order determinations contradict the humanist postulate that life is priceless. Ail second-order
determinations mar the ideals of distributional equality in society. Tragic decisions, note Calabresi
and Bobitt, are abhorrent to societies espousing equality and egalitarianisrn. As a result, societies
tend to jump from allocation policy to allocation policy in an effort to avoid the harsh reality that
such decisions are actually being made. If a bit more effort is expended, cornmon wisdom holds,
the rnistakes of the past can be corrected and "equal" or "equitable" access can be guaranteed. In
fact, the nature of the problem ensures that such a utopian ideai can never be achieved.
Furthemore, note Calabressi and Bobbit, the decision not to make tragic choices - to simply
muddle through as best able - is itself an allocation decision; some people are granted access to a
good or avoid a bad, others however, are not so fortunate. Society, in this instance, has no Say
about how resources are ailocated. While such a policy allows a society to avoid the unpleasant
consequences of making tragic choices, it does not change the fact that such choices are being
made. The decision to provide resources "to al1 those requiring them" is also, they note, a fiction
adopted by societies to disguise the fact that tragic decisions are being made. Because societai
resources are ultimately finite, each dollar ailocated to cancer care, for exarnple, is a dollar that

c m o t be allocated to providing prenatal care for expectant mothers or for providing education,
shelter, or other social s e ~ c e s .
Using the Calabresi and Bobbit framework, health care expenditure in Ontario can be seen
as a tragic choice in which the province makes the first order allocation decision, but relegates
responsibility for the second order decision to others (regional planning authorities, physicians and
hospitals). Thus, resource allocation, by defining the monies to be made available for the
provision of specific seMces within an institution, ultimately determines who in society will have
access to which scarce resources. Because of the gravity of this decision, methodologies must be
developed to enable hospitals and physicians to detemine explicitly the costs and benefits of the
allocation decisions they make.
In this paper, we describe a resource allocation tool for heaith care institutions in Ontario.
In particular, we will focus on Toronto's Mount Sinai Hospital. The rnethodology described in
this paper is an application of multi-objective, linear goai prograrnming.
1.2 Mount Sinai Hospital

Mount Sinai Hospital was founded in 1923 by a group of women fiom Toronto's Jewish
cornrnunity who were inspired to create a facility to meet the medicai needs of the Jewish
community, provide a teaching centre for young Jewish doctors, and serve the city of Toronto
(Mount Sinai Hospital 1994). The vision of a strong, educationally focused institution, serving
the cornrnunity-at-large, with specific emphasis on Toronto's Jewish community, is a deeply
ingrained element of the organizationd culture at Mount Sinai.
In 1923, Mount Sinai was a tiny 33-bed facility located on Yorkville Avenue. Conditions
were, by today's standards, primitive. Despite its inauspicious beginnings, Mount Sinai has
become one of the foremost teaching hospitals in Canada (Mount Sinai Hospital 1994). The

hospital tripled in size in only ten years, growing to 84 beds by 1933. In 1950, the hospital moved

fiom Yorkville Avenue to a 330-bed site at 550 University Avenue. In 1956 Mount Sinzi became
officially filiated with the University of Toronto. This relationship was cemented in 1962 when
the hospital became a member of the University's undergraduate medicd education program. The
increased ernphasis on teaching and research, in addition to a cornmitment to the generai patient
population, ied to rapid expansion. In 1973, the hospital moved fiom 550 University Avenue to
its present location at 600 University Avenue. In l995/96, Mount Sinai admitted 2 1,6O 1 patients.
At that time, the hospital had a staff of 15 15 FTEs and a medical complement of approximately
600. The hospital's global budget in IW5/96 was approximately $138 million.

2.0 Background
The genesis for this research cornes f?om an earlier application of operations research
techniques to surgical process scheduling. When attempting to validate a simulation model of the
surgical process, it was discovered that the master surgical schedule (the document defining the
number and types of rooms available, the hours roorns will be open, and the service or surgeons
who are to be given priority for operating room time) described a pattern of care much different

from the services actually delivered. This observation was consistent across five hospitais in
downtown Toronto. In some instances a variation of 40 percent was observed between the
seMces defined by a master surgical schedule and the services delivered by a hospital (Blake et al
1995). A number of hypotheses were suggested to explain the discrepancy between planned and

delivered senrices. These included random fluctuations7ernergency demand, and periodic

shortages of resources outside the surgical suite. While no definitive cause was apparent, it was
evident that the allocation of resources was non-optimal. However, determining a more
appropnate allocation of resources proved to be a difficult problem. For instance, the naive
solution of granting more resources to seMces that exceeded their allotment and reducing the
resources of seMces using less than their allotted time, creates a perverse incentive for physicians
to consume more than their prescnbed allocation.
Baker (1992) argues that scheduling is a subordinate decision to the rneta-problem of
identi%ng the number and types of products (or services) an organization wishes to offer and
ensuring the appropriate mix of production resources to achieve strategic goals. According to
Baker there is little point in solving the scheduling problem, if the organizational meta-problem,
which Blake and Carter (1 996) label surgicalprocess scheduling, has not been appropnately
solved. It is this problem, the question of identifjmg the number and types of products a hospital

wishes to offer and ensuring the appropriate mix of production resources, that will be addressed in
this thesis.
2.1 A Review of Hospital Funding in Ontario

Public hospital funding in Ontario has undergone several changes since its inception in
1959. When Medicare was first implemented in Ontario, the Ministry of Heafth (MOH) engaged

in a line-by-line review of each hospital's budget (Lave, Jacobs et al. 1992). This technique.
which restricted the ability of administraton to move money from one program to another, was
accompanied by a rapid increase in hospital costs. In response, the province implemented a
technique known as global budgeting. Under global budgeting, hospitais were given an envelope
of funds that they were allowed to disburse as they saw fit, with certain provisos on the level and

quality of care delivered (Lave, Jacobs et al. 1992). Each year, a hospital's budget was
recalculated by adding to the previous year's budget dlowances for inflation, new program
development and life-support programs such as haemodialysis, cancer treatment, and cardiac
surgery. While global budgeting was an improvement over the line-by-line review used before
1969, it sufered from technical and philosophical shortcomings, such as an inequitable

distribution of funds between similar facilities, a lack of incentives to promote efficiency, and a
growth formula too insensitive to volume, acuity changes, or case mix changes.
To overcome the difficulties associated with global budgeting and to promote a move
fiom inpatient to ambulatory services, the MOH and the Ontario Hospital Association (OHA),
through the Joint Policy and Planning Cornmittee (JPPC),created a funding mechanism known as
transitional fnding (Lave, Jacobs et al. 1992). The basis of transitional funding is a system for
comparing the outputs of one hospital to another, via weighted cases. To this end, the JPPC

initiated a series of case rnix groups (CMG9s)' that cluster patients into broad diagnosis
categories with sirnilar resource requirements. In addition, the JPPC adopted a resource intensity
weighting (RIW9)scheme based on data fiom the State of New York that defines the relative
resources used by patients in each CMG. By surnming al1 patients faliing into each CMG and
multiplying each CMG by its associated RIW, it is possible to determine a weighted case rnix for
each facility in the province. This figure, representing a standard facility output, can be compared
to other faciiities and differences in costs between faciiities c m be contrasted via cornmensurate
units (Spencer and Topp 1993). However, to address validity concerns over the use of case data
denved from New York State information, the JPPC initiated a project, known as the Ontario
Case Costing Project (OCCP),to collect detailed resource use data at the patient level. Thirteen
Ontario hospitals, including Mount Sinai Hospital, were given grants in 1992 to purchase
sophisticated cornputer systems to capture, record, and analyze the costs of caring for a particular
individual. Information from these systems will be used to create Ontario RIWs.
While originally envisioned as a tool for ensuring an equitable growth formula, recent
initiatives by the province suggest that transitional funding may be used as a rate based mode1 for
hospitals. Starting in 1989/90 the province set aside money for equalization grants to be made
available to historically under fnded facilities (Lave, Jacobs et ai. 1992). To determine which
facilities were underfunded, the province separated hospitals into five peer groups, including hvo
peer groups for teaching hospitals, which were used to determine an average cost per weighted
case. Hospitals undefinded, relative to their peer group, were given a portion of the equalization
grant. (In 1989 total equalization funds were $25 million.) However, in 1993/94 and again in

'CMG and RIW are trademarks of the Canadian lnstitute of Health Information.

1994/95 the province announced that, as an incentive for promoting operational efficiency, it

would claw back fnds fkom facilities that were over funded relative to their peer group.

This policy effectively changed the nature of hospital funding in Ontario; it replaced global
budgets with a system in which hospitals are paid a standard rate for seMces delivered up to a
mmirnum total revenue which is determined by the province. Although the province has, under

political pressure, deferred action on its policy of clawing back funds, hospitals in Ontario face
potentially senous economic consequences if this policy is adopted and their management
practices do not anticipate it. In this thesis we will investigate techniques for case mix
management under the assumption of an extemally constrained, rate-based Nnding mechanism for
hospitals.

Constrained rate based fnding changes the nature and importance of strategic planning
for hospitals. Hospitds in Ontario, because they are private corporations, are generally free to
choose the mix and volume of cases they will produce, subject to informal seMce agreements
with provincial and regional planners and legislative restrictions that forbid hospitals fiom
incumng deficits. Under global budgeting, a hospital rnay produce any volume and mix of cases it
desires (excluding specially funded life support prograrns) so long as the volume and rnix is within
the production capability of the institution and the total cost of providing services does not exceed
the global budget. While under runs are acceptable, pressure fiom physicians, who are funded
separately on a fee-for-service basis, to maximize services ensures that significant under runs are
rare.

The hospital case mix selection problem may be formulated as a linear programming
program (Dowling 1976). Assume an institution provides n different senices, each approxirnated
by a single case mix group (CMG). Furthemore, assume that the number of services delivered by

the institution is given by the vector x, (j = 1, 2,

...,

n) and that the goal of the institution is to

maximize the total number of seMces it provides. (We will discuss institutional goals in more

detail later.) Under these assumptions, the hospital case selection problem, under global
budgeting, may be approximated as follows:

(3.1 ) Strategic planning under global budgeting:

subjeci to:

vlxl

+ v2x2 + ... +

vnxn s G

Global Budget - Constaraint O

...

alnx,

< b,

Technology Comtrai~zt I

a 2 p I + a 2 2 ~+2...

atnxn

6,

Technology Consiraint 2

6,

Technoiogy Constraini i

allxl

q2x2

...

a,,x,

a,?x,

... + alnxn

bounds:

F = The fixed institzrtional costs of prodirction.

vI = The variable cost of producing CMG j.
G = fie global budget aliocated by the province.
ay = 7he amount of resozrce i required to prodme CMG j
b, = The mmimum capaciiy of resource i.

When soIved for X, Mode1 3.1 rehirns a vector of CMGs that maxirnizes institutional
caseload, subject to constrants on the total amount of money that may be spent (Constraint O),
the technological and human resources available to provide care (Constraints 1 though i), and
bounds on the minimum and maximum number of cases that may be produced. Technology
constraints (C onstraints 1 through i) represent limits on the availability of enabling resources, such
as operating room time or bed days on nursing units.

II

Model 3.1 approximates the decisions hospital managers make, either implicitly or
explicitly, when determining the volume and rnix of seMces their institution wiU provide. While
the decision described in Model 3.1 is to maxirnize case volume, any mix and volume of cases that
does not exceed the global budget is clearly feasible Yoman econornic standpoint. This is
because market efficiencies for hospitals are notably absent under global budget mechanisms. If.
for exarnple, the objective in Model 3.1 was defined as purely profit maximization, the "optimal"
solution for the hospital would be to provide no services whatsoever. However, in Model 3.1 we
have chosen, for illustrative purposes ody,a naive approximation of the objective function of
physicians which assumes that, because they are paid on a fee-for-service basis, physicians have an
economically rational desire to maximize services. This approach, which has been adopted by a
number of authors (Feldstein, 1967; Dowling, 1W6), is intuitively appealing, but somewhat
simplistic (Evans 1984).
Under rate based funding, strategic planning is a more complex problem. Managers must

find a vector of seMces that satisfies budget constraints while generating sufficient revenue to
miarantee the econornic well-being of the hospital without violating technological and human

resource capacities. If we assume the existence of a vector R (= r,, r2, ..., rJ, representing the
revenue an institution receives under rate based funding for each service delivered, the strategic
planning mode1 may be approximated as s h o w in Model 3.2.

(3.2) Strategic planning under rate based funding:

subject to:

vlx,

v2x,

...

< G

vnxn

( r , - v1)xI + (r2 - v2)x2

... +

(rn - vn)xn z F

~(lowr)

ail x,

rJ =
vJ =
G =
F =
a 11 =
bl =

5(upper)

Global Budget - Cottstraint O

Break Even Constrainil O'

for al1 x,

The institzriional revemre gained by producing CMG j.

7he variable inrritrrtionai cost associated with producing CMG j.

The global budget uiIocc(ted by the province.

The fixed irts~ilutionalcosts of production.
The amount of resource i required to produce CMG j.
The mmimtrm copaczty of resource i .

In Model 3 -2, the requirement that the fixed and variable costs of production be less than
the global budget is augmented with a constraint stating that the total c o a of production must not
exceed revenue. A new constraint (0') has been added to force the selection of a mix of CMGs
that generates enough revenue to offset the fixed costs of production. Constraints 1 through i and
the bounds on each xj are carried over fiom Model 3.1. When solved for X, the mode1 described

13

in 3.2 returns a vector of CMGs that maximizes the total number of cases an institution cm
complete in a given planning period, subject to constraints on revenue generation, spending, and
technologicai and human resources.
3.1 Refning the Objective Function

The models described in 3.1 and 3-2 are examples of product rnix selection rnodels - a
well-known class of problems in operations research. In produa mix problems, decision rnakers
must determine the rnix of products necessary to achieve a given profit or cost target (Wagner
1975). in firms operating within the private sector, a particular mix of products that provides the

most profit to shareholders is generaily considered optimal. In the public sector, it is difficult to
accurately determine the value of digerent case mixes, since the shareholder is absent. Thus, it is
impossible to state definitively that any particular mix is "optimal" for a hospitai, its medical staff,
its patient population, or society. Nevertheless, in 3.1 and 3.2 the rnix providing the greatest
absolute number of cases is assumed to be optimal. This is clearly an oversimplification, since
institutions and physicians tend to value cases differently.
The literature provides few practical methods for normatively valuing health care senrices.
Brown and Smith (1994), in a study describing a model to allocate operating room time among
surgeons, suggest that services cm be valued according to perceived clinical importance, medical
severity, expenenced patient need, and "institutional preference." Their model is, however, based
on subjective assessments derived fiom an ernpirical study that cannot be easily generalized. The
literature on health state utility analysis does provide a method for comparing personal
preferences for various health States against one another. By extrapoiating from the results of
health state utility assessments, it is possible to construct estimates of the social utility of
particular services. However, the proportion of heaith state utility assessrnents appearing in the

14

Iiterature, relative to the total number of services that c m be delivered, is infinitesimai. In

addition, methodological concems surroundhg the meaning of averaged individual utility
assessments and the inability to compare utilities collected with different tools render them
inappropriate for use in a broad resource allocation model. The difficulty applying health state
utility assessments for resource allocation in the State of Oregon in the early 1990's illustrates this
point (Ubell et al. 1996). Lndeed, even the rnethodologies used to rank health state utilities for
resource allocation is the subject of debate in the literature (Tengs et al. 1996)
In the absence of a scheme for valuing hospital and medical services let us assume that
both hospitals and their associated medical staff are, at a fundamental level, constrained profit
satisfiers (Simon 1960; Galbraith 1985). That is, assume that neither hospitals nor physicians are
solely interested in profit maximization. Instead, assume that both are interested in achieving an
"acceptable" level of income. Because Ontario hospitals are pnvate, nonprofit organizations, an
acceptable income is clearly the amount of money that allows an institution to recoup the fixed
and variable costs of producing services (Evans 1984). An acceptable level of income for

physicians is, however, more difficult to postdate. The naive assumption that more income is
always better, contradicts market theories of supply and demand and ignores empirical evidence
that physicians are motivated by factors other than income (Evans 1984). Sandier (1989), in an
evaluation of physician incomes in OECD notes that despite the reimbursement scheme in place
and the level of public participation in health care financing, physician income, when compared

with the average industrial wage, shows remarkable stability. Rice (1983) notes that in instances
where physicians' ability to set the pnces for their services is restricted, the amount and intensity
of treatment provided to patients tends to increase. Conversely, the volume and intensity of
treatment provided to patients decreases when physicians are able to rapidly increase their service

15
prices. By adjusting the volume and intensity of seMces in response to changes in reimbursement
schemes, note Hay and Leahy (1982) physicians are able to achieve a target income (Hay and
Leahy 1982). It should, however, be emphasized that in a fee-for-service environment where
physicians act as independent agents, preferred level of incorne is a highly personal decision,
involving implicit tradeoffs between the ability to generate income, conduct research, teach
students, and have the leisure to enjoy the h i t s of one's labours.
Therefore, assume that physicians, as individuals, have desires for particular income and
utilities for the services they provide that are implicit in a self-reported preferred case mix. Also
assume that hospitals are indifferent to the exact seMces they provide, so long as primary
economic goals are achieved and the senices deiivered are within the parameters of the
institution's strategic plan. If hospitais and physicians are, as assumed, constrained profit
satisfiers and physicians, as the "system gatekeepers", determine the exact rnix of services an
institution offers, the objective function of the case selection problem may be stated as foilows.
Determine the volume and mix of cases that:
a) Ensures the hospital, as an institution, is able to generate sufficient revenue
to recoup the fixed and variable costs of production and so ensure its longterm econornic viability.
b) Ensures physicians are able to generate a preferred level of income.
c) 1s feasible, given the production capability of the hospital.
d) Allows physicians, as much as is possible?to perfonn a "preferred" mix and
volume of cases.
Clearly, the objectives listed above are not al1 congruent. Furthemore, not al1 of the goals
can be expressed in commensurable units. Since the decision as outlined above is not amenable to
the traditionai linear prograrnming techniques described in 3.1 and 3.2, linear goal prograrnming
will be adopted to solve the resource allocation problem.

3.2 Goal Programming

Goal programming is a mathematical approach for solving multi-criteria decisions through
the simultaneous solution to a system of (possibly) cornpethg objectives. Schniedejans (1995)
notes that goal programming can be viewed as an extension to linear programming. Linear
programs are optimization problems in which the objective is to maximize a linear fnction of
decision variables, subject to a series of signed constraints that must be satisfied (Winston 1991 ).
In canonical fom, a linear program is given by the following equations:
(3.3) A linear programming problem in canonical forrn:

Minimize :
n

Z =

c,x,

Objective Function

J = l

subject

to:

a,?

brfor i = 1 , ..., m

Technology Constrainrs

/ = 1

X,

OO

j = 1,

. , 11

Non -Negativity Requirements

where x,, x2, .... x, represent a senes of unknown decision variables and c,, q,..., c, are the
marginal contribution of each of the decision variables to the objective function 2. The $ are
technology coefficients, which describe the unit consumption of resource bi associated with the
decision variables x.. The bi are, unimaginatively, known as right-hand side coefficients. Linear
prograrns cm be simply and reliably solved by techniques such as the simplex method or the
interior-point method (Wagner 1975).
Al1 Iinear programming models are based on the following impiicit, and sometimes

restrictive, assumptions:
1) Proponionality: Each unit of the decision variable x, contnbutes exactly c, units to
the objective function and consumes exactly % units of the resource bi.

2) Additivity: The contributions of decision variables to the objective function are

independent of any particular value of the technology constraints.
3) Divisibility: Decision variables may take on fiactional values.
4) Certainty: Al1 parameters (%, b, cj) are known with certainty.
5) Single Criteria: Oniy single, uni-dimensional, goals are included in the objective

tnction.

Originally developed in the 1940's by Dantzig as a tool for allocating fnding for
armaments prograrns (hence the t e n programming), extensions that relax the more restrictive
assumptions of linear programming have been advanced over tirne. Of particular note is the work
by Charnes and Cooper (1968) to extend linear programming to resolve multiple objectives.

In traditionai linear programming problems, such as (3. L), deviations (or slack) from the
nght-hand side values can exist, since the constraints are specified as being less than or equal to
the bi. However, Chames and Cooper note each constraint can be viewed as a goal, with a target
level of b,. If the objective of the decision maker is not to maxirnize a particular function, but
rather to satis& a goal or target, then the decision problem cm be stated as the minirnization of
the deviations from a set of goals. For example:

Mznimize :

If w e note that :

4.

= U2[/

1 ' 1

ayx, - b

ayxJ - bl
j

= 1

]]

Positive hviuti>on Variable

then we may restate (3.4) as:

( 3- 6 )

Minimize :

Subject to:
a,x, +

dl- - dl* = bl for i = 1, ..., rn

1 ' 1

d l - , d , - , x j rO, f o r i = 1

. j

1, ..., n

Furthemore, if we assume weights, w,-, w, associated with positive and negative

deviations from the k, component of the i, goal and a preemptive weight Pi associated with the i,
goal, then w e may restate (3.4) as a standard preemptive goal program:

(3.8) Lexicographic goal program:

Minimize :

Subject to:

dl-, d,', x,

O, for i = 1,

..., m; j =

1, ,

A preemptive weight is a non-Archimedean value (P, > > > P,

> > > ... > > > Pd

that enforces a priority order among goals by ensuring that no value associated with a
deviation at a lower level goal can change the decision variables set by a higher order goal.
When the priority order is strictly enforced, the resulting model is known as a pre-emptive, or
lexicographie, goal programming model.

Lexicographical goal programming (LGP) models have several interesting features that
make them well suited for practical decision rnaking problems. LGP models enable multiple,
and possibly competing, goals to be integrated into a single decision model. In addition, the
concept of goal satisfaction, as opposed to optimization, is a reasonable and well established
approximation for the decision making behaviour of individuals faced with multiple criteria
decisions (Simon 1960). Goal programs, as an extension to lin=

programming models. can

be efficiently solved using existing simplex solution techniques (Ignitzio and Cavalier 1994).

However, GP applications are often criticized for naive relative deviation weights, objective
functions expressed in incommensurable units, or naive priority structures. In addition, it can

be shown that under some formulations, goal programs may renirn dominated solutions
(Schniederjans 1995). (Dominance in multiple-criteria decisions is an important concept,
derived from the work of Italian economist Vilfredo Pareto. A policy is said to be "Pareto
Optimal" if the benefits to any given individual cannot be increased without a corresponding
loss to another person or persans.) A dominated solution occurs in multi-criteria decision
making if and only if an alternative solution can be found that increases the benefit of at least
one goal dimension without worsening any other. Despite its potential problems,
lexicographical goal programming provides a simple, structured framework for analyzing and
comparing competing objectives. While other techniques, such as multi-criteria linear
programming can be shown not to produce dominated solutions, the combinatorial complexity
of a practical problem renden them unsuitable in this instance. Issues of incommensurability.
naive relative weights and naive priority structures can generally be addressed through
appropriate modelling (Schniederjans 1995).

4.0 Literahire Review

The literature on the application of multi-criteria o p t i d t i o n techniques to health care

planning is, in generd, quite sparse (Schniededans 1995). There are, however, several instances
where single criteria optimization techniques are used to model the econornic behaviour of
hospitals.
Feldstein ( 1967), presents a senes of quantitative planning models based on the
assumption that measurement and cornparison of productivity and efficiency are crucial elernents
of management in the public sector, since market forces are absent. In enterprises that produce
products that cannot be valued through market mechanisms, other techniques for evaluating and
cornparhg the cost of production must be identified. Feldstein argues that hospital inputs and
outputs should be measured on sophisticated scales sensitive to variation in caseload and quality
of care. Of particular note is Feldstein's linear prograrnming model for planning case mix:
(4.1) Fieldstein's Iinear case mix planning model:

where w is a vector of weights assigned to each class of case, A is a technology matrix, and x
represents a vector of inputs estimated by Iinear regression. To test his model, Feldstein employs
three sets of weights in the objective function: weights proportional to total expenditures per case
type; weights proportional to the average length of stay per case type; and equal case weights.
The results of these mns produce highly disparate caseloads. Feldstein concludes that objective
functions of case rnix models require more detailed analysis than has been assumed in the

literature. In particular, Feldstein calls for additional efforts to refine estimates of case type
weights, since they have a significant impact on case rnix.
Baligh and Laughhunn (1 969) describe a iiiear prograrnming model, based on a US
model, in which the hospitai is assumed to act as a profit-oriented, econornic entity with
constraints on inputs, outputs, and resource use. Bai@ and Laughhum include in their model the
notion of intermediate outputs consumed by patients who are massed into mutually exclusive
groups on the basis of resource consumption and "importance" to the institution. Designed as a
model to plan admissions, Bdigh and Laughhunn define, but do not solve, a linear product m k
problem for an inpatient acute care institution (4.2). The Baligh and Laughhum model makes the
assumption that, within global limits specified for individuai patient classifications @J and global
patient volume (H(EJ), product mix can be arbitrarily assigned. Baligh and Laughhunn also
assume the existence of a vector, wi, which defines the relative importance of patients in a
particular "importance class".
(4.2) Baligh and Laughhunn linear economic model:
Mmirnize:

Z = &w,(x,
r- 1

subject

EO:

Y,)

Where:

I
1

Decision
Variable

Description

Number of insured patients treated in class i

y,

1 Nurnber of uninsureci patients irateci in class i.

1 ~ o e f i e i e n t1

Description

W,

1 Relative importance of patients in class (medical specialty) i.

a,

1 Number of uni&of
Quanti- of &ce

fixed remurce k required to produce one unit of the j, service.

j provided to patients in class i.

Price charged for service j.

Cost per unit of variable resource 1.
Number of units of variable resource 1 required to produce one unit of the j,
service.

v,

1 Marimum acceptable &action of uninsurd patients in class i.

Description
--

Total availability of fked resource k.

Totai number of patients in ciass i.
-

Target profit for the given planning period.

Fised costs of resources in the given planning period.
Govanment subsidy during the pIannhg period.

Minimum nurnber of patients in c l a s i necessary to meet strategic. ducation, and

cornmunity service goals.

Dowling (1976) draws an analogy between strategic planning in health care and
rnanufactunng production. Although hospitals pursue multiple objectives, their raison d'erre, he
argues, is patient care. Dowling proposes a mode1 to maximize the number of medical and
surgical patients treated during a given planning period, subject to capacity constraints on

fnctional departments.

24

Dowling divides patients fkom a study site into 55 diagnostic categories, each of which is
intended to represent a hornogeneous group of patients having administrative sigdicance.
(Dowling's choice of diagnostic categories is, however, determined as much by data availability as
by clinical meaningfulness.) Each of the 55 diagnostic categories in Dowling's model represents a

group of patients with assumed cornrnon resource consumption patterns, which are modelled via a
resource consumption vector. The resource consumption vector describes the number of
intermediate products consumed by patients in a particular diagnostic category during their stay in
the hospital. Dowling estimates patient resource consumption through a heunstic allocation of
intemediate products to diagnostic categories. From these allocations, Dowling estimates the per
patient consumption of intermediate products, and thus generates a resource consumption vector.
Departmental capacities are estimated through extrapolation of available physical resources or
historical usage. Departments not constrained by physical resources are assumed to be freely
variable and are excluded fkom the analysis.

(4.3) Dowling's linear production model:

Subject to:

Where:
Decision
Variable
S,

Description
Nurnber of patients in category i to be treated

1 Total number of patients treated ( = rx,).

Description

'Is

Quantity of prduct j required by patients in category i.

Description

Maximum number of product j that can be produced.

Lowet bound on patients in category i.

5-

Upper bound on patients in categoq i.

Description

Misceiiaaeous

Product,

1 Paediatric patient days.

Producs

Med-surg patient days.

Product,

Psychiatrie patient days.

Product,

Obsteuics patient days

Product,

Newborn patient days.

Product,

Laboratory tests.

Product,

X-Rays.

Producg

Deliveries.

The Dowling model is conceptually appealing, but suffers from a number of practical
problems. Because the data required to model an acute care institution was not readily available
when the study was undertaken in the early 19701s,Dowling was forced to estimate the
consumption of intermediate products through crude methods. A multiple linear regression

26

relating departmental outputs to raw patient inputs failed to achieve sigdicant results, and so was
abandoned in favour of a heuristic allocation of departmental resources to medicai seMces based
on a weighted average consumption of departmental resources by an average patient. To estimate
the intermediate departmental products consumed by a patients from a group of medicai services.
the average consumption of departmental resources produced was rnultiplied by the proportion of
patients from each medicai specialy receiving seMces relative to the total departmentai
complement. This produced an allocation of departmental resources for each medical seMce that
was subsequently divided by the total number of patients in the seMce to produce an estimate of
average per patient consumption of resources. Dowling notes that such estimates are cmde; tests
for correlation coefficients proved weak.
The lack of accurate data clearly hinders the applicability of the Dowling model. Dowling
employs average per patient resource consumption estimates as coefficients in the technology
constraints of the rnodel linking patient input to departmental outputs. Even if such estimates are
accurate at the outset of the problem, changes to the proportion of patients receiving
departmental services, suggested by the model, may invalidate the coefficients used in the
technology matnx. Given the large changes in case mix suggested by Dowling (10-)O%), the use
of average data is clearly inappropriate. Also lirniting the applicability of the Dowling model are

the bounds %' and xio* used to set limits on the minimum and maximum number of cases that can
be performed by each medical s e ~ c e Dowling
.
sets the lower case bounds at values equal to
each service's volume in the previous fiscai period; upper case bounds are set on an ad hoc basis.
Thus, the Dowling model returns suggestions only for increases to case load. In addition, the
Dowling model contains no policy levers included that would allow a hospital to identify or
achieve target case loads for physicians. Dowling, for instance, recommends that the study

hospital increase its production of medical-surgical patient days by 12.7%, but is mute on how
this might best be accomplished. (Should al1 physicians increase their case volume by 12.7%?
Should additional physicians be recmited?) Furthemore, since economic constraints are not
included in Dowling's model, it is not possible to gauge the economic impact of its proposed
recornmendations. The most valuable information provided by the Dowling model, therefore, is
the shadow price of physically constrained resources. (Winston (199 1) defines a shadow pnce of
constraint i as the arnount by which the objective finction of a Iinear program could be improved

if the right-hand side of the ith constraint were to be increased by one unit. Shadow prices are
detemined fiom the optimal solution to the dud of a linear programming problem. For more
detail on Iinear programs and their duals see Winston.)
Morey and Dittrnan (1984) descnbe the use of a non-linear model to set service prices and
allocate fixed costs to direct seMce departments in American hospitals operating under Medicare
reimbursement policies. Morey and Dittrnan assume that hospitals operate as constrained profit
satisficers in an environment of voluntady or regdatory imposed revenue caps. Restrictive
reimbursement policies are assumed to limit the latitude of administrators to allocate costs to
service departments. In addition, the market behaviour of private (non-Medicare) patients
restricts the ability of hospitals to set the pnces for various types of services arbitrarily.
Morey and Dittman assume a hospital consisting of N departrnents, numbered j = 1, 2,

. ..,

N. Each department in their model delivers sj services, indexed i = 1, 2, ..., si. The hospital is
assumed to have a mix of Medicare and pnvate patients. Private patients are assurned to pay the
price of seMces set by the hospital, within the confines of reasonableness, as dictated by the local
market. Medicare reimbursement, however, is assumed to be the lesser of the total billed
Medicare charges,

or the departmental ratio of Medicare charges to total charges

Under these assumptions, the hospital profit planning model can be formulated as a non-linear
prograrnrning problem of the following form:

(4.6) Morey and Dittman model:

Subject to:

Where:

P,, SP,,

( i = 1, 2, . s; j = 1, 2, ..., N )

Decision
Variable

Ci

Description

1 Cost (a decision variable) dlocated to department j.

Coefficient

1
(

4
IQ,

Description

1 Dernand for mits of service i in departmentj.

1 Proportion of demand for service i in departmentj made by Medicare patients.

RHS

P,L'
Cl

C,"

Description

Global budget-

Upper bound on the price pnvate consumers are willing to pay for service i in
department j .

Lower limit on the costs that can be allocated to department j.

( Upper limit on the costs that can be aiiocated to department j.

TotaI cost of running the hospital.

Because of the difficulties associated with achieving optimal solutions to large non-linear
models, Morey and Dittman adopt a linear approximation to their model in which the department
Medicare ratios (4.5) are fixed and product pnces are determined. Once prices have been
updated, new department Medicare ratios can be calculated. The process continues, iteratively,
until pnces stabilize. Aithough this heuristic requires less solution time than the non-linear model
in test applications, the quality of the solutions returned was found to be poor.

The Morey and Dittman model is an interesting application with several conceptual
parallels to the resource allocation problem. The idea of hospitals acting as constrained profit
satisfiers in a rate-based fnding environment constrained by a global revenue limit, clearly
suppons the assumptions used to develop the objective function described in Section 3.1. In

30

addition, their use of price as an administrative control for profit satisfaction can be easily adapted
to the resource allocation problem. However, because the Morey and Dittman model is designed
specifically to identify optimal pricing policies under the US Medicare plan, its utility in other
jurisdictions is limited.
Schwartz and Lenard (1994) also describe the use of linear programming to set prices
for care under the American prospective payment system (PPS). In the United States, the
price established for each procedure is determined from the average cost of treating patients
with similar diagnoses. Prospective payment, when introduced in the mid 1980's. was

described as an effort to introduce market forces into the delivery of publicly subsidized health
care. While PPS has not reduced health care spending, it has helped to slow the rate of
inflation for health care services in the US. As a result, many private insurers have adopted
prospective payment systems similar to that of Medicare.
Under PPS, prices for services are calculated annually fiorn an average of prices
charged patients at a sarnple of institutions (Schwartz and Lenard 19%). Changes in cost due
to new technology, efficiency, or economies of scale associated with specialization are
gradually introduced into the price structure through annual updates. PPS prices can also be
used as policy mechanisms to encourage institutions to specialize in care deemed beneficial to
society and to provide disincentives for socially "undesirable" a r e . Schwartz and Lenard,
however, daim that the use of policy pricing, because it gives al1 institutions incentives to
produce the same products, does not necessarily provide society with low cost health a r e .
Instead, they argue hospitals should be encouraged, through prospective pricing , to provide the
products that they are able to deliver most efficiently. Accordingly, Schwartz and Lenard
develop a societal model for identifying the least expense health care delivery mechanism:

(4.7) The Schwartz and Lenard prima1 model:

Subject to:

.....M

CxurD,

~ ' 1 . 2

Where :
Decision
Variable

q j

Description
Number of patients of type j treated at hospital i.

Description
-

cij

1 - 1

Average cost per patient for hospitai i to treat patient type j.

D ~ s pt
c ion
~

Dj

1 h e -ber

K,

1 Totai nurnber of patients of aii types treated at hospitd i.

of patients of type j treated at al1 hospitais.

I
1
1

This model, when solved, yields a solution descnbing the lowest possible cost a society

can expend for a given arnount of health services. It also yields %', the number of patients in
DRG i to be treated in facility j. The solution to the duaf of this problem,

(4.8) The Schwartz and Lenard dual problem:

Subject to:
for ail i and j
r, - s1 d,,
rj r 0
3 2 0

where :
r, O"d s, are aral vmiubles

yields the optimal pnce (pj = r,') for DRG i necessary to ensure the societal demand for services is
met at the least possible price. This solution, note Schwartz and Lenard, requires no regdatory
body to allocate patients to hospitals; complernentary slackness ensures that whenever si*= r,'
= 0,

- C,

qjwill be greater than zero. Plainly stated, complementary slackness guarantees that the

most profitable option for an institution is to specialize in the production of seMces that are
socially rnost beneficiai. Similarly, when %j = O, complementary slackness guarantees that profit.

si', will be less than r,'

- Ci,.

Thus, institutions will have no incentive to produce products that are

not optimal for society.

Hughes and Soliman (1989) draw an analogy between case mix and product mix. Because
case mix can be thought of as a hospital's mix of produas, they argue, linear programming can be

used to select a product mix for institutions funded under the US prospective payment (ie. ratebased) model. Arguing that fixed costs cannot be readily changed, they describe the use of Iinear
programming model to maximize contribution, within the constraints of existing capacity, through
an optimal case mix.

(4.9) The Hughes and Soliman model:

Subject

to:

581

b f l d i Total Available Bed Days

d - 1

un

ofid x, Total Available Operating Room Hours

d - l
437

l f i s Total Available Labour and Delivery Hours

d - 1
487

= f i d s Total Available CT Hours

d - 1
437

n3,

Total Available MCU Days

d - 1

Where:

Decision
Variable

Description
I

Number of DRG d to be produced.

Xd

1 Coefficient 1
1
1

Description

bd

Od

1 Average operating rmrn rime for patients in DRG d (in hours).

Average length of suiy of patients in DRG d (in days).

1,

Average labour and delivay suite tirne for patients in DRG d (in hours).

Cd

Average CT scan time for patients in DRG d (in hours).

nd

Average M C U length of stay for patients in DRG d (in ciays).

rd

Revenue for DRG d.

"d

1 Average variable cost of DRG d.

34

When solved, note Hughes and Soliman, this model yields both an optimal case mix for

an institution as well as a set of dual (or shadow) prices for scarce resources; a sensitivity
analysis can be run to determine the range of contribution over which the chosen solution
remains optimai. Based on the results of a test case, Hughes and Soliman suggest LP models
be used to identiQ the marginal vaiue of change to scarce resources. Case mix, they suggest,
can then be controlled by adding or removing physicians or by modifying the availability of
other scarce physical resources.
Robbins and Tuntiwongpiboon (1989)also present a model des igned to maximize short
term profit by case rnix management, but add consmaints specifying lower bounds for each

DRG.
(4.10) The Robbins and Tuntiwongpiboon model:

Subject to:
Total Availab le Diagnostic Tesring Hours
Total Availab le Bed Days
Total Availuble Nrrrsing Hours
Total Availab le Pharmacy Budget

Minimum Number of Cases ; y = 1, 2, .... 487

Where:

Decision
Variable
%i

Description
-

--

Number of DRG d to be produced.

Description

1 Average diagnostic testmg for parie*

bd

- -

--

. . --

in DRG d (indays).
--

1 Average length of stay of patients in DRG d (in d i s ) .

4i

Average nursing time consumed patients in DRG d (in hours).

Pd

Average pharmacy cost of patients in DRG d (in doars).

-

rd

Revenue for DRG d.

Vd

Average variable cost of DRG d.

-- - -- -

When solved, this mode1 yields both an optimal case mix for an institution as well as a
set of dual (or shadow) prices for scarce resources; a sensitivity analysis can then be nin to
determine the range of contribution over which the chosen solution remains optimal.
Robbins and Tuntiwongpiboon draw particular attention to the effect of shadow price
on case mix selection. The shadow price of a particular DRG, not its contribution. they argue.
determines whether an economically rational institution w il1 increase, or decrease, its capacity
to produce a particular type of case. Such results, they note, may be counter-intuitive.
However, Robbins and Tuntiwongpiboon, like Hughes and Soliman, suggest no policy actions,
other than adding or removing medical staff, that would allow an institution to modify its case
mix.

In addition to the Iiterature describing single criteria decision models, there is a srnaIl
body of research on the application of muiti-criteria planning models to health a r e .
Schniederjans, in a comprehensive review of goal programming published in 1995, lis@fewer

36

than fifty health related papers. This result is surprising, given the complex nature of decision
making in health care organizations and the conflicting goals created by hospitals' dual
organization structure (Harris 1977). Of the papers listed by Schniederjans few are applicable
to the resource allocation problem; most describe staff mix (Lee 1973). capital budgeting
(Jackman 1973; Chae, Suver et al. 1985), regional planning (Nelson and Wolch 1985),
marketing (McGlone and Calatone 1992), or equipment purchasing (Lee 1973) applications.
Rifai and Pecenka (1989). however, use a health care setting to illustrate how goal
programming can be used to al locate resources in situations with multiple, competing
objectives. The Rifai and Pecenka model is not an exhaustive study of decision making in
hospitals. It is simply an illustration of GP methodology through a sample problem, consisting
of an operating room, recovery room, and a nursing unit, which serves patients undergoing
one of four types of surgery (tonsillectornies, appendectomies, hernia repairs, and
cholecystectomies). Rifai and Pecenka contrast the single objective linear programming
problem of maximizing profit,
(4.1 1) The Rifai and Pecenka LP model:

Subject to:
3x,

4x2 + 8x3
8 5 + 2 3 + 4x3
4x, + 6x2 + 4x3
+

+
+

6x4 i 1,100
2r, s 1,400
4x4 i 400

Operating room hours

Recovery room hozrrs
Bed h y s

to a goal programming model with objectives of achieving a target profit of $50,000 and
making complete use of productive capacity:

(4.12) The Wai and Pecenka GP model:

210x,

2 6 0 ~+ ~280x3
3x,

4x2

300x4

8x3

6x4

dl-

- dlT=

di

50,000

d2*= 1,100

8x1

2u,

4x3

2r,

d,- - dg = 1,400

4x,

6.5

4x3

4x4

d i - d4*=

400

Profit Target
Operuting room hours
Recovery room hmrs
Bed dqys

The linear prograrn in 4.1 1 identifies a mix of 100 hernia patients (x,) and 50
cholecystectomy (x,) patients and generates a total profit of $43,000, but leaves 900 hours of
recovery room tirne unused. The goal program retums a rnix of 130 tonsillectomy patients ( x , )
and 88 hemia (x,) patients, with a maximum profit oF$52,230, that makes complete use of al1
existing resources. Rifai and Pecenka use the results fiom this model to show that goal
prograrnming cm b e used to balance the competing objectives inherent in health care delivery and
provide a strategic advantage in the marketplace.
Panitz (1988) describes the use of both multi-objective linear programming and goal
programming to select service rnix in mental health facilities. Panitz describes a problem facing
community mental health centres in the United States in which management must select the
mental health seMces (adult outpatient services, child outpatient services, chronic mental illness
programs and client seMces management) they wish to deliver and the staffnecessary to do so
such that:
1) The absoIute number of clients served is maxirnized.

2) The number of staretained to provide client services is maximized.

3) The number of "client hours" delivered is maximized.

The selection of the seMce rnix and staff, however, is limited in practice by the following
constraints:
1) The seMces offered must meet local demand requirements by age, sex and
rninority status.

2) The services offered must be feasible given the selected staff.

3) The services offered, in terms of staf hours, must be consistent (ie greater

than or equal to) good clinical practice.

4) The seMces offered must be less than or equal to the physical capacity of
the study site.

5) The total revenues generated by the selected services must be less than or
equal to the availabie funding fiom US Federal, State, charitable, and
private sources.
6) The cost of delivenng each service must be minimized.

7) The delivered seMces rnust meet legislated requirements.

Panitz decomposes the problem into three related sub problerns based on the trio of

objectives Iisted above. Each model identifies a different service rnix for the CMHC. However,
the selection cntena varies between sub-rnodels. The first sub-mode1 selects a service mix based

on the absolute number of clients served; the second selects service rnix by staff; the third uses
client contact hours as the decision variable. The objective functions of the sub-models are:
(4.13) The Panitz model:

yx,

Mm

M a

2 R,

Mmimize Clients

xi Mminiize St@ Hours

B,x, Mminrize Revenue

Mm
t

Min

C,x1 Minjmize COSZ

Where:

Description

Decision
Variable

xi

1 Number of clients to be provided service i.

Yi

Number of staff to be hired to provide seMce i.

Wi

Number of client hours of seMce i to be provided.

1 Coefficient (
Description
1 H, 1 Client contact per hour of service i.
I

Ei

Staff hours per hour of service i.

Bi

Billing per hour of service i.

ci

1 Cost per hour of service i.

I
1

Panitz's models are subject to a number of constraints including revenue limits and
restrictions on cost in several budget categones. They are also constrained to be technologically
feasible with respect to the physical limitations of the study facility and existing clinical practice
guidelines. A number of additional constraints are also included to maintain consistency between
model sub-components and to define the relationships between dernand for pnmary services and
secondary (inpatient services) which were sub-contracted at the study site.
Panitz employs multi-objective linear prograrnrning (MOLP) to solve each of the submodels separately and together in an aggregate model. The client mode1 identifies two efficient
solutions, the staff model identifies three and the client hour model identifies two. When the subrnodels are combined, a total of twenty-four efficient solutions are identified; the efficient
solutions found by the sub-models, however, are not entirely replicated by the aggregate model.

Of the twenty-four solutions identified by the aggregate model, Panitz rejects twelve out of hand
as being unacceptable, because they do not achieve an a prion requirement of minimum levels for
some of the seMces offered by the CMHC.

40

Panitz aiso converts the aggregate MOLP mode1 to a goal prograrnrning problem and

resolves it, using the goal program as a filter for the MOLP to reduce the number of efficient
solutions presented to the decision rnaker. The results of the goal program were not, however,
satisfactory; large variations were identified in the level of seMces provided when achievement
levels were ailowed to fluctuate by relatively small arnounts. Panitz therefore incorrectly
concludes that goal programming is ineffective for service mUc selection problems, despite the fact
that the model retums, as it should, a lexicographie minimum for a given set of goals.
The Panitz model is of note for a variety of reasons. It encompasses the notion of rnultiobjective optimization for service mix selection within the context of health care. The model
includes constraints for both revenues and costs and is, therefore, similar in nature to the resource
allocation problem in acute care hospitals. There are, however, a number of difficulties with
Panitz's approach as well as his conclusions. From a purely practical standpoint, multi-objective
linear programming is simply too cumbersome for a problem of the magnitude of the resource
allocation problern. Panitz's aggregate model includes a total of 12 variables and 36 constraints.
Since even a restncted version of the resource allocation problem has well over 12,000 variable
and 6,000 constraints, the number of possible solutions necessary to identie the entire problem
solution space for a problem of this size is enonnous. To fully describe the possible solution
space, millions or billions of vertices must be enumerated. Ignizio and Cavalier (1 994) note that
the total number of extreme points for a problem with n variables and m constraints approaches

{[(n + m)!] / [m!n!]). Thus, the time required to generate the solution space would be
prohibitive. In addition to the time required to generate the complete solution space, the number
of solutions that must be presented to the decision maker is too large for rational selection. Thus,
a filtering algorithm would be required to narrow choices to a manageable number. Therefore, it

41

would be unlikely that the decision maker would be able to identiS, the "optimal" solution; at best,
such an approach would produce a satisficing solution - exactly the kind of solution that a preemptive goal programming approach produces.
4.1 Literature Review Summary

A review of the literature suggests that there has been inadequate coverage of the
resource allocation problem in acute care hospitals. While extensive literature exis for the
single criteria problem, little work appears on the application of multi-criteria techniques to
health a r e . Single criteria models, however, typically yield naive results. For example, the
single criteria objective of profit maximization, yields solutions in which the most "profitable"
cases in an institution's portfolio are maximized at the expense of al1 others. Sirnilarly, the
objective of rnaximizing caseload, as assumed by Dowling, merely transfers a preference for
cases based on "profitabilityn to one based on cost or resource consumption. Simple, low cost
or low intensity cases, are generally preferred under case maxirnization models to higher cost,
higher intensity cases. While single criteria models produce significant theoretical results,
there is little to suggest that the solutions to most single criteria models are, in any sense,
practical. Because physicians are independent agents whose behaviour is not controlled by the
hospital, their case mix (and thus the case mix of the institution) cannot be dictated. The naive
application of industrial planning models to health care is clearly inappropriate; planning
models must employ more realistic objective functions that capture the complexities and the
subtleties of manag ing health care organizations. Goal programrning provides an appropriate
framework for formulating the resource allocation problem. Unforhinately, few useful
examples appear in the literature.

Although there is reluctance to view health care as a decomposable product (Harris 1977),
we can, for the purposes of allocating resources, draw an anaiogy between a hospital funded
under a rate-based methodology and a manufacturing firm (Evans 1984). While the analogy of
hospital as production fim is not perfect, it does provide a starting point for analyzing the
resource allocation decision.
Overall, the revenue generated by a firm manufactunng a good, or selling a s e ~ c eis, the
product of sales and unit price. Ifwe assume a nmi. consisting of Y independent production units
(factories, sales centres, franchises) with differing costs of production, sells N products,
, x,,, ,,x
represented by the vector X = ( x , , , x , ~.-.,

and that each product is sold for a

..., x,),

pice given by the vector R = (r,, r,, ..., r,), then the total revenue generated by an organization is
given by:
(5.1)
Y

Total Revernre

.v

rnxy,,
y = l n = l

Sirnilarly, if F represents the fixed costs of production and V = (v,,,vlZ,..., v,,, vE,
v,)

..

..

represents the variable costs of production, then the total cost incurred by an organization is

described by:

Total Cost = F

v Y n ~

Clearly, the income an organization receives for the seMces it delivers is simply total
revenue less total cost. In equation f o m this is:

Total Incorne

y = L n = l
Y
N

y = l n = l

Interpolating fiom the general case to an acute care hospitd, we note that revenues,
represented by the vector 9 are set by the province and k e d costs of production, given by the
scalar F in ( 5 3 ) cannot in the short term (ie less than one year) be changed by the hospital.
Therefore, hospital income is a fnction of the variable cost of production v, and the total
quantity of any good or seMce produced 5.
In this chapter we describe a two stage algorithm to adjust both the volume and costs of
production in an acute care hospital. Phase 1determines a volume and mix of cases for each
provider that:
a) Ensures the econornic stability of the hospital.
b) Guarantees physicians, as rnuch as possible, a preferred level of income.

c) 1s technically feasible.
d) Minirnizes, as much as is possible, the disturbance to providers' preferred mix of
services.

In Phase 1 (Volume Model), the costs of production are held fixed and an optimal provider
case mix is detemiined. Using the case mix determined in Phase 1, an allocation of resources
(revenue, overhead, beds, operating room tirne, etc.) c m be inferred. This information is used in
Phase II (Cost Model) to determine how the unit costs of production and unit consumption of
resources should be altered to ensure physicians are able to produce a preferred rnix and volume

44

of cases, while remaining within the econornic and resource boundaries determined in Phase I (see
Figure 5.1 .)
Figure 5.1 A conceptual o v e ~ e w
of the resource allocation ale;orithrn.
Conceptual O verview

Resource Allocation Algorithm

Province

Volume Model
(costs AEltrmd Flxpd)

Initial Resource Assignment

(BaMd on Volume Rasulb 6 Phy+tdan Ractlte Pdmm]

Cost Model
(Voiume hsumad Fixad)

5.1 The Volume Model

Assume a hospital delivers N diEerent services, each represented by a case mix group.
Furtherrnore, assume in this institution there are Y physicians, each of whom delivers some subset
of the N different services to their patients. Thus, the variable )hi represents the number of CMG
n's delivered by doctor y, and the array X = (x,,, x,, ..., x,,, x,
of physicians

..., )x,

describes both the output

(1,x, for ail y) and the output of the institution ayz,Q.

Assume that associated

with each physician y is a vector,',P

= (p,, , p,, , ...,

p,,'),

which represents his or her preferred

volume and mix of cases. If the amount physicians are able to bill the province for professional
services in relationship to a particular CMG is given by the vector B, = (b,, b,

...,

h),then we

may estimate a physician's preferred income as:

Assume, also, that each provider c m s p e c e lower and upper bounds on the number of
each case he or she wishes to produce,

(5-5)
Lower, s x,,,, r Upperp

for d l y, n

as well as bounds on the total billings he or she wishes to receive.

(5.6)

Lower,,.

(;

C bnxyn

Gpperbr. for ail y, n

n = l

The bounds on billings s h o w in 5.6 represent limits on physician willingness to trade income for
leisure. The lower bound represents the minimum amount of income a physician deems
acceptable. Under the assumption that additional physician workload is required to generate
increased income, the upper bound represents the minimum amount of leisure time considered
acceptable.
If the fixed costs of production in the hospital are given by the scalar F and the variable

cost of delivering CMG y in conjunction with physician n is v, then the total cost of production
is:

Total Cost

v x

+
y = l n = l

P Y"

Similarly, ifthe revenue accruable to the hospitai under a rate-based scheme fiom the
provision of a particular senrice n is r,, the total revenue accruable to the hospital is the lesser of
the giobal budget, G, or:

(5.8)
Y

Total Revenue =

.v

r,xyn

y = l n = l

Finally, assume that the hospital can estimate upper and lower bounds on the number of
each CMG that should be produced:

xp s Ulpper,, for n = 1, 2,

Lower, s

...,

y = l

Lower bounds, in this instance, represent the hospital administration's estimate of the
minimum number of each CMG required to meet the health needs of the hospital's catchment
population; upper bounds represent the hospital's estimate of the maximum number of each CMG
that should be delivered.

Ifwe assume, as stated in Section 3.1, that hospitals and physicians are, foremost, profit
satisficers and that hospitals will defer to the production wishes of physicians, if economic goals
are met, we may fornulate the volume-rnix selection problem as a preemptive linear goal
program:

5.10 Volume Mode1 (Continues on next page)

IV

Y
Y"

n - l y - l

Subjecr

2) F

5E

v x

P' yn

n - l y - l

+
;

- di

G -

To:

Proft

for al1 y
for all y

for al1 y
IV

4b)
n- l

IV

-ORjn

diR,- diR, = n - 1 ( 1 -qHardm) ORp PJ;

for al1 y
for all y

5.1 O Volume Model: Goal Programming Model for Case M k Selection

for al1 y; n
=

Upper>,Pre/erred - P;

for all y: n

p, - Lower, Prpierred

for al! y; n

Gl&

for sorne selected y; n

s Upper,,

Ahoftue

Lower,

.4hro&e

s xv. s Upperl,

for al/ y: n

W here:

Decision
Variable

Description
-

)5n

Coeficient
rn

Description
The institutional revenue associateci with the production of CMG n.

v~

1 The variable institutional costs associateci with the production of CMG n by doctor y.

bn

1 The physician b d h g associated with the production of CMG o.

OR,

1 The amount of operating roorn time associateci with the production of C M G n by doctor y.

LOS,

1 The nurnber of bed days associateciwith the producrion of CMG n by doctor y.

1
1

RBS
F

Profit
G

b;
Lowerby

p'm

Description
The fiscd costs of production.

1 A user-defmable profit requirernent

The global budget or revenue cap for the institution, as determineci by the provincial government.

1 The target billing (desired incorne) for doctor y.

1 The (bard) lower bound on income for doctor y.

Doctor y's preierred level of production of CMG n.

Lowernaw

The hard lower bound on the number of CMG n's that rnay be produced by al1 doctors.

UppernaO*

The hard upper bound on the number of CMG n's that may be produced by d l doctors.

LowerWRcfd

The soi? lower bound on ihe number of C M G n's doctor y can produce.

U~ptr,,%~-

The sofl upper bound on the number of CMG n's doctor y can produce.

UpperW*b"'

The hard upper bound on the number of CMG n's doctor y can prduce.

1 Lo~rr,*"'~ 1 The hard Iowa bound on the number of CMG n's doctor y can produce.
ORA-,,,

BedsA,ble
qHard,

The nurnber of CMG n's assigned to doctor y.

The total amount of operating room tirne available.

1 The total nuinber of bed days available.

A fraction (O s qHard,
r 1 ), whch represents the minimum proportion of cwrently assigned bsds
considered acceptable. Thus C,(1 - q H U P,'represents the minimum acceptable nurnber of brds
that can be assigneci to doctor y. For example, rfqHard, = 0.1, physicians are willing to accept, at

most, a 10%reduction in bed days. (Note: WhiIe the fhction qHard,, is defmed as applicable to al1
physicians, it may be replaced with an individualfy determined value.)

1
1

Description
--

A t'raction (O s qH&,

s l), wtiich represents the minimum proportion of cmentiy assigned OR time

6 ( 1 - qHard,-JP, 'represents the minimum acceptable nurnber of OR
minutes that can be assigned to doctor y. For example if qH=GR = 0.2, physicians are wiIling to
accept, at most, a 20% reduction in operating room time. (Note: While the fiaction qHa&, is defmed
as applicable to al1 physicians, it may be replaced with an individually determineci value.)
ansidered acceptable. Thus

--

--

A f'raction (O s @II&-

--

s 1), which represents the minimum proportion of currently assigned direct

cost dollars considered acceptable. Thus (1 - qHarmP,' represents the minimum acceptable
number of direct cost dollars that can be assigned to doctor y. For example, if qHard,,
= 0.3,
physicians are assumeci wiliing to accept, at most, a 30% reduction in direct cost dollars. (Note: Whle
the kaction &is definexi as applicable to d l physicians, it may be replaced with an individually
d e t h e d value.)

zy

Deviation
Variables

d i , d;

Description

1 Deviations from the objective of achieving a pre-defined level of profitability

1 volume of cases above the institutional break even point.

at a

Deviations fkom the objective of achieving a pre-defined level of profitability at a

volume of cases below the institutional revenue cap (or global budget).
dB;, d,.

1 Deviations h m the objective of achieving doctor y's preferred billing.

g
y does not produce more CMG n's
Deviations 60rnthe objective of e n s u ~ doctor
than hislher preference.
Deviations from the objective of ensunng doctor y does not produce fewer CMG
ri's than hisher soft lower bound.
--

Deviations from the objective of ensuring doctor y receives his or her minimum
number of bed days.
Deviations From the objective of ensuring doctor y receives his or her minimum
number of operating room minutes.
Deviations from the objective of ensuring doctor y receives his or her minimum
nurnber of direct cost dollars as implied by minimum bounds on direct cost.
-

--

---

Deviations from the objective of ensuring doctor y receives his or her minimum
number of direct cost dollars as implied by maximum bounds on contribution.

Weight

Description

w;, w,+

Penalty assigned to deviations 6om the objective of achieving a pre-defined level of

profitability by producing a volume of cases resulting in revenues greater than the
institutional break even point.

w,+

Penalty assigned to deviations fkom the objective of achieving a pre-defined level of

profitability by producing a volume of cases resulting in total costs less than the
institution's revenue cap.

WC,

I
I
Wcc

Penalty assigned to deviations fiom the objective of achieving a physician's preferred

Ievel of billings.
Penalty assigned to srnail deviations (within soft bounds) corn a provider's preferred
rnix of cases.
Penalty assigned to large deviations (outside sofl bounds) fiom a provider's
preferred mix of cases.
Description

Preemptive
Weighting

pi

Preemptive weight assigned to goal one - econornic targets.

pz

Preemptive weight assigned to goal two - minimum preferred resources.

PI

1 Preemptive weight assigned to goal three - case mix targets.

5.1.1 Volume Mode1 Goals

The volume mode1 is a linear GP mode1 with three pre-emptive goals. Under goal P,, the
program first identifies a mix and volume of cases for providers 1 through Y that is economically
viable for both the hospital and its associated providers. This is achieved through constraints
associated with P, that ensure a case mix generating sufficient contribution (revenue - direct cost)
to recoup the fixed costs of production and achieve a desired level of profitability, without
exceeding the institution's global budget, is selected. Associated constraints also ensure the case
mix selected under goal P, provides physicians with a preferred level of income.

Preemptive goai P, operates by minimizing deviation from desired economic goals. (w,'d;
+ wF*dF-)in this equation is the penalty associated with deviations fiorn the objective of achieving

a defined level of profitability by producing a volume of cases sufficiently large to recoup the fixed
costs of production. d i , negative deviations from the objective of achieving profitability, are
heavily penalized in the model (ie w; O), while d,', positive deviations from the objective of
profitability, are not (ie w,-

= O).

This weighting scheme is used to reflect the fact that profit

underruns are considered to be much less desirable than are profit overruns.
(wE-d+ wE+dEb)
is the penalty associated with deviations from the objective of ensunng

profitability without exceeding the institution's global budget. In applications of the model,
positive deviations from the goal of remaining within budget, d,- are heavily penalized (ie w,' >>
O), while negative deviations from the goal of remaining within budget are not (ie w; = O). This

weighting reflects the gravity with which budget overruns are viewed in cornpanson to budget
underages.
w B ( Iy (dB; + dB; ) is the penalty associated with deviations from the objective of
ensuring each physician achieves a desired amount of billings. Both positive and negative
deviations from a physician's billing targets are assumed to be undesirable and are therefore

minimized by the model. Negative deviations fiom target billings are assumed to represent a loss
of income for physicians; positive deviations represent lost recreation. Since both are undesirable,

positive and negative deviations fiom billings are weighted equally.

53

Preemptive goal P, is designed to ensure, as much as is possible, al1 physicians receive a

minimum amount of physical resources such as bed days, operating room minutes, and direct cost
dollars. Minimum levels of physical resources are determined by surnrning the produa of the

minimum price (or length of aay or operating room time) a physician is wiUing to accept for a
particular seMce by his or her preferred number of cases. These constraints are placed in the
model to ensure any selected case mix can be converted to commensurate changes to physician
practice in the second phase of the model.

Preemptive goal PZis a model induced objective, rather than an objective translatable to a
specific organizational goal within the hospital. Wherever possibie, the volume model retums a
solution that can be translated into a feasible set of practice change recommendations. (The
translation of cost model recomrnendations into a set of cornmensurate practice goals is described
in Section 5.2 - The Cost Model).

Preemptive goal P, identifies, h m arnong the many that could be selected under
preemptive goals P, and P,, the particular case mix that comes closest to achieving the desired
case rnix of providers. The "goodness" of a case mix is determined by a two-part, piece-wise
linear penalty function that attaches small penalties to modest deviations from a target and
significantly larger penalties to more substantial deviations:

54

In the above equation w, (d& + dm') represents the penalty associated with smali
deviations from a desired volume for a particula. doctor and CMG. In modeliing terms, w, (d,'
+ d*,

) is the penalty for deviations between a target level and the "soft bound" for a given

physician and CMG. w,

(d+=& + d,'*'

) is the penalty for more signifiant deviations fi-om a

desired target for a particular physician/CMG. The penalty function defined by P, is graphically
depicted in Figure 5-3.
5.1.2 Profitability

Constraints 1 and 2 represent restrictions on case mix arising from requirements for
profitability. At a findamental level, every organization, regardless of size, sector, or ownership
status, must achieve profitability; on the whole, revenues must meet or exceed the costs of
producing a given set of goods or services, if the organization is to remain solvent. Constraint 1
defines a lower bound on case volume necessary to achieve institutional profitability: The case

rnix returned by the model should, if possible, generate revenues

a,1,r,

cost of production (F) plus the variable cost of production ( ~ , ~

v, ,

equal to the fixed

plus any requirements

for profits or allowances for loss (Profit).

Negative deviations from the objective of achieving a defined level of productivity are
heavily penalized via preemptive goal P,; positive deviations are not. Thus, Constraint 1 functions
like a standard LP greater than or equal to constraint. Therefore, while the volume model may
recommend any case mix, those case mixes that do not generate at Ieast enough contribution
(revenue - direct cost) to offset the fixed costs of production are heavily penalized.

5.1.3 Global Budget I Revenue Cap

Constraint 2, conversely, d e h e s an upper bound on case volume derived from the dual
requirements that the hospital generate a profit and yet remaining within the revenue cap assigned
by the province.

2)

C C

vpx,

d i - d i = G - Profil

n = I y = l

In general, the total cost of any case mix selected by the volume model should not exceed

the revenue cap granted by the province. Thus, the total cost of production (F +

x&

should be less than or equal to the global budget. Moreover, since institutional profitability is, for
volumes greater than break even, the difference between the provincial revenue cap and the total
cost of production, the total cost of production should be at Ieast $Profit less than the global
budget. Thus, Constraint 2 sets a target for the volume model, which States that the totai cost of
the case rnix selected should amount to no more than the global budget, less desired profit. In
preemptive goal PI,negative deviations ( d a fkorn this goal are not penalized (ie w; is set to
zero.) Positive deviations (d,') , however, are penalized (ie w,- > O). Thus, Constraint 2
functions in a manner similar to a standard LP Iess than or equal to constraint. Constraint 2. while
allowing any case mix to be selected, penalizes those mixes in which the total cost of production
is greater than G-Profit.

Figure 5.2 Volume model profitability constraint space.

Revenue
r*x

Totd Cod
F+ v ' x

M d t y due to d i

Cas Volume

Mdty due to d;

(x)

Figure 5.2 illustrates the profitability constraint space for a simplified example in which the
hospital and its associated medical staff produce only one product. In the figure x represents the
case volume that may be selected by the model. The total cost of production, as shown in the
figure. is (F + vx). Total revenue, in this example, is equal to rx at values of x where ni s G. At
values of x where rx > G the total revenue accruable to the hospitai is limited to G. If profit is
ignored and break even requirements are assumed. the preferred solution region for the volume is
the triangular area BCD. However, when a desired profit is specified, the preferred region
changes ffom the tnangular area BCD to the polygon H'H"CIN1'.

57

Thus, the preferred solution space, when profit is specified, is bounded by Constraint 1 on
the left and Constraint 2 on the right. Constraint 1, defines the line segment passing through

H'H". In the single produa example Constraint 1 is:

1)

(r -

V)X +

di

di = F

Profit

Or
r x - Profit

di - di = F

vx

Thus, the line segment passing through H'H" is the minimum value of x that may be
selected without violating preferences for profitability. Mathematically, H'H" is a vertical line
passing through the x-axis at the point where (r-v)x = F+Profit. Thus, H'H" is the point at which
total contribution is equal to the fixed costs of production plus any required profit. Altemately,
H'H" may be thought of as the point at which total revenue exceeds the total cost of production

by the required amount of profit. The variables, d i and d,*, in Constraint 1 are slacks that allow
the line segment WH"to deviate both to the left and right of the point defined by (r - v)x = F +
Profit. However shifts to the left, which imply positive values of d i , are heavily penalized, while
shifls to the right, which imply positive values of d,' are not penalized. Thus, values of x between
O and (r-v)x = F+Profit will not generally be selected by the volume model. H'H" may therefore

be thought of as a lower bound on the preferred solution space for the model, restricting the value
of x selected, in most instances, to be greater than or equal to the point defined by (r-v)x =
F+Profit.

Of course, institutional revenues are capped by the province. At larger values of x, as

shown in Figure 5.2, institutional profit declines. There is, therefore, an upper limit on the value
of x that rnay be selected. In the single product example, Conaraint 2 is:

2)

+ vx +

d i - d;

G - Profil

This constraint defines the ne segment passing through 1'1". Constraint 2, therefore, is
the maximum value of x that can be selected by the volume model without violating preferences
for profitability. Constraint 2, as shown in Figure 5.2, is a vertical line passing through the x-axis

at the point where the total costs of production are less than the global budget by an arnount equal
to the required profit. The variables d,' and dETare slacks that allow the line segment 1'1' to
move to the lefi and right of the point where vx + F = G - Profit. Shifis of the line to the left,

which imply positive values of d i are not penalized in preemptive goal P , Shifts to the nght,
which irnply a positive value of d,- are heavily penalized in P,. Therefore, Constraint 2 hnctions
as a bound on the preferred solution space for the model, restricting the value of x selected, in
most instances, to be less than or equal to the point defined by vx + F = G - Profit.
5.1.4 Physician Billings

Constraints 3 through 3a define a relationship between case rnix and physician billings.
Constraint 3 is a standard goal programming formulation stating that the sum of each physician's
billings, over al1 cases,

(zy1,b_

) should be as close as possible to a stated preference b-".

59

Recogniting, however, that changes to case mk, and hence physician billings, may be
necessary for the econornic stability of the hospital, Constraints 3a and 3b define hard upper and
lower bounds to lirnit the magnitude of change to any individual physician's billings.

Deviations fiom the physician biliing targets are minimized via goal P,
5.1.5 Implied Bounds on Physical Resources

The volume model is one cornponent of a two phase algorithm to set volume a d o r prices
for seMces in an acute care hospital. The algorithm is, in essence, a methodology for distributing

a set of physical resources to a group of doctors. If available resources are insufficient to meet
the demands of surgeons, or if physicians' preferred use of resources creates an econornically
infeasible solution for the hospital, the algorithm redistributes resources. Because total resources
consumed are a function of the number of products or seMces produced in a hospital and the cost
or resources consumed to deiiver them, the algorithm makes recornmendations regarding both
case rnix and case cost.

In the volume model, product prices are assumed to be fixed and volume is allowed to
Vary; the cost model (described in Section 5.2) assumes volume is fixed, but allows price to Vary.

The two phase approach, enables the model to deliver separate, but equivaient, solutions to the
resource allocation problem. The first, generated by the volume model, describes the "best"
volume and mix of cases, if physicians make no change to their practice. The second, generated

by the cost model, describes the changes necessary to physician practice to achieve the same

results as those identified by the volume model.

While the volume model can operate independently of the cost model, information
contained in the cost model c m be used to set lower bounds on case mix in the volume model.
For example, bounds in the cost model stating that a given physician wiii accept no more than a
10% reduction in length of day for his or her cases, represents a defacto constraint on the

minimum number of cases that may be assigned in the volume model. Since the cost model
translates the volume model into practice recommendations necessary for physicians to achieve
their preferred case rnix within the confines of available resources, a minimum number of cases
rnay be inferred for each physician as follows:

w,LOS,

44

( 1 - qHmdLm)LOS,

diedsv- diedSY
=

for each y

n=l

n=1

Constraint 4a States that the total number of bed days allocated to doctor y (1, L O S 3
should equal the sum of the product of the minimum acceptable length of stay (1 - qHardLos)
LOS, and the preferred volume (p,')

for each case. Preemptive goal P,, in which only the

negative deviations from bed day targets (&,*)

are penalized, makes constraint 4a function like a

greater than or equal to constraint in a standard linear program. Thus, constraint 4a can be
interpreted as a flexible bound on the minimum number of cases that may be recomrnended in the
volume model. A positive value for d-B,

will be returned oniy in instances where physicians

have defined practice constraints that are so restrictive that no solution can be found to the
volume model, without violating bounds in the cost model.

Lower bounds on cases may also be inferred fiom minimum acceptable bounds on
operating room case time:

xy.oRp

4b)

d~R;-

n=1

1 (1 - @rdOR) ORlmp i

for each y

n =1

or direct cost:

In addition to soft bounds on the minimum number of cases that must be produced, a
maximum number of cases may also be inferred for each physician from bounds on contribution:

Constraint 4d is a soR bound that restricts volume model case selection so that the total

contribution assigned to a physician does not exceed the ability of the physician to produce it fiom

his or her preferred mix.[r, - (1 - qHard,-JvJ,

the maximum difference between institutional

revenue and the physician's minimum direct case cost, is by definition, also the maximum
contribution that can be generated fiom each CMG. By summing the maximum contribution per
case, a maximum contribution cm be determined for each physician. Any case mix selected by the
volume model, therefore, must not result in a contribution exceeding this maximum. Because

* , ,only
,,,d
constraint.

is penaiized in P,, Constraint 4d fnctions like a standard less than or equal to

62

Constraints 4a through 4d strengthen the practical appiicability of the resource allocation

algorithm by generaily assuring al1 recornmendations suggested by the volume model c m be
translated into cost model recommendations.
5.1.6 Case Mix Targets

Constraints 5 through 5c define objectives for case mix and volume for each valid
combination of physician and CMG. constra.int 5 is a standard goal prograrnming objective
setting the target volume for each physician/CMG combination ( q,,) to a predefined value (p,').

for all y; n

In practice, p,,,' might be specifically set for a given physicianKMG or it may be defaulted
to a histoncal value (ie. the previous year's volume).
Constraints 5a and Sb are complex constraints stating that if any deviations fiom a target
volume of cases for a particular physician1CMG combination are necessary, they should fa11 within
the range defined by Power,

P"fd

UPper,,, P"f-d].

In Constra.int 5, the variables dm- and dm'

define the negative and positive deviation from the target value of each physicianKMG
combination. Deviations from target case volumes are penalized in P, by a weight of w,. Since

P, rninimizes deviations, the mode1 d l , if possible, always assign the value of the preferred
volume (p,')

to each 5.However, in circumstances where the target case volume cannot be

achieved, the model will select an altemate value for 5.

It is well known within the field of mathematical optimization that a senes of m linear
inequality constraints comprising n variables defines an n-dimensional, convex, region in space.
Furthemore, it is also known that the maximum or minimum of any linear function (if it exists)

63

defined over a convex region occurs only at an extreme point of the region. Thus, if n > m, at
least n-m constraints active at the extreme points are derived fiom non-negativity constraints (ie x

r 0) and are thus satisfied as exact equalities at an extrerne point (Ignino 1994). Thus, solutions
to linear prograrnrning problems occur at the intersection of constraints; the f o m of the solution
is always n-m non-basic variables (eg x, = 0)and m basic variables (eg % r O). In practical terrns,
solutions to linear programming problems occur at hyperplanes defhed by two or more
constraints. In the instant model, when a target case volume cannot be achieved, the model wiil
not select an arbitrary value for G.Instead, the value retumed will be at the bounds defined by
two or more constraints. Thus, we would expect in circurnstances where a target volume cannot
be achieved that the model will typically return a value for

bower,

, Upper,

PI=C~~ZTC~

P"f"d],

at either end of the range

rather than somewhere in the middle. (Mid-range values for case

volume occur only when another constraint, for exarnple operating room time or bed days, is
active within the range defined by power,

Rd'""d

U P P ~P"fmd]).
~,~

The fact that linear programming solutions occur only at extreme points makes the
appropriate selection of upper and lower bounds for case volumes critical, and is, in fact the
prirnary reason that rnany single dimensional models for case rnix selection in acute care
institutions tend to be impractical. If changes to case volume are constrained only by aggregate
physical resources such as bed days or operating room time, case mix models tend to suggest
large changes to a relatively small number of products. A common recommendation of such
models is to abandon unprofitable product lines and to entireiy convert the productive capability
of the institution to the delivery of more profitable products. See, for exarnple, Hughes and
Sol iman ( 1989). While mathematically optimal, such recommendations are not implementable
and are thus of little practicai value (Ignizio 1994).

Unlike most fims, hospitals are not completely within the control of their managers;
Hams ( 1977) describes the hospitai as a dual organization consisting of a formal management
structure and a medical hierarchy that, because of the ethical responsibility of physicians to act as
agents for their patients, cornpete for influence and control in an "environment of subtle intrigue".
Gloubeman and Mintzberg (1994) are even less optimistic. A hospital, they argue is not two
organizations, but four: management, nursing and d i e d care givers, medical staff, and directors,
each of whom have their own goais and objectives. Because of the separation of authority,
responsibility, and affinity within the "four faces of health care", physicians they argue are
essentially unmanageable. Since management dictate does not influence physician practice,
hospital administrators are left with the ability to manage only what they cm actually control.
Thus, they tend to manage the physical resources necessary for physicians to care for their
patients, as a proxy for controlling physician behaviour.
As a result of the dual nature of management in health care organizations, standard

product mix models denved from industrial experience are inappropriate for case mix selection;
managers do not have the mechanisms at hand to easily control product mix. Hughes and
Soliman (1989) suggest that physicians be added or removed from the medical staff30 control
case mix. While adding physicians may be a simple policy to implement, removing physicians is,
in practice, difficult. Clearly then, any recommendations for changes to case rnix must be jointly
acceptable to both a hospital's management and its associated medical staff. Since
recommendations for large scale change to case rnk are not likely to be greeted with enthusiasm
by physicians, who are the ultimate arbiters of demand, their implementability is somewhat
suspect. In response, a number of authors use bounds to lirnit the magnitude of volume change
suggested by their models. See Robbins and Tuntiwongpiboon (1989) for exarnple. This makes

65

the selection of bounds extremely critical, since in many instances, model suggestions for change

will correspond to either the minimum or maximum case bounds. Unfortunately case bounds are
generally set on an ad hoc basis (see Dowling 1976). Furthemore, if the case bounds tmly
represent a division between an acceptable and an unacceptable case volume, then some (or dl)
case recommendations returned by a single criteria case mix model are barely acceptable.

In this model, the issue of change to physician practice is approached fiom the standpoint
that large scaie dismptions are generally unacceptable; a solution returning recornmendations for
srnall changes to a larger number of physician1CMG combinations is preferable to a solution
recornmending wholesale revisions. In ternis of penalties, we can picture this as a soup bowlshaped function with small penalties in the region near the target and steeper penalties in regions
further from the target. See Figure 5.3.

67
15 and the lower prefe&zd value for this physicianlCMG combination is 10, the negative volume

variation is assigned an objective of 5. d,-&

, therefore, has a positive value whenever dm- is less

than 5 ( >5, is greater than 10). Sirnilarly, d+& has a positive value whenever dm- is greater
than 5 ( x , ~is less than 10). However, in preemptive goal P,, only d C m - is penalized. Thus,
whenever ) ~ mis in the range [O, 10) the total penalty is w,d-range [IO, 15) the total penalty is wcd,'.

+ w,d',-&

When x, is in the

At ]hi = 10, the total penalty is O. At values of

above 15, the total penalty fiom constraint 5b is zero. At values of

above 10, however,

Constraints 5 and 5a contribute toward the penalty function in a manner analogous to the
discussion for Constraint 5b. ~ o w e r , ~ ~ - *and upperWRfd are therefore referred to as "sofi

bounds" on the value of 5;although representing a valid preference, they may be violated by the
model if necessary.
In addition to Constraints Sa and 5b that define soft bounds on case volume, the bounds
given by:
-4bmiute

Lower,

A bsoLure

for uii y; n

represent absolute limits on the value of x, that may be assigned by the mode1 for any particular
physician/CMG combination. These limits are referred to as hard bounds, since they cannot be
violated.
5.1.7 Operating Room Time

Constraint 6 is a technology constraint lirniting the case mix selected by the model such
that the total operating room time consumed by dl physicians (
equai to the total availabie OR time.

xy1,O R ,

x,, ) is less than or

5.1.8 Inpatient Beds

Cmstra.int 7 lits the case mVt selected by the mode1 so that the total number of bed days
consumed (

x,

LOS, 5 ) is less than or equal to the total available bed days.

5.2 The Cost Mode1

The output of the volume model X = (x,,, x,, ..., x,,, x,

..., )x,

is an array of assigned

case volumes that can be decomposed into physician-specinc case vecton (Eg.
)

= (%i,

>43

....

Using this information, it is possible to infer an allocation of institutional resources for dl

Y physicians in the model. Specifcally, we may infer a total assignrnent of hospitai revenue,
direct cost. contribution to overhead recovery, operating room tirne and inpatient beds through
the following equations:

(5.1 1)

The allocation of resources detennined above is based on the case load selected through
the goals and pnonty structures in the volume model. While the mix of cases retumed by the
volume mode1 may be "best" from the perspective of the hospital and physicians as a group, it
may not suit the particular desires of individual physicians. Since physicians, the arbiters of
demand in the health care process are independent agents, there is little practical point in
assigning case mixes to physicians that they do not value; such assignments will simply be
ignored. It is, however, possible to reconcile financial objectives with physician autonomy,
through an algorithm that identifies the minimum case "cost" reductions required for a physician

to achieve his or her desired case rnx, wMe remaining within the total envelope of resources

granted by the volume model. (Cost here is used in a broad sense and is understood to include
both case costs and technological resource consumption.) An algorithm that identifies changes to
practice necessary for physicians to achieve their preferred mix of cases without violating the
economic results of the volume model solution has a number of practical benefits. It allows both
the hospital and its associated medicai stafFto jortly achieve their economic and clinical goals. In
addition, the aigorithm provides a choice for physician decision makers: They may either elect to
change their case rnix or their practice. This mitigates the onerous task of dictating to physicians
a case mix; instead, the alternative view of resource allocation facilitates a negotiation process

between the hospital and its medical staff regarding the number, type, and prices of senices that
will be delivered.
5.2.1 The Cost Mode1 Algorithm

We can infer preferred resource allocations by multiplying resource consumption

coefficients listed in the volume model by physicians' preferred case mix:

r ~ $

Direct COSP.=

vynp&

Revenue''

If the resource allocation determined by the volume model meets or exceeds a physician's
preferred allocation, then we may surmise that the physician is completely satisfied by the assigned
case mix and does not need to change any of his or her resource consumption parameters. It
should be noted that satisfaction requires the following condition to hold tme:

(~evemre&,,,,

Revenue Y' ) and ( C~ntribution~~,,,

s Contributiony' )

and (Direct CostA,,,

and

Direct C o s P )

s ORY')and ( ~ e a S $ ~ ~ ,2,Be&"* )

Those physicians whose resource allocation preferences are met by the case mix assigned
by the volume model cm be eliminated fiom consideration for cost reductions. Physicians whose
preferences are not satisfied by the volume model must consider changes to their clinical practice
if they are to achieve their desired case mix. The cost model is an algorithm that determines, for
al1 unsatisfied physicians, the minimum change in length of stay, operating room use, and variable
case cost necessary to satisQ case rnix and volume preferences. This is accomplished through a
second goal program in which case volumes are held fixed while costs are allowed to Vary. To

promote implementable solutions, costs are allowed to deviate relatively freely between their
prescnbed levels and a user deterrnined soft lower bound. Deviations in cost beyond the soft
lower bound, though possible, are heavily penalized. The cost algorithm is as follows:
Stage I : Calculate Resource Allocations & Eliminate Completely Satisfied Physicians
1) For al1 Y physicians, determine satisfied/dissatisfiedstatus as outlined in Equation
5.13. Designate set S to be the set of al1 physicians satisfied by the resource

allocation determined by the volume model. Designate set D to be the set of

physicians not satisfied by the resource allocation detemined by the volume
model.
iii) Update the resource availabilities for physicians in set D by calculating the
following:

Direct Cost,,,,

Direct Cost Y *
fis

Stage 2: Determine if LOS Reductiom are Required

1)

Determine if bed availability constrains preference achievement for physicians in D.

Calculate:

(5.15)
/

L d S

= Be4,fabk

C &LJkY'

FD

IfVB, r O, available beds are greater than physician preferences. Proceed to

Stage 3. If V,,, < O. available beds are insufficient to meet the preferred demands
of al1 physicians in D.

ii)

Determine the set of al1 physicians in D who are satisfied with their allocation of
bed space, by evaluating:

Place al1 physicians who are satisfied with their bed allocation into set SB. Place al1
dissatisfied physicians into set D,.
iii)

Update the availability of bed space for physicians in set D, by calculating the
following:

Note: The variable costs of production associated with nursing, after suggested
reductions in Iength of stay are implemented, are:
/

v p = v

"

ny'

* [LOS, - I,]

(~ursrn~)

LOS,,

* 'P

for al2 y EDB;n

Where:

Variable

Description
The total variable cost associated with the production of CMG n
by doctor y.

v~

~ The
) variable cost of nursing associated with the production of
CMG n by doctor y.

The histoncal length of stay associated with the production of

CMG n by doctor y.
'YI

The mode1 assigned length of stay (5, is a decision variable).

A parameter (O s P a 1) representing the expected proportion of
nursing costs saved with each patient day eliminated.

Stage 3: Determine OR T h e Reductzons Required

1)

Determine if operating room time availability constrains preference achievement

for physicians in D. Calculate:

If V,, 2 O, available OR time is greater than physician preferences. Proceed to

Stage 4. If V
,
, < O, available time is insufficient to meet the preferred demands of
al1 physicians in D.

ii)

Determine the set of aU physicians in D who are satisfied with their allocation of
operating room tirne by evduating:

Place al1 physicians who are satisfied with their OR allocation into set S,. Place dl
dissatisfied physicians into the set Do.
iii)

Update the availability of operating room time for physicians in set Doby
calculating the following:

Stage 4.-Determine Cme C'mi Reductiotzs Repired

1)

Determine the set of ail physicians in D who are satisfied with their allocation of
contribution, direct cost or beds by evaluating:

Contribzitioz,iIOned
5 ContributionY '
And
Direct Costilo,,

Direct Cost-"*

And

Place al1 physicians who are satisfied with their allocation of contribution, direct
cost, or beds into set S,. Place ail dissatisfied physicians into the set DK.Note:
The set DKpurposely includes al1 physicians dissatisfied with their bed allocation,
since changes to length of stay automatically result in changes to direct cost.
iii)

Update the requirements for contribution demanded of physicians in set DKby

calculating the following:

iv)

Update the availability of direct cost for physicians in set DKby calculating the
following:

Direct " = Direct ' Y

Direct CostY
E

SK

S~age5: Identzh Minimzm "Cost" Chmges to Physician Practice

Once physicians who are in some, or dI, senses dissatisfied with the allocation of
resources assigned by the volume mode1 have been identified, it is possible to determine the
minimum practice changes necessary for each to conform to the volume model's resource

allocation. while retaining their preferred mix of cases. By so doing, it is possible to create a
situation in which both the institution and its associated medical stafFare simultaneously able to
achieve their separate goals.
Modifications to physician practice must, however, be considered within the context of
medical stafFcompetition and/or cooperation. Physicians, as we have noted, maintain an am's
length relationship with the institution in which they practice; while they work in a particular
hospital, they do not work for it. Since physicians are independent agents, sharing a set of
common resources, they are to a limited degree, in competition with their fellow doctors for
access to scarce hspital resources (revenue, beds, operating room hours, etc.), that allow them to
care for their patients and, hence, generate income. However pure competition is the exception
rather than the nile among physicians. It is comrnon, particularly in teaching hospitals, for
groups of physicians to enter into cooperative docation arrangements, or group practices. In a

76

group practice, physicians pool both their access to hospital resources and the revenue generated
by access to these resources. Group practices enable physicians to balance incorne over time,
promote education teaching and research objects, and give the group bargainhg power with
other physicians and the institution when lobbying for resources or influence. Physicians, in
essence, surrender autonomy in exchange for stability, continuity, and secunty of access to
resources. Nevertheless, while cooperation is common, competition remains an integral
component of resource allocation decisions in acute care hospitals (Brown 1994).
The degree to which competition and cooperation exist between the medical staff of an
acute care institution plays a role in determining how resources are ultimately allocated. If we
assume cooperation between al1 physicians, we may postulate a mode1 for minimum physician
practice changes as follows:

(5.25) Cost Model: Continues on next page.

Subject to:

for al2 y E DO
for all y

Do

for al1 y

Do : n

;n

( 5 . 2 5 ) Cost Mode1 - Goal prograrnming mode1 for determinhg practice parameters

3,)

1 py:kyn

Direct CostYIn

for all y

D K fly

E DB

n = 1

.v

3a ')

p&,,,
n = l

cr,,

y$
,),

Direct Cost,Gin for al1 y E DK fl y E DB

for a i l y

DK fI y P D,; n

for a l l y E DK fI y
for all y E D,; n

Cn LOS'

for all y E DB;n

qHardLOSLOSps
s
qHardOAORynr op 5 ORy, for all y E Do;n
qHardDl,DirectYn s ,k s vYn for all y E DK;n

Dg; n

Where:
Decision
Variable

Description

Length of stay assigned to the production of CMG n by doctor y.

Operating rmm case time assigneci to the production of CMG n by doctor y.

1
1

Direct case cost assignecl to the production of CMG n by doctor y.

Description
Doctor y's preferred volume of CMG n.
- -

--

--

institutional revenue achieved by the production of CMG n.

Variable institutional costs associated with the production of CMG n by doctor y.

Reductions in direct cost expected fiom decreases in length of stay associated with the
production of CMG n by doctor y.

Description
The total number of bed days available for ail doctors dissatisfied with the ailocation
of bed days determineci by the volume rnodel.
The number of bed days assigned to doctor y in the volume rnodel.
Doctor y's preferred average LOS for patients in CMG n.
A real number (qHadLoss qSoALOss 1) representing a sofi lower bound on
deviations from doctor y's preferred LOS for CMG n. (Note: qSoft,, * LOS, can
be replaced with a value representing an alternatively determined lower bound.)
A real number (O s qHard,,, sqSoft,,, ) representing a hard lower bound on
deviatinns f h m doctor y's preferred LOS for CMG n. (Note: qHard,,, * LOS, c m
also be replaceci with a value representing an altematively detennined bound.)
The total mount of operating room tirne available for al1 doctors dissatisfied with the
allocation of operating room tirne detennined by the volume rnodel.
The arnount of operating room time assigned to doctor y by the volume mode[.
The average case tirne associated with the production of CMG n by doctor y.
A real number (qHar6, s qSo%, r 1) representing a soft lower bound on deviations
fiom doctor y's preferred case time for CMG n. (Note: q S o h R* O R , can be
replaced with a value representing an aiternatively detenriined lower bound.)
A real number (0 s qHi~.&,

qSo&d representing a hard lower bound on

deviations fiom doctor y's preferred case time for CMG n. (Note: qHa& * O R ,
can be replaced with a value representing an aitematively determineci bound.)
5

I
1

- -

RHS
F"
Profit,,,,
Direct costybh

Description

1 The updated fixed cost of production less the contribution of doctors satisfied with
1 the allocation of direct cost determineci in the volume model.

1 The total

--

- --

( k e d - v ~ i a b l e ~ o s t s ~ s i g to
n eal1
d docton in the model.

The minimum of the direct cost assigneci to doctor y by the volume model or the
direct cost inferreci for doctor y by the volume model's allocation of contribution or
allocation of bed days.

A real number ( qHard,,

s qSoftD,, s 1) representing a sofi iower bound on
deviations Grom doctor y's prefemed direct case cost for CMG n. (Note: qSoft,,,
v, cm be replaced with a value representing an altematively determined lower
bound. )

s qHard,,, s 1) representing a hard lower bound on

q H a ~ d ~ A
~ real nurnber (qsoft,,,
deviations fiom doctor y's prefened direct case cost for CMG n. (Note: qHard,,
v, can be replaced with a value representing an altematively determined bound.)
Deviation

Description

Variables
Deviations fiorn the objective of ensuring ail available bai days are allocated to those
doctors who are dissatisfied with the dotment of beds determined by the volume model.
Deviations fiom the objective of ensuring that LOS recommendations returned by the
cost rnodel are as close as possible to doctor y's preference for CMG n.
Deviations from the objective of ensuring that LOS recornrnendations returned by the
cost model are within doctor y's "soft bounds" for CMG n.
Deviations f?om the objective of ensuring that al1 available operating room tirne is
allocated to those doctors who are dissatisfied with the allotment of time determineci by
the volume mdel.
M a t i o n s from the objective of ensuring that case t h e recommendations returned by
the cost rnodel are as close as possible to doctor y's preference for CMG n.

Deviations fiom the objective of ensuring that case tirne recomrnendations retmed by
the cost rnodel are wittiin doctor y's "soft bounds" for CMG n.
Deviations flom the objective of ensuring al1 direct cost dollars are allocated to those
doctors who are dissatisfied with the allotment of contribution determined by the volume
model.
Deviations tom the objective of ensuring that direct case cost recornmendations returned
by the cost rnodel are as close as possible to doctor y's preference for CMG n.
Deviations fiom the objective of e n d g tha direct case cost recommendations returned
by the cost model are within doctor y's "soR bounds" for CMG n.

Weight

Description
Penalty assigned to model recommeudations for LOS that are less than a physician's
stated preferences. (Positive nxommendations for change in LOS are not permitted
and are, therefore, not penalized)
Penalty assigned to model recommendations for LOS that exceed physician's sofi
lower bound. (Note that negative deviations &om soft bounds, implies recommended
values are within the sofl preference range specified by a given physician. They are.
therefore, not penaiized.)
Penalty assigned to model recommendations for case t h e that are less than a
physician's stated preferences. (Positive recommendations for change in case thne
are not permitted and are therefore, not penalized.)
--

- -

Penalty assigned to mode1 recommendations for case t h e that exceed a physician's

sofi Iower bound. (Note that negative deviations fYom sofi bounds, implies
recomrnended values are within the soft preference range specified by a given
physician. They are. therefore, not penalized.)
-

p
p

--

Penalty assigned to model rewmrnendations for direct cost that are less than a
physician's stated preferences. (Positive recommendations for change in direct cost
are not pennitted and are therefore, not penalized.)
Pends. assigned to model recommendations for direct case cost that exceed a
physician's sofl Iower bound. (Note that negative deviations fiom sofi bounds.
irnplies recommended values are within the sofl preference range specified by a given
physician. They are, therefore, not penalized.)

freemptive
Weighting

--

Description

PI,

Preemptive weight assigned to goal I - iden@ a solution consistent with physician

preferences that does not exceed the number of available beds.

PI2

Preernptive weight assigneci to goal 2 - minimize changes to LOS preferences.

Preemptive weight assigned to goal 3 - identify a solution consistent with physician
preferences that does not exceed the number of available OR minutes.

PI

1 Preemptive weight assigned to goal 4 - minimize changes to case tirne preferences.

Preemptive weight assigned to goal 5 - identi@ a solution consistent with physician
preferences that does not exceed the revenue cap.

Px

1 Preemptive weight assigned to goal 6 - minimi:

changes to direct cost preferences.

5.2.2 Cost Mode1 Goals

The cost model is a linear GP model with six preemptive goals. Under goals P,, and Pl?,
the model identifies the change to dissatisfied physicians' pattern of LOS necessary to achieve (as
much as possible) physicians' preferred mk and volume of cases without violating restrictions on
the number of available bed days.

Since dB,- takes on positive values only when the total bed days consumed is greater than
available bed days, Pl,, identifies whether a feasible set of length of stay recommendations can be
identified. P,2 selects, fiom the many possible sets of recomrnendations that can be retumed by
Pl,, the set of recomrnendations that cause the least disturbance to overall physician practice.
Goals P2, and P determine the minimum change to each dissatisfied physician's pattem
of practice necessary to achieve (again, as much as possible) a preferred mUc and volume of cases
without violating restrictions on the amount of operating room time available.

Goals P,, and P3,determine the minimum change to each dissatisfied physician's pattern of
practice necessary to achieve a preferred rnix and volume of cases without violating restrictions
on direct cost or profitability.

5.2.3 Available Beds and Length of Stay

Constraints la through Id represent restrictions on allowable recommendations for LOS.

Constraint l a is a standard goal prograrnrning objective that sets the target length of stay for each
doctorKMG combination so that the total number of bed days consumed by a11 doctors ~,,~,,p,'

l,m, where 5, is the decision variable), in most instances, is less than or equal to the total available
number of bed days.
In Constraint la, the deviationai variables ,d,

d,+

are both included for the purpose of

exposition, which would ordinarily indicate that the objective of P,,would be to rninimize the
deviation from the total number of bed days allocated. However, in P,,only positive deviations
are penalized. Thus Constraint Ia can be read as: "Accept any allocation of length of stay.
However, penalize those allocations that cause more bed days to be consumed than are physically
available." Note that the available number of bed days (Bed<,-,J

in the t-ight hand side of

Constraint l a represents the total supply of unallocated beds f i e r the requirements of al1
physicians in SB have been removed from the global resource availability. (Removing the resource
requirements for physicians in SB fiees resources for physicians in D,. Satisfied physicians may
have been allocated more bed days than they need to produce their desired mix and volume of
cases. Thus any "surplus" beds allocated to these physicians in the volume mode1 can be allocated
to dissatisfied physicians - physicians in D,.)

Constraint lb is a standard linear prograrnming constraint that iimits the recornmended

lengths of stay retumed by the cost model so that the total number of bed days ailocated to each
physician in the cost model (Enp',

I,for each n) is at least as great as the number of bed days

allocated by the volume model. Constraint lb is included to maintain equity between the volume
and the cost model; no physician is penaiized for participating in the cost model, since the

resources assigned by the cost model t o each physician are at least as great as those determined by
the volume model.

l b,

C P ~ S ,r BehL1oned

for

U Iy
~ E

D ~n ;

n = l

Constraint l c is a goal programming objective. It defines objectives for each doctor/CMG

pair that penalize length of stay recommendations diffenng fiom the preferred length of stay.

lc)

I, + ciLi-

ciLi=LOS,,,,

forally

D,; n

Constraint 1d defines objectives for each doctor/CMG pair that penalize model
recommendations for length of stay that exceed a predefined sofi lower bound. Constraint Id is
analogous to the sofi bounds on case volume defined by constra.int 5-Sc of the volume model. In
the cost model, however, only a sofi lower bound is defined for length of stay, since an associated

85

bound constrains the model from recornrnending lengths of stay greater than physicians'
preference.

d~;

LL,

for a l l y E LI,; n

- d * . = qSoff,,LOS,
u,"

Associated Bound:
q~*&OS, s

The right hand side of Constraint Id (qSoft,,

LOS,

LOSd is a user-definable soi3 bound on

length of stay for each doctorICMG combination. As shown above, the soft bound on length of
stay defaults to a proportion of the target value (ie qSoft,

is a real number in the range [O, II).

In practice, soi? bounds rnay be individually specified for a particular physiciadCMG combination
or may be set in relation to an intemal or extemal benchmark.
5.2.4 Operating Room Availability and Case Time

Constraints 2a through 2d represent restrictions on allowable recornmendations for

operating room case time. Constraint 2a is a standard goal programming objective that limits
recommended operating room time for each doctorfCMG combination such that the total amount
of operating room time consumed by al1 doctors

(&En pm0odis less than or equal to the total

arnount of operating room time available.

As in Constraint 1a, deviation variables are defined for operating room availability,

although in practice model recornrnendations typically ensure that the total operating room time
consumed by al1 physicians is less than or equal to the total arnount of operating room time

86

available. By definhg 2a as a goal progranunhg objective rather than a less than or equal to
constraint, a feasible (though undesirable) solution cm be obtained in instances where the sum of
physician preferences for operating room tirne exceeds the total amount of operating room time
available.
Constraint 2b is a standard linear programming constraint that limits the recornrnended
case times retumed by the model so that the total amount of operating room time ailocated to
each physician (

p,'o,

for all y doaors) is at least as great as the amount of operating room

time ailocated by the volume model. It is included to preserve equity between the cost and the

volume model.

Constraint 2c is also a standard goal programming objective. It defines objectives for each
doctor/CMG pair that penalize case time recomrnendations (O, ) that differ from a preferred case
time (O&).
2c)

O,

dip - diF

ORp

for al1 y G Do; n

Constraint 2d defines additional deviation variables for each doctor/CMG pair that
penalize case time recomrnendations that exceed a predefined soft lower bound.

87
As was the case in Constraint Id, the sofi lower bound for case times is assumed to be a

proportion of the target value. This value, however, may be individually specified for a particular
physiciadCMG combination or set to an interna1 or extemal benchmark.
An associated bound:

lirnits mode1 recornrnendations for case tirne to the range of values between [qHard,,OR,,,,

O&,].

5.2.5 Profitability and Direct Cost

Equations 3 through 3c represent constraints on changes to direct case cost. Constraint 3

is a goal programrning objective that Iimits direct cost and length of stay for each doctor/CMG

combination so that the total contribution by al1 doctors

resulting from a reduction in length of stay (

x,

~,,,,~,
p,'(r, - IZ, )) plus the savings

1,p,'(q, - y,&,)

) are greater than or equal

to the fixed costs of production plus the total profit determined by the volume model
(F+Profit,,J.

Constraints restricting direct cost (3 - 3c') are more complex than other constraints in the
volume model. The terrn

(x,&

p,'

(q,- y,

) describes contribution towards the fixed

costs of production coming fiom physicians who are dissatisfied with their allocation of bed days

via length of stay reduction. Physicians who are dissatisfied with their allocation of beds are, by

88

default, dissatisfied with their ailocation of direct coa, since reduction in length of stay implies a
change in the direct cost of a case. Thus, physicians dissatisfied with their bed docation are
automatically dissatisfied with their docation of direct cost, since fewer direct cost dollars are
required whenever length of stay decreases. Nevertheless, rnonies freed up through length of stay
reductions are available to arneliorate requirements for other reductions in direct coa that may
have been mandated by the model,
Interestingly, the exact nature of change in direct cost arising from marginal changes to
length of stay are not particularly well understood; Mount Sinai, like al1 hospitals in Ontario, has
no data to descnbe exactly how costs change with length of stay. For instance, if an average
patient in a particular CMG receives seven x-rays during an average seven day length of stay. will
a reduction in LOS of one day result in one fewer x-ray being delivered? Although it rnight seem

reasonable to assume that one fewer x-ray rnight be delivered, an argument can be advanced that
seven x-rays constitutes the standard of care. Several care givers at the study site expressed the
opinion that as lengths of stays are shortened, the intensity of nursing care increases; the same
amount of care, they argue, is delivered, only the period in which it is delivered changes. In light

of this controversy, we have chosen to adopt a consewative estimate of the impact of length of
stay changes on direct cost. Specifically, we assume that all direct clinical expenditures
(pharmacy, laboratories, diagnostic imaging and nutrition), with the exception of nursing,
constitute a standard of care; without specific initiatives to modi@ these components of care, they
remain part of the cost of delivering a particular CMG. Only direct nursing costs are assumed to
Vary with length of stay. Again, the exact rnanner in which direct nursing cost decreases with
length of stay is not known and cannot be presently obtained from data maintained by Mount
Sinai. However, if we assume that some proportion, Y,of the average daily nursing cost is saved

89

with each day LOS is reduced, we may estimate the reduction in direct case cost with marginal
change to length of stay as:

(5 -26)
Reduction in Direct Cost =

If q,= TvW,-

(lvun,ng)

LOSp

Y * [LOS,

- iyn]

and y, = IfvWo'riinmS1
/ LOS, equation 5.26 may be written as:

( 5 -27)

Reduction in Direct Cost = a,

-,l,y

Constraint 3, therefore, sets an objective for profitability such that recommendations for
change to length of stay when combined with recornmendations for changes to direct cost is equal
to the reduced fixed costs of production plus any requirement for profit.
Constraints 3a and 3a' enforce equity between direct recomrnendations returned by the
cost model and the volume model. In general, constraints 3a and 3a' are designed to ensure each
physician is assigned as many direct cost dollars in the cost model as was assigned by the volume
model. Nevertheless, an equity constraint based on assigned direct cost dollars is not applicable in

al1 circumstances. Recall fiom 5.22 that physicians are considered dissatisfied with their
allocation of contribution (eg belong to the set Dd if they receive fewer direct cost dollars than
desired, or if they are required to generate additional contribution, or if they are displeased with
their allocation of bed days. Equity constraints related to assigned direct cost dollars are
applicable only when the direct cost dollars implied by the ussigned contribution and arsigned

reductions in beddays are greater than the direct cost dollars assigned by the volume mode!. If
irnplied direct cost dollars, determined by subtracting the volume model assigned contribution

90

fiom the preferred revenue (Revenu& - C o n t n b u t i ~ n ~ ~ are

~ &less
, than the direct cost dollars
assigned to a physician by the volume model (Direct CostYMow),the physician rnay be required to
reduce his or her direct cost expenditures to generate required contribution. SimiIarly, $a
physician is required to reduce his or her length of stay to accommodate model recomrnendations
for reduced bed days, direct cost savings (p,,,'(a,,, - yyn&,)) may imply fewer direct cost dollars
than were assigned by the volume model.

p i k,,

3a)

Direct Costi,,,

for aZl y

DK n y e D,;n

for aZIy

D,

n = l

ny

D,;n

Where:
Direct

= Min

Direct Cost,;,,, = Min

Direct C ~ s tY, , , ~ ,,

V Y D,
@

Re venuepiefimed - Contri bu!ion,iIDned

Y
Y
Re venuePrefimd
- Cotltributzon,,,,,

ify

N
Y
CostPreferred

P-;(

DB

n=l

Constraint 3a States that the direct cost dollars assigned to a physician by the cost model
must be greater than the minimum of the direct cost assigned by the volume model or the direct
cost inferred by the volume model's allocation of contribution. Constraint 3a' extends the equity
constraint defined in 3a to cover instances where the physician is dissatisfied with his or her
allocation of bed days and must, therefore, reduce length of stay. The direct cost assigned by the
cost mode1 must be greater than or equal to the minimum of the direct cost assigned by the

volume model (Direct CostYa,,&, the direct cost Uiferred by the contribution allotted by the
volume model (Revenue*,,

- ContributionyMo&)or the direct cost implied by the bed days

assigned by the volume model (Direct CostY,

- 1,p,'(q,

- y,&,).

To illustrate how Constraints 3a and 3a' function, consider the following examples.
Suppose a physician with the following preferences:

Table 5.2.1 Example p hysician preference

Physician Practice Parameter

Calculation

Preferred Value

Revenue
Variable Cost

InpWo (rn- v,d

Contribution

$10

has been alloted the following resources by the volume model:

Table 5.2.2 Exarnple volume mode1 aIlocation (oversupply)

Physician Practice Parameter

1
1
1

Revenue
Variable Cost
Contribution

Preferred Value

1
1
1

VoIume Mode1 Allocation

1
$90 1

$100

$10

1
$120 1
$150

$30

Constraint 3a and 3af, equity constraints, are designed to ensure physicians receive as
many, or more, resources under the cost model as were allocated by the volume model.
However, since the volume model in this example aliotted more direct cost to the physician than
he or she requires to support his or her preferred case mix, there is no reason to set the minimum
direct cost to $120. Such an allocation would imply that the physician, in order to produce his or
her prefemed case mix, wouid have to increase direct cost to consume the additional direct cost
dollars assigned by the volume model. Clearly, this is nonsensical. Therefore, an alternate lower
bound on the amount of direct cost dollars rnust be identified.

92
An obvious, but incorrect, rule for assigning a lower bound is to select the minimum of the

direct cost assigned by the volume rnodel or the physician's preferred direct cost; in this example,
$90. This bound is incorrect, since it ignores requirements for contribution assigned to the
physician by the volume model. In the example problem, the physician was assigned a
contribution of $30 dollars at the conclusion of the volume model. Ifthe lower bound on direct
cost is set to the physician's preferred direct cost, the total contribution generated by the example
physician wiil be less than or equal to $100 - $90 = $10. Since a minimum contribution of $30 is
required from this physician, the lower bound on direct cost must be equd to the minimum of the
direct cost allotted to the physician, or the physician's preferred revenue less the contribution
assigned by the volume model. In this case, the minimum direct cost would be the minimum of
($120, [$1 O0 - $3 O = $701).
The above determination of the lower bound on direct cost, however, ignores the impact
of length of stay reductions. Assume that changes in length of stay resulting from the volume
model's allocation of bed days to the example physician results in an automatic decrease in case
cost of $25. In this instance, the exarnple physician's preferred case cost is no longer $90, but
$90 - $25 = $65; with no other changes to practice the most the example physician will spend to

produce his or her preferred case rnix is $65. Thus, assigning a lower bound of $70 for direct
cost is nonsensical, since it implies that the physician must raise the direct cost of his or her
procedures. Therefore the minimum direct cost that can be assigned to this physician is $65.
As further illustration, assume that the example physician descnbed above has been
assigned the following resource profile by the volume model:

Table 5.2.3 Example volume mode1 ailocation (undersupply)

Preferred Value

Physician Practice Parameter

Revenue

Volume Mode1 Ailocation

$100

%50

$10

$5

Variable Cost

Contribution

Assume also, the physician has been forced to reduce his or her consumption of bed days, which
results in an automatic reduction in direct cost of $25. ln this example, to ensure equity between
the volume and cost model, the Iower bound on direct cost that c m be assigned via the cost
model is $45; the sum dlotted by the volume model. The difference between the physician's
preferred revenue and assigned contribution is $100 - $5 = $95. The maximum direct cost that
the physician can possibly incur, after bed day reductions are considered is $90 - $25 = $65.
However, both of these amounts are greater than the direct cost ailotted by the volume model.

Therefore, the minimum amount that c m be guaranteed to the physician in the second example is
$45.

Constraints 3b and 3b' are standard goal prograrnming objectives that state. where
possible, the cost model should assign direct cost

(u,
less any anticipated direct cost savings

attnbutable to a reduced length of stay (-(rr,- y,$,)),

the preferred case cost ( v , ).

for each physicianKMG combination to

36) ,k

+ d i - d& -

for ail y E D,

V~

P'

3b') kyn -

% + y,,,,$,, + dip - dip = vyn

for al1 y

y e D,;n

DK n y

D,;n

Constra.int 3c defines additional deviation variables for each doctor/CMG pair that
penalize direct case cost recomrnendations exceeding the predefined soft lower bound (qSoft,,
Direct).
3c) diF + d -

- d'

/or al1 y

qSoJD,,vr

D, : n

F m

Associated Bound:

qHardD~m$irect,

km r vp

As shown in 3c. the soft bound for direct case cost defaults to a proportion of the target

direct case cost (qSoft,,

v, ). However, the default vaiue may be replaced with a specific value

for a particular doctor and CMG or it may be set in relation to an intemal or extemal benchmark.
5.2.6 Total Cost

Constraint 4 limits length of stay and direct cost recomrnendations made by the volume
mode1 so that the total cost of production for the institution results in a profit equal to that
allotted by the volume model. Constraint 4 limits total cost such that the fixed cosrs of
production (F) plus the variable costs of production due to physicians dissatisfied of direct cost

(&Sn pp*km),less any anticipated savings due to reduced length of stay, ( &,
yQ
,)

p,*(<5,

are less than the global budget (G) minus the profit allotted by the volume model

(Profi~*omJ-

Constraint 4 is included in the cost model to ensure the total cost of production retumed

by the cost model does not exceed the total cost of production determined by the volume model.
5.3 A Numerical Example

To illustrate how the volume and cost models function, the following exarnple is
provided. Assume a hospital consisting of four surgeons: Doctors A, B, C,and D. Doaors A

and D are orthopaedic surgeons, while B and C are General Surgeons. The four physicians have
the following clinical and financial practice patterns.
Table 5.3.1: Exarn~IePhvsician Profile
Dr.

Dwcription

Volume

Direct
Cost

Amputation

MultipleJoint

Gastrostomy 8
Colostamy
Mjr lnfednal8
Recta1Rucedure
Gastrosbrny 8
Colostomy
Mj Intestinal 8
Rectal Procedure
Mjr C M

Procedures
Muiple Joint
Reofacement

26.4

672
28.8
100.8

2.4

24

iis exarnple are fictitious.

Contribution
(in vri)

OR

The

LOS
( L O h)

Provider'
Billing

Table 5.3.2: Example Physician Summary

Volume

Doctor

Total
Direct
Cost

Total
Contribution

Total
Hospital
Revenue

Total Bad

Total

Days

Provider
Billing

24.0

149,215.1

355,463.1

504,6782

261-8

73,850.0

93.6

828,9522

178,185.5

1,007.138.5

1.3826

503,541.6

129.6

1,262,265.6

271,685.9

1.533.951.4

2,288.2

689,961.0

4.8

20,9328

10.104.5

31,037.2

31-2

25,548.0

262.0

2261.3665

815.439.0

3.016.805.4

Total

Assume that the example hospital has fixed costs (F) of $865,437.40. a global budget (G)
of $3,076.805.40, OR availability (ORA-)
(Beds,-,,J

of 54,000.0 minutes, 3,900 bed days available

and a break even profit requirement (Profit = O). From the data listed in Table 5.3-2

it is evident that providers' preference is to deliver a rnix and volume of cases that satisfies the

overall constraints of the global budget, since

1,

p,' r, ($3,076,805) is equal to G. However,

it cm also be seen that demand for operating room time ( ~ , ~

p,'~,,
the amount of time available, as is the demand for bed days

a, E,p,'

= 54,208.8) is p a t e r

than

1,= 3,963.8). Additionally

the preferred mix of the four providers in the example does not generate sufficient income to
offset the fixed costs of production; since fixed costs are $865,437.40 and the mix of cases
preferred by the surgeons in this mode1 contributes only $8 15,437.40 towards overhead recoveq
the hospitai faces a potential shortfall of $50,000. To ensure the long term viability of the
institution, the hospital and its associated providers rnust, therefore, either change case mix or
practice patterns.
Let us assume that the hospitalysdecision makers choose to apply the following relative
weights to deviations fiom the objectives of the volume and cost models:

"-3.3

Examp r Penalty Weights

Negative deviaons from profi goals.

1
1

Positive deviations from profit goals.

0.0

Negative deviations tom remaining within the global budget

0.0

Weight

Description

- -

Positive deviations tom remainingwithin oie global budget

Positive and negative deviatians h m physician billing targets.
Positive and negave deviations h m case targets within soft bounds.

1
1
1

Value
1,000,000.0

1,000,000.0
1,000.0

O.1

Negative and positive deviations h m case targets between soft and hard
bounds.
-

Positive and negave deviations h m length of stay targets within soft

bounds.
Positive and negative deviations h m length of stay targets between soft
and hard bounds.
Positive and negative devialions from case time targets wilhin soft bounds.

O. 1

O. 1

Positive and negative deviations from case time targets between soft and
hard bounds.
Positive and negave deviations h m direct cost tergets within soft bounds.
Positive and negave deviations from direct cost targets between soft and
hard bounds.

Furthemore, assume that decision makers place bounds on physician's earnings such that
no individual provider's billings are to vary by more than

* 75% from the level listed in Table

5 -3.1. Finally, assume that limits have been specified so that changes in physician-specific case

volume greater than 10% of the value listed in Table 5.3.1 will be heavily penalized.
Under these assumptions, the volume model identifies three provider case mix arrays. (See
Appendix A for a listing of the example model in LP format.) The first, determined under preemptive goal number one, identifies one of rnany possible case mixes that allows the hospital to
meet its economic goals. The second, identifies a rnix that incorporates minimum bounds on

physician practice. The third, incorporates physician production preferences. Al1 arrays satisfy
economic and technological constraints. Results for each physician, appear below.

Table 5.3.4 Volume Model Case Mir Recommendations

Docbr

Referred

CMG

Volume
Under
Goal P,

Volume

C
D

Volume
Under
Goal P,

Volume
Under
Goal Pz

185

2-40

223

240

2.23

350

21-60

22.43

21-60

22.43

251

26-40

9.09

8.00

9.09

253

6720

87.67

91.33

87.67

251

28.80

17.47

19.05

17.47

253

100.80

114.21

11234

114.21

125

240

3.03

3.03

3.03

250

240

0.00

0.00

0.00

25200

256.13

255.75

256.13

Total

Table 5.3.5 Resource Allocation - Volume Model

Volume

Do&

Toial
Direct

Cod

Total
Contribution

Total
OR
Time

Total
Bed
D~P

Total
Provider

Billing

519,610.4

367,149.68

3.731-15

264.24

73.850.40

777.121-85

1,007.130.94

230,009.09

21,167.77

1,36021

503,561.11

131.67

1,259,856.51

1,514,345.67

254,489.16

28,545.82

2,242.17

689,961.60,

3.03

11,945.66

25.736.73

13,791.O6

555.26

33.38

25.548.00,

256.13

2.201.384.76

3.066.823.74

865,439.00

54.000.00

3.900.00

1,292921.00.

24.66

152,&0.72

96.76

O
Total

Toial
Hospiial
Revenue

The results fiom the volume model indicate, for this example problem, the economic goals
of the providers cm be achieved, as can the hospital objective of breaking even, within defined
constraints on revenue, operating room hours, bed days, and profitability. The results in Table
5 . 3 . 5 show the volume model c m idente a feasible solution to the resource allocation problem.

As can be seen fiom the data in Table 5.3.5, the rnix of cases selected by the mode1 is

econornically feasible; the total revenue consumed by this volume and mix of cases, $3,066,823 is
less than the global budget (G) of $3,076,805- The selected mix, in addition, generates enough
contribution ($865,439.00) to offset the fixed costs of production. Technological constraints

restriaing operating room tirne and bed days are satisfied by this allocation. Overall institutional
profitability and physician billing goals are also achieved within the specified constraints on
technology and the maximum change to physician practice. Nevertheless, to enable physicians to
obtain their desired case and volume mix, within the confines of existing resource availabilities, a
number of changes to physician practice patterns can be entertained. The cost model can be used
to determine the minimum changes to physician preference necessary to achieve these goals.

Cost Model Stage 1:Calculate Resource A Ilocations and De termine Completeiy Satzeed
Physicians
1) In this example, no physician is completely satisfied with the resources assigned by the
cost model. Physicians 4 B, C, and D are therefore designated as belonging to set D.
See Table 5.3-6.

Table 5.3.6 Example Satisfied / Dissatisfied Table

ii)

'

Satislhd
wih
Contribution

S a t i f i d with
OR Time

Satisied wini
Bed Days

Satisfied?

Set

SIosied with

Revenue

Satisfiad with
Direct Cod

Doctor

Since no physician is completely satisfied with the allocation of resources

determined by the volume model, updates to resource availabilities are not
required. The total availability of resources is therefore as given in Table 5.3.7
r

Table 5.3.7 Updated Resource Availability

I

Total Availability

Global Budget

Fied Cos&

866.439.0

OR Erne (Minutes)

54,000.0

Bed Dav

3,900.0

Cost Model Stuge 2: Determine

1)

3,076,805.4

if LOS Reductions are Repired

By evaluating equation 5.16 it can be determined if reductions to length of stay are

required.

Because V,, < 0, the global availability of beds is insufficient, in this example, to
meet the providers' preferred patterns of practice. Therefore, adjustments to LOS
patterns will be necessary for some, or all, physicians to achieve their preferred
volume and mix of cases.
ii) From Table 5.3.6 it can be determined that physicians A and D are satisfied with
the allocation of bed days made by the volume model. Thus, SB= { A D} and DE
=

iii)

{B,C).

By subtracting the bed preferences of doctors in SB from the total number of bed
days the remaining availability of beds for physicians in DE can be detennined.

Cosl Model Stage 3: Determine rf OR Time Reductions Are Required

1)

By evaluating equation 5.19 it can be determined if reductions to case tirne are

required.

Because V,, < 0, the availability of operating room time is insufficient, in this
example, to meet providers' preferred patterns of practice.
ii) From Table 5.3.6 it can be determined that physicians A and C are satisfied with
the allocation of operating room time made by the volume model. Thus, So = {A,
C ) andDo= {B, D).
iii)

By subtracting the OR time preferences of doctors in SB from the total availability

of operating room time, the remaining availability of operating room time for
physicians in Do can be determined.

Cost Modei Stage 4: Detennine

1)

rf Addilional Case Cost Reductions are Required

From Table 5.3.6 it can be seen that no physician is satisfied with the allocation of
contribution suggested by the volume model. Physicians A, B and D are
dissatisfied because ContributionMod y > Contribution,,,
Y. Physician C is
dissatisfied because Revenueaokd Y < Revenuhf- Y. Therefore Sc = (a)and Dc
= (A, B,C,and D).
By subtracting the contribution preferences of doctors in Sc fiom the total profit
generating requirement, the profit generation requirements for physicians in Dccm
be determined.

ii)

F" = F'
=

ontr ri but ion^"

Profit -

866,439 + O

fiSc
- 0

= 866,439

Note: For this example, we will assume a value of Y (the proportion of average
daily nursing costs saved with each one-day reduction in average length of stay) of
1.0. h addition the following nursing costs will be assumed:
Table 5.3.8 Example Direct Nursing Cost Profiie
Doctor

CMG

Description

185
350
251
253
251
253
125
350

Amputrtb'onExcept Upper
Muple Joint Replacement

B
C

~y8Colostamy
Mjr Intesiinal & RectalRocecfiae
Gasfrastomy & Colostomy
Mjr Intesiinal & Rectal Procedure

MjrC~nocedures
Multiple Joint Replacement

Volume
24
21.6
26.4
67.2
28.8
100.8

2.4
2.4

Direct
Cost
9,411O
.
5.8624
i;loOS.6
7.618.8
11,229.0
9,314.2
3,937.0
4,785.0

LOS

Direct Nuning

Ca9t
29.0
8.9

17.5
13.7
23.1
16.1J
11.0
20

5.000.0
2,000.0
6,000.0
3.500.0
5,500.0
4,500.0
1.800.0
2.400.0

Cost Model Stage 5: Identzfi MNNmum "Co~r"

Changes to Physician Practice
Having determined that changes are required to length of stay, case time and direct case
cost, the cost mode1 can be run to identiS, the minimum changes to individual physician practice
necessary to ensure each physician is able to achieve his or her preferred mix and volume of cases,

IO2

while remaining within the envelope of resources allocated by the volume model. For this

example, soft bounds were assumed to be 10Y0of preferred practice; hard bounds were assumed
to be *75% of preferred practice.

Under the above assumptions, the cost model produces three distinct sets of
recommendations for each physician. The first set of recommendations, determined under preemptive goals P,,and P,,identifies the smallest change necessary for each physician dissatisfied
with the volume model bed allocation to achieve a preferred case mix without exceeding his or
her assigned bed day supply. (It should be noted that the cost model reallocates "surplus" beds
not required by satisfied physicians to dissatisfied physicians. Thus, resource allocations rnay
change between the two models. The structure of the cost model, however, ensures that total
physician dissatisfaction wiil not increase with reallocation.) The second set of recornmendations,
determined under pre-emptive goals P,, and P , identifies the change necessary for physicians to
achieve their preferred case mix without exceeding the allocation of operating room time
determined by the volume model. The third set of recornrnendations, determined under
preemptive goals P,, and PJ2, identifies the change to direct cost, in addition to any cost savings
achieved through reduced length of stay, necessary for physicians to achieve their preferred mix of
cases. Results for each physician appear below:

Table 5.3.9 Re :ommended Physician Profile*

Dr.

CMG

Volume

Description

Direct
Cod

AmputEttion
9,411.00
2.40
Except Upper
350
5,862.44
Multiple
2 1-60
Joint Replacement
12,006.55
Gas~stomy&
26-40
Colostomy
253 Mjr Intestinal &
7101.83
6720
Rectal Procedure
C
251 Gashstomy &
1 1,229.00
28.80
Colostomy
253 Mjr Intestinal &
9162.77
100.80
Rectal Rocedure
D
125 Mjr Chest
3,937.00
2-40
Procedures
350 Muftipie Joint
4,785.00
2.40
Replacement
*suggested revisions to practice are s h o w in bold with italics.
185

OR

LOS

rie
(Min)

Provider
Billing

74.00

29.00

10,746.00

159.00

8.90

2,225.00

255.00

17.50

6,051.O0

21482
222.0
216.0

f 83.00

256.09

Table 5.3.10 Resource Allocation - Cost Mode1

1

Doctor

Total
Volume

Total
Direct
Cost

24.0 -

149,215.1O

93.6

C
D
Total

Total Hospital
Revenue

Toal
Conibution

Total
OR

Total
Bed

Time

Da~s

Total
Rovider Billing

504,678.24

355.463.14

3.6 12.00

794215.93

1,007.138.45

212.922.52

21,167.77

1.364.79 )

503,556.55

129.6

1,247.002.54

1,533,951-43

286.940.89

28,166.00

2,242.17

689,961.60

4.8

20,93280

10.104.46

1.053.83

31.20

25,547.99

!54,OOO.~

3,900

1,92,916.54

25200

2,211,366.38

31,03726 i
3,076,805.38

261.84

73,850.40

6.0 Refining the Mode1

The numencd example in Chapter 5 demonstrates that a methodology targeting change to
volume or practice parameten can be used to i d e n t e an allocation of resources that jointly
satisfies the preferences of both hospitals and their associated medical staff; the model retums an
array X = (x,,, x,, ..., x),

which defines a mix of cases for each physician that best satisfies the

economic objectives of the institution and its providers, given an array of provider preference
coefficients p,* and a set of global resource availabilities. The mode1 also returns a set of
recommended changes to practice necessary for physicians to achieve their preferred rnix of cases,
if dissatisfied with the allocation of resources implicitly defined by X. While the example model
has s h o w the concept of resource allocation via god prograrnming to be technically feasible, a
number of refinements are required to accurately reflect day-to-day realities faced in the operation
of an institution as complex as a medium-sized teaching hospital. In this chapter we discuss
refinements and extensions to the base model necessary to reflect operational reality and promote
practicai application.
6.1 Refining the Model's Scope

The volume and cost models defined in Chapter 5 are generic and can be applied to any
acute care hospitai. To maintain tractability, however, the scope of the model will be limited to
the patient population and physicians cornprising the Surgical Planning Council at Mount Sinai
Hospital. This restriction, while narrowing the range of the decisions that can be made, greatly
reduces the data required by the mode1 and eases the computational tasks of model execution and
analysis. Since surgical patients are largely scheduled on an elective basis, they represent a
population more amenable to case mix selection techniques than medicd patients, who are largely
adrnitted on an urgent or emergency basis. Accordingiy, a sample of data was obtained fiom

Mount Sinai cornprishg ail patients assigned to the Surgical Planning Council during the fiscal

year 1994/95. The data consists of two distinct files - a file of costing inf'ormation maintained by
the hospital's Decision Support Department and a file of demographic and seMce data maintained
by the hospitai's Health Records Department. The files consist of 14,260 records, iinked by a

unique encounter number. See Appendix C for the database layouts of the two files.
In addition to restricting the patient population to the Surgical Planning Council at Mount
Sinai Hospital, it was determined d u ~ development
g
that a restriction on the number of cases
included in the mode1 was necessary. In teaching hospitals such as Mount Sinai, the distribution
of patients presenting conditions is influenced by the composition of the medical staff and their
specialties. Nevertheless, Mount Sinai provides a diverse range of medical seMces (emergency
medicine, outpatient surgery, etc.) The pattern of seMces provided at Mount Sinai thus tends to
exhibit the statistical characteristics described by Silver and Patterson (1985) in their discussion of
Distribz~tionby Vahe OBV. DBV is an empincal observation, common in a wide variety of

contexts, in which a small proportion of an organization's products (= 20%) account for a large
proportion (= 80%) of the organization's total annual dollar usage. Distribution by value is
determined by calculating the weighted sum of product volume and product cost (DBV

=p

* v).

Figure 6.1 shows a chart of distribution by value for surgical "products" at Mount Sinai. In the

figure it can be seen that 73.67% of the total dollar cost of surgical services at Mount Sinai is
consumed by 20.2 1% of the total products, representing approxirnately 69.12% of the total
population served. Moreover, the four most "valuable" products (CMG253, 25 1, 125 and 3 S2),
representing 1.O4% of the total number of products delivered, consume 18.5% of al1 the resources
of the surgical planning council (see Table 6.1).

Figure 6.1 Product value analysis - Surgical Services, Mount Sinai Hospital.

Mt Sinai Hospital
Product Value Analysis
100%
90%

80%
70%
6u%
50%

40%
30%
20%
10%
0%

0%

30%

20%

10%

40%

SO%

70%

80%

100%

90%

O/. Num of CMGs

Table 6.1 Product Value Analvsis

-

--

CMG

Descnpt~on
253 ~ W OINTESTINAL
R
/%O RcTALPR
251 ~GASTROSTOMY& COLOSTOMY PROCED
-

254 IMAJOk INTESTINALAND RECTAL PR

31 1 MAJOR PANCREATICPROCEDURES
479 THYROIO PROCEDURES
035 BiADDER AND URETHRAL PROCEDURE
383 JOINT REPLACEMENT FOR MALlGNANT
289 INFLAMMATORY BOWEL DlSEASE
736 CHEMOTHERAPY

256 LESS EXTENSIVE ESOPHAGEAL. STOM

356 FRACTURED FEMUR PROCEDURES WIT

184 MAJOR RECONSTRUCTIVEVASCULAR

76 MAJOR HEAD AND NECK PROCEDURES
725 MAJOR LEUKEMlA AND LYMPHOMA PR

131 RESPIRATORY NEOPiASMS

Volume

1
1-

Total Cost

Cum%
Total Funding Nurn CMGs

211 1 $2.538.896.00 S2.199.358.54

1131 S2.044.881.Mi $1.542.611.54
$1,078.910.50
$1.72137.01
f 1.382.930.94
$1,544.697.23
%[email protected]
981
$911.559.&
13
9804.30900 S42.465.66
228
f770.298.00 $706,530.75
$1.005.308.45
11 76
$548,313.00 $580.63.25
31
$511,284.00 S478.513.M
106
S491.923.00
198
S480.879.00
70
S473.507.00
56
$460.731.00
27
$418,014.00
40
$416,256.00
28
$366085.00
59
$351,828.00
-

0.26%
O 52%.
078%
1.04%
1.30%
1.55%
1.81%
2.07%
2.33%
2.59%
2.85%

3.11%-

Cum %
Volume

3
2.27%
311%
4.03%
2380%
24.77%
25.46%
25.55%

27.15%
27 78%
36.a
36.24%

Cal

Funding

5 57W
9.48%
12.21%

6.59%
11 91%

15 43%
18.50%

16!57%

3.98%
23 39%

M 07%

25 75%
27 84%
2984%
31 71%.

33.13%
34.46%

23 99%
25.64%
26.51%
28.m
30.84%
32.31%
33.52i

CMG

Vdume

Desaiphon

56 LE% INSERTION WITHOUT CC (MNRH)

379 OTHER MUSCULOSKELETAL PROCEDURE

1 127

~THER
RESPIRATORY PROCEDURES W

78

Total Cost

s320.609.W

Total Funding

Curn %

Curn %

Num CMGs

Vdume

S7.473.90
$191,979.40

81-

$311.790.00

331

$307.462.001 $290.926.801

5.70%
5.96%

42.34%
4291%

6 . 4 ~ 1 43.56%1

Cum X

Cosl
45.12%

Curn %
Fundng

45.93%

44.20%
44.68%

4753361

4601%]

Silver and Peterson (1985) argue that products should be divided by value into three
goups (4B, and C) for management analysis and control. "A's" represent the most important
5- 10% of products that, because of their cost or volume, require in-depth management analysis

and detailed tracking. "B" products are secondary products (50 - 60%) that require a moderate
level of control. "C" class products make up the remainder of an o r g ~ t i o n ' product
s
mix.

"C" products are small, inexpensive, or rarely used items that comprise 30 - 45% of an
organization's total product Line-up. Since "C" products are not critical to the overall success of
an organization, they need only be considered in aggregate.
Using the Silver and Petenon framework, Mount Sinai's surgical products can be
separated into three groups for the purposes of the model. "A" group products, which encompass
the first four CMGs, represent important products that must be accurately captured and tracked in
the model. "B" products, consisting of 1 12 CMGs that consume 67.53% of the total surgical
resources of Mount Sinai Hospital, are products that must be tracked and captured in the model,
but with less accuracy than A products. "C" products, because of their small impact on the
overall function of the organization, can be aggregated into a composite CMG. By aggregating
Iow value products into a comrnon "CMG", the number of unique CMGs in the mode1 database
can be reduced fiom 386 to 1 17; the number of unique doaor/CMG pairs cm be reduced from
1,758 to 902. In practical terms aggregating low value products allows the number of constraints

in the volume model to be reduced From 8,586 to a more manageable 4,402; in the cost model, the
number of constraints can be reduced ftom approxirnately 10,923 to 5,787.

1O8

Restricting the scope of the model to patients of the surgical planning council at Mount
Sinai Hospital and aggregating low value products, though reducing the generality of the model,
allows the sue of the resulting goal programs to be greatly reduced in size. This makes solution
on an ordinary desktop computer, using comrnercialiy available linear programming solvers,
operationally feasible. The two models are conjointly implernented in Microsoft Visual BasicQ
Version 3.0 and make use of an Microsofl ~ccess' database for data storage and retrieval. A
commercial linear prograrnrning solver, CPLEP, fiom CPLEX Corporation, is used to solve the
cost and volume models through an implementation of a sequentiai multiplex aigorithm (Ignizio

and Cavalier 1994). The software implementation of the model runs on an ordinary IBM
compatible PC under the Windows operating system. A Windows interface allows users to
interact with the model, set run time parameters and review results. Mode1 runs typically require
22 minutes; solver time for the individual models, once identified, built and forrnatted is between 2

seconds and 1.5 minutes, depending upon the number of constraints that appear in the model.
6.2 Determining Physician Billings

Identifjing the relationship between physician billings and hospital costs is key to the
proper hinction of the volume model. In Ontario, both physicians and hospitals are independent
agents that bill a central insurance agency (the provincial government) for the services they
provide to patients. The societal costs associated with an episode of care, which are both borne
by the provincial insurance agency, are therefore:

Sociefaf Cost per Episode of Care

Fee For Servicepwa,,

Prospective

Subject to
Restrictions on total hosptial spending.
Restrictions on total physician biflings.

Since the volume model attempts to strike a balance between hospital revenue and physician
billing objectives within the constraints of technology and doctor preferences, it is necessary to
establish a relationship that Links physician fees for an episode of care to the institutional revenue
and/or the institutional cost of that episode.
Although the provincial govemment in Ontario insures both physician and institutional
services, obtaining information describing the components of health care cost is, in practice,
difficult. Provincial statutes guaranteeing rights to privacy are designed to prevent access to
patient-level OHIP information. In addition, physicians as a group are (understandably) reluctant
to make income information public; hence, biliing information is sensitive data. As a result,
physician billings are rnaintained by the province in data structures separate fiorn institutional
billings. The two data elements are not routinely linked; data iinking hospital costs to physician
billings for specific health care products are therefore difficult to obtain.
Because of the absence of patient-specific case information and the separation of physician
information fiom institutional costs, it has been difficult to determine the total public costs of
s Ontario. Coyte et al. (1995), for instance, describe a study
specific health care s e ~ c e in
undertaken at the Hospital for Sick Children in Toronto to determine the costs of treatment
protocols for paediatric femoral shaft fractures. To identiq the cost effectiveness between the
two protocols Coyte et al. quantified the costs and benefits of each. Cost information was
obtained fiom the study site's accounting systems. Physician billings, however, could only be
estimated through consultation with a panel of experts. Coyte et al. developed a sewice summary
for each protocol, based on the consensus of the expert group. The estimated treatment patterns
were then converted into physician billings by relating each seMce to its appropriate billing
through the OHIP Schedule of Benefils.

Although appropriate for a cost-benefit study with a s m d number of clearly defined

treatment protocols, such a technique is inappropriate for a more extensive analysis of costs with
less well defined protocols. Gathering consensus data on treatment protocols for each of 400 500 CMGs, would require an inordinate amount of expert t h e . In addition to being tirne
consuming and costly, there is no obvious method for validating such protocols. Estimated
billings would be, at best, an educated guess. It was therefore decided to extrapolate the
relationship between physician billings, hospital costs and revenues from a sarnple of patient
records and physician billing data. Accordingly, a datafiie was obtained fi-omMount Sinai
Hospital's case costing system and linked to information maintained by the provincial insurance
agency. A total of 8,039 records identifj4ng patients discharged during the period from
September 1, 1994 through November 30, 1994 was obtained from the hospitai's medical records
and matched with costing information maintained by the hospitai's decision support system. The
information abstracted included patient demographics (se% age, and provincial health insurance
number), clinical information (admission date, discharge date, most responsible physician service,

CMG, length of stay and procedure), and direct cost of hospital care.
Case cost information and patient encounter number was stripped from the data, to
preserve confidentiality, and the information was passed to the Information Planning and
Evaluation Branch of the Ontario Ministry of Health. The Ministry of Health, using the health
insurance number as a key, extracted dl OHIP billings for each of the 8,039 records for the penod
between November 1993 and December 1994 in which Mt. Sinai's institution code appears.
(Claims for an extended period are required since patients discharged between September 1, 1994

and November 30, 1994, may have been admitted to the hospital at a much earlier date. Over the
course of an extended stay several physicians rnay submit numerous claims al1 relating to a single

111

individual.) The set of claims were then matched and surnmed for each patient for the period
between their admission date and discharge date at Mount Sinai. Case costing information was

then re-appended to the dataset. The data was then separated into inpatient and outpatient sets
and a multiple linear regression analysis was run on physician biiIings, using institutional cost and

revenue as independent variables.

Initial results of the regression analysis were not prornising. A coefficient of detemination

8 = 0.17 was obtained when the regression anaiysis was run on the entire data set.

An analysis of

the data indicated that a large number of records relating to a specific group of procedures had
no OHIP billing data. See Table 6.2.

Table 6.2 CMGs Commonly Missing Billing Data

CMG

# Records
Total Records
with No OHlP
in Sampk
Data
607
121
75
179
69
694
127
64
45
197

Description

611
624
648
619
609

VAGINAL DELIVERY
ANTEPARTUM NON-COMPLICATING
NEONATES WEIGHT > 2500 GM
FALSE LABOUR, LOS < 3 DAYS
VAGINAL DELIVERY WlTH COMPLICATIONS

736
623
196
604
602
513
852
628
601
639
646
253
254
610
735
625

CHEMOTHERAPY
.ANTEPARTUM COMPLICATING DlAG
CARDIAC CATHETERIZATION
CAESAREAN DELiVERY
CAESAREAN DELIVERY WITH COMP
OTHER TRANSURETHRAL OR BlOPS
PROCEDURE CANCELLED (MNRH)
NEONATES WEGHT 1000-1499 GM
REPEAT CAESAREAN OELIVERY WI
NEONATES WElGHT 2000-2499 GM
NEONATES WEIGHT > 2500 GM WI
-MAJOR INTESTINAL AND RECTAL
MAJOR INTESTINAL AND RECTAL
VAGINAL DELIVERY AFTER CAESA
RADIATION THERAPY
NEONATES WEIGHT < 750 GRAMS

I
-

33
20
19
18
17
15
11
11
10
10
10
10
1O
9
9
9

47
68
108
51

18
18
41
28
36
147

90

58
32
16

13

As can be seen, the rnajority of records lacking billing information were obstetric or

gynaecological CMGs. Comrnon research practice suggests data points obviously in error either

be corrected or eliminated from regression anaiysis. However, the large number of records
lacking information in this patient subset suggests an exogenic error rather than chance
occurrence. Therefore the data, as it relates to obstetnc/gynaecology patients, is suspect.
Fortunately few of the CMGs in Table 6.2 are surgical procedures. It was thus possible to lirnit
the records in the regression analysis to those matching patient data in the hospital abstract, in

which the patient was considered to be under the care of the hospital's Surgical Planning Council.
The regression analysis was then re-mn using a total of 1,553 surgical inpatients and 1,465
surgical outpatients.
The results of the regression analysis, using only surgical patient data, revealed significant
relationships between physician biliings and institutionai costs and revenues, particularly in the
case of inpatients. Physician billings for inpatient surgical cases were found to Vary with direct
case cost; the value of ?, the coefficient of determination was 0.727. See Tables 6.3 - 6.5.
Table 6.3 Regression Surnmary Inpatient Data
Standard Error

505.899

Table 6.4 ANOVA - Regression Anaiysis for Inpatient Data

df
Regression
Residual
Total

MS

SS

1057713768.591

1551
1552

396953806.038

l~irectCase Cost

Coefficients

4132.758

1057713768.591
255934.111

1454667574.629

Table 6.5 Regression Model for Inpatient Data

Significance F

0-2681

Standard Enor

0.004f

t~tat

64.2871

P-value
0.000e+001

Table 6.3 indicates the existence of a strong linear relationship between the direct cost of
hospital seMces and physician billings. In equation form this relationship is:

(6.3) Regression Line - Inpatients

Billing

= 367.992 +

0.268d i r e c t COSI

Both the intercept and the regression coefficient are signincant at the 0.001 level. A plot
of the data and the corresponding regression Iine cm be found in Figure 6.2.

Figure 6.2 Regression plot, inpatient physician billings.

I

Regression Plot Inpatients

Predicted OHlP vs Actual

/ OTdal OHlP

~redictedTdai OHlP

i
i

i
O

1OOOO

XX300

30000

40000

50000

60000

irect Case Cost

Regression results for outpatient data indicated a weaker relationship between physician
billings and institutional costs or revenues. Physician billings for outpatient surgical cases were
found to vaiy with both direct case cost and institutional funding; the value of ?, the coefficientof
determination, was found to be 0.347. See Tables 6.6 - 6.8.

Table 6.6 Regression Surnrnary Outpatient Data

R Square
Adjusted R Square
Standard Error
_Observations

0.348
0.347
158.713
1463

Table 6.7 M O V A - Remession Anahsis for Out~atientData

-

df
Regression
Residual
Total

--

SS
19633196.910
36827427.100
56460624.010

2
1462
1464

--

MS
9816598.457
25189.756

.-

Significance F

389.706

Table 6.8 Remession Mode1 for In~atientData

lntercept
Direct Case Cost
-Hospital Revenue

Coefficients
-3.759
0.408
0.254

Standard Error
9.740
0.041
O.Oi6

t Stat

-0.386
9.847
15.402

P-value
0.700
0.000
O. 000.

Tables 6.6 through 6.8 indicate a linear relationship between the direct cost of hospital
outpatient services and physician billings. This relationship is given by the following equation:
(6.4) Regression Line - Outpatients
BiIIirzg = -3.759

0.408*Direct Cost

0.254 *Revenue

Both regression coefficients were found to be significant at the 0.00 1 level. However, the
intercept of the fitted line was not found to be significant. This indicates a high degree of
uncertainty as to the exact value of the y-intercept in the regression analysis. A 95% confidence
interval places the intercept in the range [-22.866, 15.3471. A plot of the data and the

corresponding regression line, showing predicted billings against hospitai revenue can be found in
Figure 6.3.
The regression analysis can be used to draw a number of conclusions about the econornic
behaviour of physicians. The large value of the intercept for the inpatient regression (a = 367.99)
reveals the economic value of beds to physicians; physicians achieve a larger proportion of their
income from "turning beds ove? than from providing ongoing care post-operativeiy. In the
context of surgical patients this clearly indicates that physicians receive much of their income fiom

Il5

performing a procedure, which happens only once per admission, rather than from seMces to
patients throughout their post-operative stay. The regression analysis for inpatients also suggests
that physicians should be amenable to suggestions for decreases in length of stay, since such
reductions ensure a greater number of patients can be "turned-ove? in any given bed, thus
increasing physician billings. The positive relationship between physician billing and direct case
cost - for both inpatients and outpatients - also indicates physicians have an econornic rationale
for tnggenng the consumption of as large a number of hospital resources as possible.
Figure 6.3 Regression plot, outpatient physician billings.

Regression Plot Outpatients

Predicted OHlP vs Actual

ig

Predicted Total OHlP

200

400

600
800
1000
Hospita1 Revenue

1200

1400

1600

The billings estimated from regression equations are, however, only a surrogate for
physician billings. Clearly, they do not represent the individual billings of particular physicians.
Frequently, several physicians can (and do) perform services for patients dunng an episode of

116

hospitalization, for which they are reirnbursed by the central insurer. By combining the billings
from ail physicians, as in this analysis, income is attributed only to the physician deemed "most
responsible". The fees charged by anaesthetists, for instance, are considered in this analysis to be
part of the income of the most responsible physician. While this is clearly an approximation that
departs fiom reality, it has ment for use Ui the cost and volume models, since the unit of analysis
for case mix selection is the moa responsible physician or clinical tearn.
6.3 Applying Appropriate Benchmarks
As defined in Chapter 5, the cost mode1 tends to target physicidCMG combinations that

are either high volume or high cost ("A" or "B" products in a Silver and Peterson distribution by

value analysis), rather than cases that are, by some normative standard, inefficient. While there is
a clear rationale for minimizing the number of cases for which practice changes are suggested,
such a policy is clearly inequitable. To ensure equity, the practice of less efficient physicians
should, wherever possible, be targeted for improvements before changes are suggested to the
practice of efficient physicians. However, to make a normative evaluation of physicians' pattern
of practice, a yard stick or benchmark against which to measure practice is required.

Gift and Mosel(1994) define health care benchmarking as the continual and collaborative
discipline of measunng and comparing the results of key work processes with those of the best

performers. Benchmarks can be established by comparing against intemal processes, extemal

competitors within an industry, or functional equivalents in non-related industries. By introducing
benchmarks, or normative yardsticks, into the aigorithm that selects CMGs for practice change,
the model can be made to serve as an automatic mechanism for institutional benchmarking. If
normative benchrnarks are integrated into the model, suggested recomrnendations are not oniy

117

necessary for economic welI being, they also indicate areas where irnprovements should be
expected.
6.3-1 LOS Benchmarks

Ident@ng appropriate benchmarks for inpatient length of stay is relatively

straightfoward. Each year the Canadian Institute for Health Information (C IHI) publishes a list
of average length of stay for each case mix group for ail hospitals in Canada. This list, commonly

known as the database [en@ ojskzy, is distnbuted to all hospitals and is used by institutions to
rate their own performance. Length of stay data for 1994-95 was obtained corn the C M I via
Mount Sinai Hospital's Health Records Department and incorporated into the modei's database.
The following algorithm was then integrated into the Visual Basic prograrn to rnodify the manner
in which soft bounds for length of stay constraints are set, thereby eliminating inequitable practice
recomrnendations:
If a user gives a specific bound on any doctorKMG combination, it is always used
and never replaced by a default value.
If no bounds are indicated by the user for a given doctodCMG combination, the
mode1 assumes a default value - a percentage variation from the target - for both
for example) c m be set by
the hard and soi? bounds. The default value
the user at run time.

If the database length of stay is greater than the assumed hard bound determined in
step 2, but less than the target length of stay, the database length of stay is used as
the sofl bound.
If the database length of stay is less than the default hard bound determined in step
2, the soft bound is set to the hard bound.
Using benchmark data to set soft bounds eliminates naive model solutions by biasing the
model's selection of physicianICMGs towards inefficient products a d o r inefficient providers.
The sofi bounds in the model represent lirnits on undesired, but acceptable deviations on physician
practice parameters. The goal program, when selecting physician practice patterns attempts, first

and fokmost, to minimize or eliminate any deviation eom the desired preference pattern of
physicians. If not al1 practice preferences can be achieved, the model wil attempt to find a
solution for each physician/CMG combination that is somewhere within the range of undesired,
but acceptable deviation. (In practice, the mathematics tends to force the solution to the soft
bound rather than to an area between the target and the sofi bounds.) If no acceptable solution
can be found for each physician/CMG combination within the range of acceptable deviation, the

model will identi@ practice parameters outside the range of acceptable deviation (ie, the area
between the hard and soft bounds). These recomrnendations are, however, highly penalized to
discourage the model fiom retuming solutions that are difficult to implement.
6.3.2 Case Cost Benchmark

Comparative data to benchmark inpatient direct case cost was obtained from the Ontario
Case Costing Project (OCCP) for the fiscal year 1994- 1995. The data was integrated into the
mode17sdatabase and the Visual Basic routine to select bounds for direct case cost was modified
in a rnanner similar to that outlined for length of stay data. Comparative outpatient data was not
available from the OCCP. Therefore an intemal benchrnark was identified for each outpatient
Day Procedure Group @PG) by averaging the direct cost of al1 outpatients in the 1994195 dataset
obtained from Mount Sinai7scase costing system.
6.3.3 Operating Room Tirne Benchmark

Since comparative data to benchmark operating room case time was not available from the
OCCP, an interna1 benchrnark was used to set hard and soft bounds for operating room case time.
The average case time for each CMG (and DPG)was calculated fiom the model database. These
figures were used to set sofi bounds for operating room case time in the database in a manner

119

similar to that outlined for length of stay data. A listing of the comparative data used in the mode1
appears in Appendk D.
6.4 Operating Room Setup and Tumaround Time

Case time information stored by Mount Sinai's Health Records Department, by

convention, lists only skin-to-skin time, the clinicaiiy important component of operating room
resource consurnption, in the permanent patient record. Stewart (197 1) notes, however, that case
time is only one of many activities that comprise the surgical cycle. These activities include:
Se~up:The tasks necessary to prepare an operating theatre for a procedure. Setup
includes unpacking sterile instruments, preparing drapes, secunng drugs, supplies
or prosthetics.
Posztzoning: Preparing and positioning the patient for surgery.
Induction: Anaesthetising the patient.

Surgey: Case time

Recovery: Ceasing the flow of anaesthetking agent and retuming the patient to
normal respiratory fnction.

C l e m p : Removing waste, gathenng soiled instruments for sterilization and

sterilizing soiled surfaces.
Although case time consumes the majority of the surgical cycle, setup and cleanup time
account for a significant proportion of total operating room use. Hancock and Isken ( 1992) note
that productive operating room utilization rarely exceeds 80%. Despite the impact of setup and
cleanup time on overall operating room resource utilization, accurate information on the duration
of setup and cleanup times is difficult to obtain. While most institutions employ automated
booking systems to record critical components of the surgical care process, these systems often
focus on the capture and management of the case time component of the surgical care process; at
Mount Sinai standard setup and cleanup times are assumed when surgical cases are booked, even

120
though fields in the automated booking system's database exist for recording actual setup and
cleanup tirne.
Mitha and Reid (1996) in a study of setup and cleanup at Mount Sinai note that cleanup
and setup time are not distinct activities; it is ofien difncult to discem when setup ends and
procedures begin, particularly when cases are long or involved. Similarly, cleanup activities may
overlap recovery or even case time. As a result, setup and cleanup activities are frequently not
logged or are inaccurately recorded. Nevertheless, because setup and cleanup aflfect resource
capacity and patient volume, an accurate estimate of these parameters is required for the
allocation model.
In the summer of 1996, Mount Sinai Hospital engaged a consultant to review its surgical
senrices. The consultant, in an effort to determine the productive efficiency of the operating
rooms at Mount Sinai conducted a "case-to-case7' analysis to estimate average setup and cleanup
intervais. Using a sample of 12,945 cases drawn fiom the penod between April 1995 and June
1996, average inter-procedure times were determined by measuring the interval between patient

exit and procedure start for sequentially scheduled procedures taking place on weekdays between
0630 and 1700 hours. Inter-procedure intervals greater than 90 minutes were considered to be

errors and were deleted fiom the analysis. Using this algorith, the following inter-arriva1
intervals were determined:

Table 6.9 Case-to-Case Analysis Inpatient Surgery

SeMce

1 General Surgery

1 Gynaecology

1
1

# of Cases

2,168

Neurology
Ophthalmology

1 Oral Surgery

1
1

Procedure Tirne

2:23

1
1

Case-to-Case Tirne

0:32

18

1:44

0:26

1,274

?:O8

0:17

127

3:21

0:32

Orthopaedic Surgery

1.309

2:31

0:31

Otorhinolaryngology

729

156

0:21

Thoracic Surgery

435

3:08

0:33

Urology

345

1:36

0:28

Vascular Surgery

290

2:12

0: 34

# of Cases

Procedure lime

Case-to-Case Time

0:16

O: 34

Table 6.10 Case-to-Case Analysis Outpatient Surpery

Service
Gastroenterology
General Surgery

r-

0:44

--

0:22
-

0:14

11

0:38

0:21

Ophthalmology

969

0:56

0:13

Oral Surgeiy

47 8

1:19

0: 16

Neurology

Orthopaedic Surgery

1 Otorhinolaryngology
Plastic Surgery
Thoracic Surgery
Urology
Vascular Surgery
Other

--

2.1 03

1
1

O:%

Gynaecology

260

054

15

55
149

1:16
0:45

0:20

O:18
0:13

0:30

0:19

53

2:26

0:25

1:33

0:31

57

022

O:32

--

A plot of inpatient inter-amval time indicates a bimodai distribution, with times clustered

at either 20 or 30 minutes. Of particular note are the case-to-case times of the surgical seMces
compnsing the DentaVEyeENT tearn, which are significantly shorter than the case-to-case times
of other services. Accordingly, two distinct inpatient case-to-case time distributions are assumed

122

in the ailocation model. Surgical procedures performed on DentaVEydENT patients are assumed
to require a 19.26 minute setup and cleanup tirne. Procedures perfomed on al1 other inpatients

are assumed to require an average of 30.9 1 minutes. These estirnates compare weii to the
assumed setup and cleanup times used when booking cases at Mount Sinai. (DentaVEyeiENT
procedures are assigned a standard setup and cleanup time of 20 minutes; al1 other procedures are
assigned setup and cleanup times of 30 minutes, except orthopaedic cases which are assigned
setup and cleanup times of 40 minutes.)

.Figure 6.4 Frequency plot of case-to-case intervals for inpatient surgery at Mount Sinai.

Operating Room
Interamval Frequency (Inpatient)

10
15
20
Case-to-Case Tinte

25

30

35

(Minutes)

A plot of outpatient inter-&val time indicates a uni-modal distribution of setup and

cleanup times, centred at a mean of approximately 15 minutes. Accordingly, the allocation model
assumes al1 outpatient procedures require a setup and cleanup time of 14.97 minutes. This figure

123

also corresponds closely to the standard 15 minute setup and cleanup time assumed by the
hospital when booking outpatient cases.
Figure 6.5 Frequency plot of case-to-case intervals for outpatient surgeiy at Mount Sinai.

Operating Room
Interamval Time (Outpatient)

10

15

20

25

30

35

Case-to-Case Tlme

Minutes

6.5 Non-Deterministic Demand

During development of the resource allocation model, a number of individuals from the
study site expressed concerns regarding the model's ability to set case volume targets for
physicians. The pattern of patient arrivals to the hospital and the diseases presented, they note, is
neither entirely randorq nor entirely deterministic. A model that identifies volume for physicians
may, therefore, return solutions that cannot be implemented, since neither the physician, nor the
institution has complete control over emergency patient amvals.
To address concerns over the implementability of deterrninistic volume targets, the scope
of the model was, as noted, lirnited to the subset of patients at the study site fdIing under the

124

auspices of the Surgical and Perioperative Planning Council (one of four major clinicd categones
at Mount Sinai Hospital). Since surgery at Mount Sinai is largely elective (85.93 % of al1 cases
are considered elective, 0.20% are urgent and 13.87% are emergency), these patients represent a
population amenable to case mix seleaion.
To hrther strengthen the implementability of the recornrnendations suggested by the
model, the number of ernergency cases for each CMG in the database was cdculated and input
into the volume model as an assumed hard lower bound on the number of cases that must be
selected (eg

x,r,

2 z(~rner~en
Cases)
c~ V

CM&). This bound constrains the number of cases

suggested by the model for each CMG to be greater than or equal to the number of emergency
cases that were performed in the previous year. This ensures that, as an institution, provisions are
made for a number of non-deterrninistic arrivais. The constraint does not, however, ensure a
minimum number of emergency patients for each physician.
6.6 Team Level Data Aggregation
Both the cost model and the volume model use the physician and CMG as the basic unit of
aggregation; each

represents a unique combination of physician y and CMG n. Physician and

CMG were selected as the unit of aggregation to give the models sufficient granularity to produce
implementable results, since there is oflen a great deal of variability between the case mixes of
physicians within the same medical service. Moreover, practice patterns c m Vary between
physicians with sirnilar case mixes. Nevertheless, initial reaction fiom decision makers at the
study site when presented with a surnmary of model output aggregated at the physician level
suggested a surfeit of information. Decision makers were overwhelmed with the arnount of data
presented by the model. (Detailed model output for physician based practice can exceed 12,500
separate recornmendations or approximately 52 pages of print.) In addition, the decision makers'

frame of reference did not appear to encompass a physician level view of the data; decision
makers expressed a desire to view the data at a more aggregate unit of analysis, the clinical team
level, which is the level of aggregation comrnonly used at the study site for purposes of planning
and control.

Mount Sinai, in response to directives fiom the Canadian Council for Health Facility
Accreditation, adopted an organizational structure emphasizing tearn activity and decentralized
patient care in 1994. The structure consists of 19 clinical teams, each centred around a comrnon
patient population. The 19 tearns are grouped into four clinical divisions, known as councils:
Medical, Pennatal, Academic and Cornrnunity Health, and Surgical. The Surgical Council at
Mount Sinai Hospital, which is the focus of this study, consists of six teams, having the following
resource consumption protile:
Table 6.1 1 Resource Profile Surgical Council Mount Sinai Hospital 1994/95

Team

Cases

Bed Days

General & SpeciaHy

Surgery

1,823

1,949

67.142

Musculoskeletal

1,194

6,786

120,939

1 Tharacic Surgery

517

1,297

32.800

OR Tirne

Direct Cost

Contribution

934,059.00

$ 1,232.307.50

$ 2,516,790.00

$ 5,025,525.20

1$

598.558.00

604,669.20

Oncology

3,535

17,684

325,784

$ 6,820,616.00

$ 7,304,516.30

DentaEyeENT

3,114

2,563

206.167

$ 1,562,517.00

$ 3,793,919.40

14,260

42,963

Total

Aggregating information on the basis of team rather than council is not simply a matter of
grouping physician results by team. For clinical and administrative rasons a 1 :1 correspondence

does not exist between physicians and tearns. Mount Sinai's Health Records Department, which

is responsible for creating and maintainhg standards of coding for medical records, has defined an

algorithm for assigning patients to clinical teams that groups patients into clinical and

126

administratively significant groups for purposes of evaluation and control. (Team budgets, for
example, are based on the assigned patient population.) The assignment of patients to clinical
teams, however, may be made independently of the most responsible physician. Thus, a
physician's patient population may be assigned to several teams. Tearn level and physician levels
of aggregation are therefore similar, but not strictly compatible. To support a team based

allocation, separate views are maintained in the rnodel's underlying database for both physicians

and teams. Both are developed frorn a common underlying database of patient specific records.
Either view may be chosen at mn time.
While physician level aggregation provides the greatest data granularity, for the sake of
brevity of output, team level data aggregation will be assurned in production runs of the model.
By so doing, we posit the team, rather than the physician, as the arbiter of demand and assume

that the team is able to exert control over its associated physicians. While this is clearly a
simplification, it is consistent with the operating philosophy of the study site, supports the
decision makers' preferred frarne of reference, and significantly reduces the computational efforts
required to generate a solution.
6.7 Fractional Case Miar Recommendations

When presented with initial results, decision rnakers at the study site also expressed
concem over the number of fraaional case mix recommendations retumed by the model. Such
recornmendations, they noted, could not be implemented and therefore did not make practical
sense. Decision makers therefore recomrnended that, wherever possible, integer solutions should
be returned; at least two decision makers, when asked to provide feedback, suggested rounding
model results post-hoc. There are, of course, technical difficulties associated with rounding
fraaionai rnodel results post-hoc. Such a procedure may violate physicai constraints such as

available beds, operating room time or budget. The more rigorous approach of defining dl
decision variables as integers, presents solution difficulties. The amount of time required to solve
a large integer programming problem with thousands of variables and tens of thousands of

constraints would likely prove prohibitive, particularly where nine separate models must be
executed to achieve a solution. Therefore, an intermediate alternative was adopted.
As noted in Chapter 5, the solution values retumed by tinear programming models occur

at extreme points of the solution space, which are points defined by two or more constraints.
Therefore, the volume model tends to make case mix recomrnendations that are exactly equal to
the preferred volume, the volume implied by the user-defined soft bounds on volume, or the

volume implied by the user defined hard bound on volume. (Mode1 recomrnendations may, of
course, result in intermediate values being selected. In these instances, another constraint, such as
bed availability, is active.) Therefore, if the user defined bounds on volume are rounded to integer
quantities pt-ior to execution of the volume model, fiactional case mix recomrnendations c m
largely be elirninated. In modelling tems, volume model Constraint Sa, which sets soft upper
bounds on case volume, must therefore be revised so that smallest integer greater than or equal to
the implied soi? upper bound is returned as a case rnix recommendation:

Similady, volume model Constraint Sb, which sets the soft lower bound on case volume
must be revised so that the greatest integer less than or equal to the implied by the soft lower
bound is retumed as a case mix recomrnendation:

Hard bounds on case m k recornmendations mua also be revised so that integer values will
be returned by the model:
A bsolrire

A bsolure

These revisions rninimize the number of recomrnendations for case mix change with
fractionai values, while p r e s e ~ n gconstraints on scarce resources. In addition, integer model
recornmendations can largely be achieved without having to declare dl solution variables as
integers. Thus, reasonable solution times c m be preserved, while achieving practical case mix
recommendations.
6.8 Competition and Cooperation

In Chapter 5 , it was noted that the resource allocation model presumes cooperation
between physicians. The degree to which physicians cooperate with one another, however,
affects the nature of equity between volume and cost model recommendations. Under the
assumption of cooperation, equity is preserved between the cost and volume models by
constraints that ensure physicians are granted at least as many resources (bed days, operating
room minutes, and direct cost) by the cost model as was assigned by the volume model. For
example, cost model constraint 1b:

l6)
n : l

PL

t Bed,&~~ed

for al/ y E D ~11;

ensures each physician dissatisfied with his or her allocation of beds is assigned at least as many
bed days by the c o a model (Znp & , J as was granted by the volume model ( B e d ~ . ~ ~ &These
).

constraints ensure no physician is penalized for participating in the volume model; the cost model,
under the assumption of physician cooperation, nearly always satisfies conditions for Pareto
optirnality: no physician is any worse off and &en one or more physicians is better off for having

participated in the cost model. (Pareto optirnality cannot be guaranteed, however; while no
physician is worse off for participating in the cost model, there is no guarantee that any physician
will be made better off)
While the assumption of cooperation between physicians promotes equity between the
cost and volume models and is dmost always Pareto optimal, it does not ensure a resource
allocation for individual physicians and tearns that is commensurate between the two models.
This is best illustrated by the following example:
Scenario: The Surgical and Penoperative Planning Council at Mount Sinai
Hospital has been asked to reduce its total costs by $3 million from $39.48 million
to $36.48 million. The hospital has determined that $1 .O million cm be cut by
reducing operating room time by 50.000 minutes and eliminating 2,000 bed days.
Nevertheless, an additional $2.0 million must be eliminated through changes to
volume, mu< or practice. A constrained rate-based payment rnodel is in effect.
Default model weights are listed in Appendix E.

When mn under the assurnptions in the scenario described above, the allocation model
produces the results listed in Table 6.12. As can be seen fiom the table, recornmendations made
under the volume model are cornmensurate with recornmendations made by the cost model at the
aggregate level. However, on an individual team (or physician) level, the recommendations are
not equal between the two models. The volume model, for instance, recommends a total
allocation of 295.722 operating room minutes be aliocated to the Oncology Team, while the cost
model recommends the Oncology Team receive 296,577 operating room minutes, an increase of

130

854 minutes. Because the cost model is designed to promote pareto optimality between the two

models, aii physicians or tearns dissatisfied with their allocation of resources, as assigned by the
volume model, are guaranteed to receive at least as many resources in the cost rnodel. Resources
which are surplus to any individual physician's or tearns' preference, however, are made available
to those teams that are dissatisfied and are re-docated in the cost model. in the example
illustrated by Table 6.12, OR time surplus to the needs of the GI Team is reallocated to the
Oncology Tearn. While this approach promotes pareto optimality and minimizes overail
dissatisfaction with the assigned resources, it does not produce a set of equal, or commensurate
solutions for individual physicians or teams. Accordingly, physicians and teams, who represent
the clinical decision makers in this problem, are not free to individually choose which of the two
solutions (volume or cost model) they prefer. This decision c m be made only by, or on behalf of,
ail physicians or teams as a group. In the example listed in Table 6.12, if the GI Team prefers the
allocation implied by the volume model solution, the Oncology Team cannot choose the allocation
implied by the cost model, since the GI Team OR allocation declared "surplus7'in the volume
model and reallocated to the Oncology Team, will now be consumed by the GI Team. Because
physician and team autonomy is an important concept within health care, it may be unredistic to
assume that a consensus on either of the resource allocations can always be reached arnong al1
physicians and/or tearns.
When initiai model results were s h o w to key decision makers, severai expressed doubts
that physicians would always be N i n g to "subsidize" their colieagues' practice, particularly if
subsidies crossed group practices or clinical expertise. Instead they suggested that the model be
revised so that comrnensurability can be maintained between the two sets of recornrnendations.

Table 6.12 Example Scenario Results: Non-cornmensurate Recommendations Between Volui

Sumrnary of Physician Prelerences

and Cost Model

Summaty of Cosi Model Recmmendalions

Summary of Volume Model Recommendalions

k

Bed

OR

Dga

Time

4,077

12,173

222.870

1.823

1,948

66,243

Doc
Bibngs

Oveci
OxI

1 2,124.680

$ 4.122,967
2

640.592

840,838

132

By revising the equity constrahts in the cost model, so that each physician/team receives
no more resources than were assigned by the volume model, cornmensurate solutions cm be
enforced between the two models. This revision, in which physicians and tearns are assumed to
compete individually for resources, is realized through the foliowing changes to cost model equity
constraints:
(6.5) Cost Mode1 Revisions to Enforce Cornpetition Between Physicians
.v

16)

C pJ,,

Beds*,-,

for aii y E D,

a = I

.v

26)

,
' , O~~A'II~,
O

for ail y E D,

a - 1

,v

30')

p,i($,

- a p - y,l,,)

r Direct Cost6k Jbr di y E

D,

n-v E D,;n

n - l

Revised constraints lb and 2b restrict cost model recomrnendations so that the total
number of bed days and operating room minutes assigned to each physicianklinical team is less
than or equal to that assigned by the volume rnodel. Constraints 3a and 3a' restnct model
recommendations for direct case cost reductions such that the sum of the variable cost of
producing a physician's preferred case mix is less than or equal to the minimum of the direct cost
assigned by the volume model, the direct cost irnplied by the volume model's allocation of
contribution, or the direct cost implied by the volume model's allocation of bed days.
In Table 6.13 output fkom the model, nin under the assumption of cornpetition between
individuai teams, is listed. As c m be seen fiom the table, solutions between the two models are
completely cornmensurate; teams may individually select either the cost mode1 or the volume
model recomrnendations, without affecthg the choices of their peers.

Table 6.13 Examde Scenario Results: Cornmensurate Recommendations Between Volume and Cost Mode1

Summary of Physician Yre ferences

Casa

Bcd
Days

OR

Doc

Dircc!

Timc

Billings

Cost

Summaq of Vulunie Mode1 Recomrnmdaiions

Cascs

Bcd
Dap

OH
Timc

Dircct

Doc
Billing

Summlry of Cos1 h l d e l Kccommnidilionr

Casa

Cor!

OR

Bcd
Da)%

Doc
Billings

Dircct

Timc

S 2,124,680

S 4,122.997

Cmt

Gl Tcam

4.077

12,684

222,870

S2.124.680

$4,574,532

4,619

1 1,668

226.787

5?,124,680

4,122,997

4.077

1 1,668

222.870

General&
Spccially SUE

1,823

1.949

67.142

5 M0,532

51,232,308

1,938

1,612

66.243

S M.592

84,U38

1,823

1,612

64,243

htwulmkeleial

1.1W

6,786

110,939

SI ,090,653

S5.025.525

1,359

6.016

120,362

Sl,OW.653

2,335,152

1,194

6.016

120.362

5 1,030,653

$ 2,335,152

517

1,297

32,800

S 301,918

5 640,669

631

1,132

26,913

S301,918

489,898

517

1.132

26,913

S 301,918

Oncology

3.535

17,684

325,785

S2,798,559

S7JM.516

4,430

15,671

295.722

42,798,559

5,776.503

3,535

15.671

295.722

$ 2,798,559

S5,776.503

DcnEyc/ENT

3.114

1.563

206.167

SI,W9,208

53,793,919

3,078

2,132

189.675

S1,5-19.208

1,345,016

3.1 14

2,132

189,675

S I.E19,?08

S 1,345.016

W.592

84,838

Thoracic Surg

1 Toiab

14.260

1 42.%3 1 975,702 1 58,505,610 1 S2,571,490 1 16,055 1

38,231 1

915,701 1 S8,505,610 1 S 14,910,W 1

14,260 (

489,898

38,231 ( 921,784 1 S 8,505,610 1 S14,910,4Q1 [

134

Under an assumption of competition, however, the cost mode1 is no longer pareto optimal
with respect to the volume model; it can no longer be guaranteed that no individual is any worse
off under cost model recommendations than he/she was under volume model recomrnendations.
Indeed, given that resources allocations under the cost model are constrained to be less than or
equal to those assigned by the volume model, it is reasonable to assume that solutions to the
volume model will not be pareto optimal under the assumption of competition. In addition,
dissatisfaction with the solution, as measured by the objective functions of the pre-emptive goal
program, is greater under the assurnption of competition (see Table 6.14). More significantly,
under this operating assumption, if al1 teams choose to adopt the cost model recommendations,
perverse situations can occur. Under the assumption of cornpetition, some teams may be forced
to reduce their consumption of a particular resource, even though other tearns may be over
supplied.
Table 6.14 Goal Programming Penalties: Cornpetition vs. Cooperation.
Goal

1 1 - Min Changes to LOS

1 2 - Min Changes to OR Tirne

3 - Min Changes to Direct Cost

Penalty under
Cornpetition

Penalty under
Cooperation

0.08

3.50

0.57

4.23

--

500.18

500.18

Despite the drawbacks associated with the assumption of competition between physicians
and teams, it remains an important component of decision rnakers' rationale when determining
resource allocations. It was therefore decided to maintain the functionality of both competition
and cooperation in the software implementation of the cost model, since both represent valid sets
of constraints and are of practical and theoretical importance. Results fiom production runs of
the model, however, will be reported for the assumption of cooperation

135

6.9 Mode1 Revisions to Run Under a Global Budget

The resource allocation model, as defined in Chapter 5, is predicated upon the assumption
of a constrained rate-based method for hospital funding. However, in early interviews at the
study site, decision makers specified a desire to adapt the model to function under the assumption
of a global budgeting methodology. Given the province's delay in implementing rate-based
model, and the need of the hospital to deal with budget reductions in the immediate future,
decision rnakers felt the model would be more appropriate if it could inction under both funding
methods.
Changes to the model defined in Chapter 5 to accommodate a global budgeting
methodology, are relatively minor; conversion of the two goal programming rnodels consists of
deleting constraints related to revenue generation or contribution and substituting the global
budget for the revenue cap as the definition of model parameter G. Accordingly, the following

constraints are deleted from the volume model, when run under the assumption of a global
budget:
(6.6) Volume model constraints deleted under an assumption of global budgeting.

Under a global budget, the parameter G in volume model Constraint 2 represents a limit

on the total costs of production. Under this definition of G, Constraint 2 is a soit bound imposing
penalties (d,') when the total costs of production (F + 1, ,v x.J are greater than the giobai
budget minus the desired institutionai profit (Profit):

2) F

v,,,,~,,,,

di -

di

- Profit

n = L y = l

In the cost model, the following constraint is deleted when run under the assumption of a
global budget:
(6.7) Cost model constraints deleted under an assumption of global budgeting.

The definition of model parameter G in cost model Constraint 4, which restrcts total
spending, rnust also be r e d e h e d to be equal to the global budget, rather than the provincially
imposed revenue cap. Thus, cost model Constraint 4, imposes penalties (dGIobd-)whenever the
total institutional cost of production:

exceeds the global budget less the profit allotted by the volume model.

7.0 Validation
Although validation, the measure of the extent to which a model represents truth
(Saunders 1985), is critical to the success of any decision involving abstractions from reality, there
is a paucity of research on validation in the literature (Finlay and Wilson 1990). Law and Kelton
(1991), note that validation is among the most diicult components of an operations research
study, since no forma1 methodology can be applied in ail cases and no absolute measure of validity
exists for cornplex decision-making models (Law and Kelton 1991).
Because the ultimate objective of operations research models is to influence the behaviour
of a decision maker or decision making group, many authors argue that acceptance of a model's
output by the key decision maker constitutes de facto validation. Validation, according to this
definition, is a process of interacting with decision makers over the life-span of a model to garner
their input, capture their decision making preferences, and build their confidence in the model's
results (Butler 1995; 1-0

and Cavalier 1994).

While noting the importance of acceptance as a measure of validity, other authors argue
for a structured approach to validation that critically investigates the elements constituting a
model. Schellenberger (1 974) argues models must exhibit technicai validity, operational validity
and dynamic validity to be considered usable decision making aids. Technical validity is defined
as a reasonable set of critena against which a model may be compared. Technical validity can be
sub-divided into four sub-components:

i.

ModeVtheoretical validity: The degree to which the underlying conceptuai model

used tu develop a decision making aid represents reality. Model validity
encompasses mathematicai, content and causal assumptions irnplicitly or explicitly
inherent in the structural component of a decision making aid.

ii.

Data vaiidity: The degree to which the information used to define an instance of a
decision making problem is representative of reality. Data validity encompasses

the accuracy, irnpartiality and the representativeness of data, as well as the

processes used to aggregate raw data into a format suitable for use in a model.
Logicai validity: Logicai validity, or model verincation, describes the fidelity with
which a conceptual model is convened into an operational decision-making aid.
Logical validity typicaliy involves processes to ensure a conceptual model is
properly translated into a working cornputer program.
Predictive vaiidity: Predictive validity refers to the ability of a model to produce
results that conform to expected outputs. Predictive validity encompasses both the
accuracy of a model in test cases and in cornparison to the real world system.
Operationai validity, according to Schellenberger is a subjective process of assessing the
impact of anomalies discovered during technical validation of model results. Operational validity

i.

..

il.

...

iii.

Degree of improvement: The relative improvement suggested by a rnodel provides

a measure of model robustness. If, for example, a model recommends a course of
action resulting in a 60% improvement in cost or throughput, errors of 5- 10% in
the input data Stream are unlikely to be operationaily significant.
Mode1 sensitivity: The effect of small changes in data parameters in the area of an
optimum gives an idea of model stability. The range of values over which a
recornmended course of action remains optimal gives a measure of model
robustness.
Implementability: To be of use as a decision making aid, a model must produce
recommendations that are achievable within the context of the real world system.
Recornmendations that are not implementable may indicate shortcomings in the
content or data validity of the model.

Dynamic validity refers to the utiiity of a model over an extended period of time. To be
dynamically valid a model must be easy to maintain. A process for periodically reviewing the
output from the model and comparing it to results from the real system must also exist for a
modei to be considered dynarnically valid.
Gass (1983) describes an application of the Schellenberger fiamework for social policy
analysis models, but notes that it is applicable oniy to existing systems. For non-existing systems,
Gass recommends validation be based on face, data and hypothesis validity.

Law and Kelton (1 991) argue, however, that validation is only one component of
establishing model credibility with decision makers. Law and Kelton describe a three part process
for buildiig credible models:
Validation of the underiying conceptual model: Al1 models are simplifed
abstractions of redity, based on assumed causal relationships. The key
assumptions defining the conceptual model upon which decision making tools are
based must be validated. Law and Kelton suggest a "stnictured walk-through" of
al1 implicit and explicit assumptions with system experts.
Verification of the operational implementation of the model: Model verification is
a process for ensuring a faitffil translation of a conceptual model into computer
code and cannot be boiled down to a simple test. Law and Kelton argue, instead,
for phased development. Moreover, they suggest a detailed review of computer
prograrns by independent agents to eliminate the possibility of "group think. As
an objective measure, Law and Kelton suggest running the model under sirnplified
conditions for which the true characteristics of the real system are known or cm be
extrapolated.
Validation of the operational model: Law and Kelton argue models should have
"high face-validity". To achieve this they suggest:
1)
2)
3)
4)
5)

Vet system results with expens.

Compare model results against system results, if possible.
Compare model resuits against theory or existing models.
Ernpirically test rnodel assumptions and input data (sensitivity analysis).
Determine the representativeness of model results.

7.1 Validating the Model

The resource allocation model presents a number of challenges that make validation
difficult. The validation fiameworks described in the literature (Schellenberger, Gass, Law and
Kelton) are primarily intended for predictive models. Validation in this context is essentially a
calibration process: A model is built, run and tested. The output of the model is compared to a
"gold-standard represented by the real world system. Once obvious errors have been identified
and addressed, the model is assumed to be valid. Any discrepancies between the model and the
real world are then eliminated using explicit or implicit correction factors, in much the same way

140

that a set of scales is calibrated to reflect the standard defined by a platinum bar in a museum in

Paris. Law and Kelton (199 1), however, argue against the use of explicit correction factors to
calibrate a model, noting that arbitrary corrections lead to a mode1 that is representative of a
particular data set, rather than the system in general. Models retain higher face validity, they note,
if correction factors are avoided.
While the Law and Kelton prohibition against explicit correction factors is correct, their
premise is flawed. All efforts at validation produce models that are reflective of a particular data
set rather than the system in general. In as much as a particular data set is reflective of the real
system, the validated mode1 reproduces reality with a high degree of fidelity. Take, for example, a
rnodel of an assembly line producing cans of soup. The model is built, nin and tested against a set
of data representing the operation of the assembly line for some arbitrary period of time. M e r al1

the bugs are worked out, the model is said to be valid and can be used to predict the behaviour of
the line at points in the future. Of course there is a limit to the predictive ability of any model.
since the underlying process is non-time stationary. Machines age, production characteristics
change, exogenous events (earth-quakes, floods, war and famine) occur, personnel are hired and
fired. and management policies and structures are updated and renewed. In short, the real-world
system, rather than being a gold standard, is itself in a constant state of change. A model
validated against data from one period, may therefore, look quite different from a mode1 validated
against data from a different period. This may be reflected in the data used to drive a model, or
in stmctural changes to the model itself. Attempting to resolve this issue by extending the

validation period will, of necessity, Iead to models with increased complexity, but will not address
the basic issue: The validated model is reflective of a specific period in time (or more generally, a
complex and perhaps unquantifiable set of operating conditions) rather than the absolute

characteristics of the system. Whether acknowledged explicitly or not, al1 models contain
calibration factors that relate to the generd milieu in which a model is designed and built as well
as the standard against which it is tested. Schelienberger's notion of dynamic validity hints at the
fact that models cannot, and do not, remain valid forever. They are an abstraction of reality.
Even if this abstraction is calibrated in such a way that it reflects reality at some particular point in
time, reality continues to evolve. Hence even the best model diverges from truth over time.
Extending the idea of validation beyond descriptive models to prescriptive or normative
models presents additional difficulties. If the purpose of a model is not to describe what is, but
rather to describe what could be, or what should be, comparing model outputs against existing
system values is either impossible or meaningless. Thus there is no "gold-standard" against which
the resource allocation model, since it is a normative model, can be adequately compared. The
resource allocation mode1 descnbes a case mix that jointly balances the economic and clinical
goals of hospitals and physicians. If institutions and physicians were able to implicitly optimize
their economic and clinicd goals, the model could be said to act in a prescriptive model. Clearly.
however, the combinatonal complexities of decisions involving thousands of variables and tens of
thousands of constraints, is beyond the scope of unaided decision makers. It would therefore be
incorrect to assume the allocation model describes the behaviour of decision makers at Mount
Sinai Hospital in any way; the mode1 sirnply identifies points on the institution's productionpossibility curve that are assumed to have certain characteristics desired by the institution and its
medical staff.
It would be incorrect to assume from the preceding discussion that model validation is
impossible. However, because of the difficulties inherent in applying normative models to
operational decision making, the approach for model validation must necessarily depart fiom the

142

gold standard of establishg correlation with historical results. Instead, validation must focus on
aspects of technical validity described by Scheilenberger: theoreticai validity, data validity, logical
validity and predictive validity. Validation in this context means that in as much as possible, the
assumptions used to create the model and the data used to define it are reasonable and refiective
of the real world. If the inputs to the model are correct, and the model is faithflly translated into
computer code, the resulting model will necessarily be valid.
7.2 Theoretical Validity

Al1 literahire addressing validation notes the importance of reviewing modelling

assumptions with system experts. Schellenberger (1974) notes the importance of documenting
assumptions during model development so t h they may be presented to system experts; validity

can only be assured if modelling assumptions are consistent with the decision maker's M e of
reference. Gass (1980) also notes the importance of documenting modelling assumptions,
particularly in the case of large, complex systems where external review is required to establish
validity. Law and Kelton (199 1) describe stnictured walk-throughs of modelling assumptions
with system experts as a key component of establishing face validity.
A two phase approach was used to validate the theoretical accuracy of the resource

model. A group of experts fiom the study site, including the vice-president of nursing and the
program director of nursing responsible for utilization management were asked on a regular
basis, to review the mode1 and its assumptions during development. When preliminary results
were available, the expens were asked to review them and provide feedback. The group made a
number of suggestions aimed at refining the model or enhancing the quality of its solutions.
These recommendations were included in the model and are described in Chapter 6. Technical

aspects of the model were reviewed with experts in the field of optimization and mathematical
modelling.
7.2.1 Major Modeliing Assumptions

The major modelling assurnptions underlying the resource allocation model appear below.
These assumptions were reviewed for veracity by both the context and content expert groups.
1)

Decision Makers: The resource allocation problem is assumed to have two distinct
decision making groups: hospital administrators and physicians. Both groups are
assumed to be constrained profit satisficers, rather than profit maximizers.

2)

Case Mix Preferences: The resource allocation model assumes al1 physicians are
able to express preferences for a particular volume and mix of cases. Facilities
exist within the model's interface to ailow users to s p e c e a specific mix for any
physician or clinicai team. In instances where a volume and rnix is not defined, a
default case rnix is assumed fiom the activities of the physician or clinical tearn in a
previous period.

Case mix preference is assumed to include dimensions of preference relating to

research, education, cornrnunity service and strategic initiatives.
3)

Decision Maker Goals - Case Mix: Decision makers are assumed to have three
distinct, but related case rnix goals. Hospitals and their associated medical staff are
assumed, first and foremost to be constrained profit satisficers. Under this
assumption, both decision makers have target income requirements that must be
met to ensure economic viability. For hospitals, income requirements must be
sufficient to cover the fixed costs of production. Physicians are assumed to have
defined target incomes they wish to achieve. Target income may be defined for
individual physicians, or defaulted to the billings generated in the previous period.

Once economic goals have been met, hospitals are assumed to be indifferent to the
actual case mix delivered by physicians. Physicians conversely, are assumed to
desire a case mix which, while preserving their income requirements, ensures that
they are able to secure a minimum slate of resources. Minimum resources are
assumed to be equal to the weighted sum of a physician's preferred case rnix and
his or her minimum allowabie practice pattern. (This goal relates to requirements
for ensuring multiple views of resource allocation and is necessary to support
assumption 7.)
Once a minimum date of resources has been identified, each physician is assumed
to desire a case mix that reflects his or her preferred case mix as closely as is
possible. NI deviations fkom preferred case mUc are penaiized. However, larger
deviations are penaliied more heaviy than smaller deviations.

4)

Case Mix Conlrol: Case mix is assumed to be controiiable. Recommendations for

increases or decreases in case mix are therefore assumed to be implementable
policies for physicians or clinical tearns. (Controiiable case mix, in this context,
may simply mean that excess demand is deferred from one tirne penod to another,
or that patients are referred to other centres.) Al1 deviations from physicians'
preferred case mix are assumed to be undesirable, whether positive or negative.
Positive deviations in this context are assumed to be undesirable because they
represent an unwanted reduction in the capacity of the physician or clinical team to
produce other, desired, cases. Negative case mc deviations are undesirable since
they represent an unwanted decrease in physician billings. Lower bounds on case
mix change are defined by the number of emergent and urgent admissions in the
previous period.

5)

Costs: Costs in the resource allocation rnodel are assumed to be either fixed or
variable. Mount Sinai's case costing system defines four types of cost: fixed direct
cost, fixed indirect cost, variable direct cost and variable indirect cost. Direct
costs are costs that are incurred by direct service departments (nursing,
laboratones, diagnostic imaging, pharmacy and clinical nutrition); indirect costs are
costs incurred by departments that support the direct service departments
(housekeeping, personnel, information systems, etc.). Fixed costs are costs that do
not Vary with the volume of patients treated in the hospital (heating, lighting,
secunty, administration, etc.). Variable costs are costs that Vary directly with
patient volume. An example of a direct fixed cost, therefore, would be the salary
paid to the manager of a nuning unit. A direct variable cost, conversely, would be
the hourly wage paid a nurse working on a nursing unit. An example of a fixed
indirect expense would be the salaries paid to the personnel who maintain the
hospital's physical plant; an example of an indirect variable expense would be the
hourly wages paid to workers who launder bed linens.

In the context of the resource allocation model, only variable direct expenses,
expenses with a direct linear relationship to patient volume, are assumed to be
variable. Ali other expenses, fixed direct, fixed indirect, and variable indirect are
assumed to be period expenses. Because the relationship between variable indirect
expense and patient volume is tenuous, variable indirect expenses are treated as
period (fixed) expenses. This assumption ensures the model makes conservative
estimates of potential cost savings arising from changes in volume.
6)

Practice Preferences: Physicians are assumed to have a preferred pattern of

practice with several dimensions (length of stay, operating room time, variable
cost). This pattern rnay be specified for an individuai surgeon or clinical tearn.
Where no pattern is defined, the model assumes a provider practice from histoncal
records equal to the average resource consumption by provider, by CMG.
A piece-wise linear penalty fnction is used to approximate physician preferences
for clinical practice. Al1 deviations fiom physicians' preferred practice patterns are
assumed undesirable. However, positive deviations in this context do not make

practical sense and are therefore disallowed. AU negative deviations from

preferred practice patterns are penalized in the model. Larger practice change
deviations are, however, penalized more heavily than smaller deviations.
7 ) Minimum Resource Repirements: The volume and cost models provide
cornmensurate views of resource allocation. The volume model sets quantity,
under the assumption of fixed product price, while the cost model sets produa
price, under the assumption of a iked quantity of product. To ensure feasible cost
model recommendations, it is assumed physiciandclinicai tearns desire a minimum
set of resources. This assumption, which is motivated by technical issues in the
solution technique, ensures in most instances a cornmensurate cost model solution
can be found once a volume model solution has been identified.

8)

Reiatzonship Between LOS and Direct Nursing: Mode1 recommendations for

changes in length of stay (LOS)are based on bed availability. In praaice,
recommendations for reduced lengths of stay are driven by a desire for decreased
cost or increased turnover. For historical reasons, however, LOS has been used a
proxy for cost and a measure of efficiency in acute care hospitals. Hospitals have
therefore focused on reducing LOS as a method of decreasing cost. (As LOS
declines, fewer beds and nursing staff are required; costs are therefore assumed to
decrease.) The model, rather than emphasizing an intermediate product, such as
beds, makes economic recommendations on the basis of cost a d o r revenue
directly. Recommendations for changes in LOS are, therefore, strictly based on
bed availability. In instances where the cost model recommends decreases in
length of stay, however, it is clear that some decrease in the direct cost of patient
care should be expected. The direct cost of a CMG is assumed to decline by a
proportion, 'f,of the average daily nursing cost with each decrease of one day in
LOS. Since the value of Y is not known and cannot be directly calculated From
current costing information maintained by Mount Sinai a value of 0.75 is assumed
for Y.

7.3 Data Validity

Saunden (1985) notes results from a model are only as good as the data used to drive it;
even the most theoretically perfect model is useless without good data. Ensuring good, valid data
is a key component of creating credible, useful model-based recommendations. The data used to
drive the resource allocation model cornes from credible, primary sources (Mount Sinai Hospital's
case costing and case costing system, the Ontario Ministry of Hedth, and the Ontario Case
Costing Project) and can therefore be considered representative. Al1 data used to develop the
model was derived fiom electronic records; no data transcription was required. Thus, the

146

underlying data is 6ee of transcription errors. The accuracy of information used in the resource
allocation model, therefore, is a fnction only of the accuracy with which it is coiieaed at source.
Coaing information accuracy is directly related to the accuracy and vaiidity of the
information resident in Mount Sinai's case costing systern. Because case costing is a new concept
for Canadian hospitals and operationai expenence with this technology is limited, Mount Sinai has
experienced some difficulty maintaining consistent costing information fiom year to year. As
experience with the case costing technology increased, the methods for assigning costs to
individual cases have been improved. in particular, the methods for assigning overhead (fixed
costs) have undergone several revisions over time. Changes in overhead assignment create
discrepancies in total case cost. (Overhead assignment is generally assigned on the basis of
workioad, through a complex process known as the simultaneous equation assignment method, or
SEAM. The cost of housekeeping services, for instance, are assigned to vanous front line

departments on the basis of total floor space a department occupies. Total costs for al1
departmental overhead expenses are then allocated to individual cases in relation to the proportion
of total direct workload a department expends on a particular case.) Thus, while the institution's
total overhead costs have remained stable fiom year to year, the allocation of these costs to
particular cases has changed. This problem, however, is inconsequentiai for the resource
allocation model since al1 costs not directly attributable to a given patient ( h e d direct cost, fixed
indirect cost, and variable indirect cost) are considered to be period expenses and therefore do not
vaq with patient volume. Assuming al1 expenses not directly attributabie to a patient to be period
expenses eliminates data validity issues associated with changes in overhead allocation methods,

since period expenses are considered in the aggregate by the model; penod expenses of $21

147

million remain period expenses of $21 million, regardless of the technique used to assign overhead
to individual cases.

In addition to revisions in the methodology used to assign overhead to individuai cases,

there has been an evolution in the standards used to define the intermediate products produced by
direct departments for individuai patients. Mount Sinai separates five direct service departments
(nursing, nutrition, pharmacy, labs, and diagnostic imaging) from al1 other departments. Each of
these direct departments produces intermediate produas, which are a workload unit (an hour of
nursing work, for instance) or an identifiable supply with a value greater than $250. (A hip
prosthesis is considered to be an intermediate product under this definition, but an aspirin is not.)
There are severai thousand intermediate products defhed in the hospital's case costing system,
each of which is detined by a clinicd standard, analogous to a bill of materials in a manufacturing
environment. Recent reviews of the standards used at Mount Sinai to define intermediate
products have suggested that errors may exist in the laboratory standards. However, laboratory
expenses account for only 8.83% of the total direct expenditures of the Surgical Council at Mount
Sinai Hospital. Assurning an error of 10% in laboratory standards, would suggest a total error of
only 0.883% in the direct costs used to develop solutions in the resource allocation model. The
vast majority of direct costs (75.91%) associated with providing care at Mount Sinai is
attributable to nursing, which has had relatively stable standards for intermediate products over
the past several years.
Given the high degree of accuracy in the costing data and the robustness of the model to
changes in overhead validation, validity must, therefore, be established only for parameters with
uncertain values: physician billings and Y, the proportion of daily nursing costs saved with each
marginal decrease in length of stay.

148
7.3.1 Mode1 Sensitivity to Changes in Physician Bling Estirnates

As noted in Chapter 6, physician billing data was obtained from the Ontario Ministry of
Health through the Freedom of Information Act for a sarnple of 8,039 patients discharged fiom
Mount Sinai Hospital between September 1994 and November 1994. Because of coddentiality
regulations, data could not be obtained on a physician level basis. (Billing information is so
carefully guarded that the Minister of Health was forced to resign in December, 1996 when a
spokesperson inadvertently revealed the name of the physician with the largest billings in the
province during a press conference. In response to the leak, the president of the Ontario Medical
Association noted that "no one, but the director of OHIP should know who the most expensive
doctor in the province is." (Coutts 1996)) Instead, the Ministry of Health provided the sum of
al1 provider billings associated with each of the 8,O39 patient's OHIP numbers between their dates
of admission and discharge from Mount Sinai. Thus, the data used to represent physician billings
in the model is clearly a proxy. However, we note that a more accurate estimate for physician
billings in Ontario is not presently available and suggest that the methodology employed in this
study is more ngorous than that comrnoniy described in the literature. There is, unfortunately, no
way to objectively validate the billing information obtained fiom the Ministry of Health without

violating physician a d o r patient confidentiality.

Schellenberger ( 1974) and Law and Kelton (1992),suggest sensitivity analysis be
conducted on rnodel pararneters for which uncertainty exists. By varying uncertain parameters,
they note, model response can be analyzed under a variety of scenarios. Drurnrnond, Stoddart
and Torrance (1987) argue sensitivity analysis is the on& technique that can be used to validate
cornplex medical-economic decisions involving uncertain parameters. If uncertain pararneters
infiuence the recornmendations returned by a model, validity becomes a question of determining

whether uncertainty represents a significant operational error. Therefore, in accord with

Schellenberger's recommendations, the mode1 was run under difZerent parameter settings for
physician billings. The P value (direct cost) of the regression coefficients for the inpatient billing
regression were varied by

* 10% from the values iiaed in Table 6.2 and the results were

compared against a base set of runs. To clearly gauge the effect of billing error estimates on
mode1 recomrnendations, the absolute value of the dif5erence between the base case and each test
case was recorded at the level of case rnix group. (Absolute value is used since case mix changes
arising fiom different P values are both positive and negative. Over a large group of CMG's,
recomrnendations for increases tend to balance recommendations for decreases, thus
underestimating the influence error in the regression parameters.) Changes in case mix arising
fiom differences in P were surnrned for each clinical team and divided by the total sum of
recommended values for each team, to produce an estimated sensitivity figure. Case mix
sensitivity therefore is:
(7.1) Case Mix Sensitivity

Case Mix Sensztzvity: = " =

'

'V

Where:

xn = Test case mix

in= Bme m n case mzx

The sensitivity of practice parameters to changes in regression coefficients, however, is

calculated on the bais of total resource requirements. Model recomrnendations are multiplied by
base case volume and summed to produce total resource requirements for each clinical team. The
absolute value of the difference in resources required between the base case and the test case are

150
then recorded. These values are sumrned for each clinical tearn and divided by the total resources
required by the team to produce an estimate of model sensitivity to error in physician practice
parameters.
The results of the sensitivity analysis on inpatient physician billing parameters appear in
Table 7.1. The analysis is based on responses fiorn a representative model run in which the
institution is required to reduce total costs by $3.0 million. Of this arnount, $1 .O million is
assumed to have been removed fiom fixed expenses by removing 2,000 bed days and eliminating
50,000 minutes of operating room time. Runs under both global budgeting and rate-based
pajment models of funding appear in the tables. Al1 mns assume cooperation between physicians
and a subsequent sharing of surplus results. See Tables 7. la and 7. lb.

Table 7.la Model Sensitivity to Changes in Inpatient Biiiing Regression Coefficient P

Direct Cost)Und r Global Funding
-

% Change in
Volume Model
Case Mix

% Change in Cost
Model Bed Days

% Change in Cost
Model OR Time

% Change in Cost
Model Total Direct
Cod

Gen 8 Spec Surg

Musculoskeletal

Thoracic
Oncology

Table 7.lb Model Sensitivity to Changes in Inpatient Billing Regression Coefficient P

Direct Cost) Under Rate-Based Funding.
% Change in Cost
Mode1 Bed Days

% Change in
Volume Model
Case Mix

Gen 8 Spec Surg

0.53%

0.20%

0.06%

Musculoskeletal

0.11%

0.34%

1.38%

Thoracic

0.63%

0.90%

1.19%

% Change in Cost
Model OR Time

% Change in Cost
Model Total Direct
Cost

0.37%

1.11%

Oncology

1.88%

1.91%

1.78%

0.04%

DenEyeIENT

0.85%

O.=%

1.15%

2.59%

The results of the sensitivity analysis on inpatient physician billing regression parameters
listed in Table 7.1a and 7.1b indicate model case rnix recomrnendations are relatively insensitive to
changes in value of P under both funding mechanisms. (Model results are, in faa, identical. A
detailed discussion of this observation is provided in Chapter 8.) A maximum change in clinical
team case mix (both positive and negative) of 2.38% is observed when the value of

P,the

152
regression coefficient on physician billings for inpatients versus hospital direct cost, decreases by
10%; a maximum change in clinical team case mix of 3.38% occurs when the value of P increases

by 10%. The impact of change in the value of inpatient billing regression parameter P on
physician practice is of the same magnitude. A maximum change in the number of bed days
assigned clinical teams of 3.78% is observed when P decreases by 10%; a change of 3.46% is
observed when

P increases by 10%.

The total amount of operating room time assigned to clinical

teams varies by a maximum of 2.45%when P decreases by 10% and by 4.05% when P increases

by 10Y0. The total direct cost dollars assigned to clinical teams is also relatively insensitive to
changes in the value of P; when

decreases by IO%, a maximum change in the total direct cost

dollars assigned to clinicai teams of 3.97% is observed, while a maximum change of 2.05% is
observed when P increases by 10%. From this analysis, it can be concluded that the mode1 is
insensitive to error in the regression analysis used to estimate physician billings for inpatient cases.
A sensitivity analysis was also run on the regression coefficients used to estimate physician
billings for outpatient services. The values of Pl (direct cost) and
varied by *IO% fiom the calculated values listed in Table 6.2 (Pl

P,

(hospital funding) were

= 0.408;

P,

= 0.254).

In

addition, the value of a, the y-intercept of the regression equation, was set at O as well as the
upper lirnit and the lower limit of the 95% confidence interval identified in the regression analysis

(a = -3.759; CI = 1-22.866, 15.3471). See Tablzs 7.2 - 7.4.

Table 7.2a Model Sensitivity to Changes in Outpatien Biing Regressioi Coefficient Pl

(Direct Cost)Under Global Funding.
- -

% Change in Volume
Mode1Case Mix

% Change in Cost
Model Bed Days

Ge" 8 Spec Surg

1.11O h

0.30%

213%

0.32%

Musculoskeletal

0.01%

0.01%

1.64%

0.81%

Oncology

0.31%

0.Ol0h

0.78%

0.42%

Change in Cost
Modei OR Tirne

Oh

Model Direct Cost

Table 7.2b Model Sensitivity to Changes in Outpatient Biliing Regression Coefficient Pl

(Direct Cost) Under Rate-Based Funding.
Oh

1
1

Musculoskeletal
Thoradc

1
1

Change in Volume
Modei Case Mix

0.01%
0.02%

% Change in Cost
Model Bed Days

% Change in Cost

Model OR Tirne

% Change in Cost
Model Direct Cost

Table 7.3a Model Sensitivity to Changes in Outpatient Biiling Regression Coefticient P,

Potentiai Revenue) Under Global Fundinp;.
% Change in Volume
Modei Case MD(

-IO%&

Team

Change in Cost
Modd Bed Days

% Change in Cost
Model OR Time

Oh

% Change in Cost
Model Direct Cost

+IO%&

Gl

1.18%

0.87Or6

Gen & Spec Surg

1-13%

0.35Oh

Musculoskeletal

0.08%

0.02%

Thoracic

0.04%

0.11%

Oncology

1.26%

O.3g0h

DentlEyeiENT

0.57%

0.08%

Table 7.3b Model Sensitivity to Changes in Outpatient Billing Regression Coefficient P,

ff otential Revenue] Under Rate-Based Fundine.

I
1 Gen

8 Spec Surg

Musculoskeletal

Oh

Change in Volume
Model Case Mix

1.13%
0.08Oh

0.35%
0.02%

% Change in Cost
Model Bed Days

% Change in Cost
Model OR Time

% Change in Cost

Model Direct Cost

Table 7.4a Model Sensitivity to Changes in Outpatient Bling Regression y Intercept a

Under Global Fundin~.
% Change in Volume
Mode! Case Mi

iange in
lei Bed t

tange in Cod
IdOR Time

iange in

4 Direct
O

0.43%

3.37Oh

Table 7.4b Model Sensitivity to Changes in Outpatiend Billing Regression y Intercept a

"-teBase(
Oh Change in Cost
Model OR fime

The sensitivity analysis on outpatient billings indicates the allocation mode1 is insensitive to
error in the values of the regression coefficients used to estimate physician billings for outpatient
cases. Case mix, in particular, was observed to be insensitive to error in the outpatient regression
analysis. A maximum change in case mix of 1.34% is observed over the entire sensitivity analysis.
Maximum change to case mix was observed to be similar with respect to change in al1 three of the
outpatient regression analysis parameters P, (direct cost), P, (potentiai revenue), and a (yidercept). Maximum change to case rnix varied fiom 0.01% - 1.1 1% as the value of P,, the

156
regression coefficient on direct cost, varied by *IO% from its calculated value of 0.408.Change
to case rniu ranged fiom 0.02%-1.26% as

a the regression coefficient on potential funding

varied from 10% of its set value of 0.254. Change to the y-intercept of the regression on
outpatient biilings produces slightly larger variations in case rnix. Case mix change varies fiom

0.000/0
to 1-34% as the y-intercept of the regression equation ranges f?om -22.9 through 1 5 -4.
The effect of error in outpatient billing regression coefficients on model practice
recomrnendations was also observed to be relatively minor. As the value of pl, the regression
coefficient on direct cost varies fiom 0.367to 0.449, changes in model recommendations for bed
days range from 0.32% - 2.78%, operating room time recommendations Vary fiom 0.03% to
1.49%, and direct cost recommendations fluctuate between 0.0 1% and 4.34%. The effect of
changes in

P , the regression coefficient on potential revenue, are sirnilar.

As Pz ranges from

0.229 to 0.279, model recommendations for bed days Vary fiom 0.02% to 2.72%, operating room

time recommendations Vary between 0.05% and 1.42% and direct cost allocations vaiy between
0.07% and 4.46%.

Error in the estimates of the y-intercept of the regression equation used to determine
physician billings for outpatient cases produces slightly larger changes in cost model allocations
than do errors in the regression coefficients. Nevertheless, the absolute impact of error in the
value of a,is still inconsequential. As the value of a ranges between the upper and lower limit of
the 95% confidence interval calculated in the regression analysis, model allocations of bed days to
clinical tearns varies by 0.32% to 3.57%, on a team by team basis. Operating room t h e
allocations Vary between 0.03% and 1.32% over the same range of a,while changes to model
recornrnendations for direct cost Vary between 0.07% and 4.66%.

157

Overall, the sensitivity analysis indicates the rnodel is, when viewed at the clinical team
level, insensitive to error in the regression parameters used to calculate physician incorne. Error
rates observed in the analysis were always less than the assumed error in the value of the
regression parameters. The largest error observed, 4.66%, is Iess than halfthe assumed error

rates of 10% used in the analysis. Thus, mode1 results cm be judged to be at least as accurate as
the regression analysis used to calculate physician billings. The statistical significance of the
values of p in both the inpatient and the outpatient regressions allows error associated with these
parameters to be discounted. The most significant error in the regression analysis is the value of

a used in the outpatient billing regression.

7-32Mode1 Sensitivity to V
As average patient length of stay decreases, there is generally believed to be a

corresponding change in average case cost. In Chapter 5, it was noted that the exact nature of
how case cost changes with length of stay cannot be estimated, since current information systems
cost patient encounters retrospectively using projected workload to distribute common costs.
This methodology produces an average encounter cost for patients belonging to a particular case

mix group, fiom which a typical cost for an average day's stay can be caiculated. Nevertheless.
O'Brien-Pallas et al. (1 992) have shown actud daily nursing workload decreases over the course
of a patient's hospital stay. Thus, assurning savings in direct cost equivalent to the average daily
nursing cost within a particular CMG rnay overestimate actual costs, if care given in the later days

of hospitalization is eliminated via changes to patient length of stay. In Figure 7.1, a hypotheticai
distribution of daily nursing workioad over an eight day length of stay, is shown, dong with the
average daily nursing workload. If reductions in length of stay result in the elimination of day 8,
average daily nursing cost will overestimate actual workload savings. If care remains constant,

158

but patient length of stay is simply shortened, average daily nursing costs may again overestimate
actual cost savings, since the total workload remains the sarne.
Figure 7.1 A Hypothetical Relationship Between Nursing Workload & LOS.

Nursng Workload NurSng Worldod

.W..,,....,#...

To test the effect of change in 'P. the proportion of daily nursing costs saved with each
marginal change in length of stay, a sensitivity andysis was run on the resource ailocation model.
The analysis assumes a representative scenario in which the institution is required to reduce total

revenues by 18% or $7.11 million. Of this amount, $3 -89 million is assumed to have been
removed from fixed expenses by removing 7,734 bed days and elirninating 170,497 minutes of
operating room time. To test the sensitivity of Y,the proportion of average daily nursing cost
saved with each day's reduction in average length of stay, was varied systematically between its

i 59

upper and lower bounds (1.0and 0.0 respectively) and compared against runs assurning 'f = 0.75.
These bounds represent practical limits on plausible change to direct cost arising fYom reductions
to LOS. Runs were made under the assumption o f rate-based fnding and cooperation between

physicians. The resulting change in cost mode1 recommendations for length of stay and direct
case cost appear in Tables 7.5a and 7.5b.

The results of the analysis indicate cost model recommendations regarding direct cost are
sensitive to changes in the value of 'P. Under certain conditions, model recornmendations for
total direct cost dollars assigned to clinical tearns when 'P > 0.75 differ significantly from total
direct cost dollars assigned at values of

s 0.75. In particular, ifbed availability is tightly

constrained and Y is greater than 0.75, volume model allocations of beds, when translated into
practice changes, result in significantly different recommendations than when 'P s 0.75
The rationale for this observation is as fouows. Each change in length of stay made by the
cost model implies a reduction in the direct cost of producing a given case. Savings due to
reduced LOS are equal to a portion (T) of the average daily nursing cost (v,,,,

/ LOS, ).

Thus, the total savings anticipated for any given physician due to model recommended changes in
length of stay is 1,p,*(LOS,

- U, 'f(v
,,,

preference for direct cost dollars is 6 p',

1 L O S 3 or

v,

1,p,'(q,

- y&.

the maximum arnount of direct cost dollars

assignable to any physician, after LOS reductions are considered, is 1,p,' vp

- 1,p,[(LOS, -

LOS$]. However, it must be noted that cost model constra.int3ar,

IJ T*(v,,-,/

3 4

Since physicians'

C PJ~,,

n = 1

- a,

,,Y 1

) r Direct Cost;,,

Direct

for d l y E D fI y E DB;n

COS^,,,,

Y
Y
- ContributionAll,,,
,
Direct Cos<$,n= MN>' RevenuePrdfimed

Direct costPtefe,-

CP&
n-i

Y,I,)

i f y E D*

162

the equity constraint on direct cost dollars, places limits on cost model recommendations. In
generai, Constraint 3a' ensures total amount of direct cost dollars assigned to a given physician is
at least as large as that assigned by the volume model (Direct Costa&).

in some circumstances,

however, it is impossible to guarantee physicians commensurate direct cost dollar

recomrnendations under the volume model (See Section 5.2.4). Thus, the lower bound on total
direct cost assigned by the cost model, as s h o w in Constraint 3af, is the minimum of the direct
cost assigned by the volume model, the direct cost implied by the contribution assigned by the
volume model (Revenu&tfmdy - Contrib~tion,,~)'), or the direct cost implied by the volume
model's allocation of bed days ( R e v e n -~ ~ p,[(LOS,
~ ~ ~ ~- l,J 'f*(v,
/,,,

LOSd]) . As

the value of T varies, the direct cost implied by the volume model's allocation of bed days
changes. When the value of 'f is large and beds are tightly constrained, the lower bound on the
minimum arnount of direct cost assigned to a physician by the volume model changes to reflect
the minimum direct cost implied by LOS reductions. Thus, the model exhibits sensitivity to the
value of parameter Y;as the value of 'f varies between 1.O and 0.0, cost model direct cost
recomrnendations change significantly, due to implied changes in cost created by reduced LOS.
As s h o w in Table 7Sb,mode1 recomrnendations for certain clinical teams, when 'P = 1.O may

Vary by as much as 15% fiom recommendations made when 'P o 0.75. Furthermore, an analysis
of individuai rnodel recommendations indicates that while the total direct cost dollars assigned to
clinical tearns at values of Y s 0.75 remains constant, specific model recomrnendations for direct
cost change as the value of P varies (see Table 7.6).

Table 7.6 Cost Model Sensitivity to Chan{ :s in V - Specifk Recomrnendat

LOS lecornmendations

Direct Cj

5t

Recommendations
-

Ui

# Cases
Between
Hard 8
Sof
Bound

# Cases
Between
Soft
Bound &
Preferred

# Cases
at
Preferred

# Cases
Between
Hard 8
Soft
Bound

# Cases
Between
Soft
Bound 8
Preferred

# Cases

at
Preferred

General8 Speciaky
Surgery

Musculoskeiebl

Thoracic

Oncology

In Table 7.6, aggregate results of the sensitivity analysis on Y,in terms of specific model

recommendations, are listed. As can be seen frorn Table 7.6, model recornmendations for changes
to particular cases tend to fluctuate as the value of 'P varies. Of particular interest are the shifts in

164

model recommendations for length of stay. As the value of 'P decreases, there is generaily (but
not always) a decrease in the number of cases where model recommendations for length of stay

' = 0.25, however, a

are set between the sofi and the preferred length of stay. At values of P
sudden shift in the model's LOS recommendations occurs. The number of Musculoskeletal cases
where model recornmended lengths of stay are between the soft bound and the preferred length of
stay drops by 9 and the number of Musculoskeletal cases in which model recommendations for
length of stay are equal to preferred LOS increases by 9. At Y = 0.0,there is an increuse in the
number of Thoracic (4) and Oncology (29) cases in which model recommendations result in
length of stay between the soft bounds and the preferred LOS. This effect is due to cost model
Constra.int 3c, which defines hard and sofi bounds on direct case cost:

In cost model Constraints 3c, direct case cost is defined as the model assigned case cost

($d less any savings attributable to reduced length of stay (q,- y,

l,,,). Deviations of this

"revised" case cost fiom the target value of v, are considered direct cost deviations and are
penalized accordingly. An associated bound ensures the total negative variation is less than the
difference between the preferred value of v, and the hard lower bound on v,.

Thus, at values of

T>O, length of stay recornmendations are bounded by constraints on both minimum length of
stay and minimum case cost. As the value of P increases, the model's freedom to set direct case
cost is restricted by reductions already imposed due to changes in length of stay. Ifthe value of Y

165

is large and beds are tightly constrained, rnodel decisions on direct cost may be, as suggested by
the sensitivity analysis, significantly restricted.
From the sensitivity analysis on Y, it is apparent that the proportion of daily nursing costs
saved with each day's reduction in length of stay, affects cost model recommendations at the
clinical team level. Thus, 'P, represents a possible source of error in the model, under specific
operating scenarios. While it would be preferable to idente the tme value of Y,such
information is not presently available. Therefore, a value for V must be assumed. Noting that it
is unreasonable to expect that no savings in nursing costs can be derived fiom a reduction in
length of stay, a value for 'P greater than O is indicated. It also unreasonable to expect that al1
nursing care provided to patients is completely variable. Therefore the tnie value of Y must be
less than 1.0. As a conservative estimate, a value of 0.75 will be assumed for Y in ali runs of the
model. By assurning !

= 0.75,

an implicit assumption is made that 25% of al1 nursing activities

are fixed and therefore do not Vary with patient length of stay.
7.3.3 Data Validity Summary

Models, Schellenberger (1 974) argues, can never completely represent reality. The
purpose of validity, it is argued, is not to ensure exact correspondence between model and reality.
but to identiQ where the model diverges From reality and ascertain if known deviations are
operationally sigdicant. The preceding sensitivity analysis indicates that rnodel output

Vary

with change in uncertain parameters. On the whole, however, the effect of such changes are small
or can be rninimized through appropriate parameter selection.
Changes to the regression coefficient P (direct cost) used to calculate billings, produced a
maximum error of less than 4% under the assumption of both global budget and rate-based
methods of tnding. Changes of less than 2% were observed when p,, the coefficient for direct

166

cost in the outpatient billings regression, was varied. Changes of 110% in the value of P,, the
coefficient for hospital funding in the output billing regression, gave rise to a maximum error of
less than 5%. Changes in the value of a,the y-intercept of the outpatient biliing regression, also
produced a maximum error ofjust less than 5%.
Changes in the value of 'f,the proportion of average daily nursing costs avoided with
each marginal change in patient length of stay do not generally affect model results, unless
inpatient beds are tightly constrained. When beds are tightly constrained, recomrnendations for
direct cost dollars may Vary by as much as 15%, if the value of 'P is aggressively set.
7.4 LogicaI ValiditylModel Verification

Law and Kelton (1992) define model venfication as the process of ascertaining the
veracity of a computerized implementation of a theoretical model. While emor fiee programs can
never be guaranteed. the possibility of errors can be reduced through simple tests and checks.
Adopting Law and Keiton's technique, a simplified version of the model, based on the scenario
described in Section 5.3 (see Appendices A and B) was run and compared to a model generated
and solved manually. The cornputer generated models were observed to be identical to the

manually generated models. See Tables 7.7 and 7.8. From this test, it was assumed that the
model is logically valid.
Table 7.7 Volume Mode1 Error Cornparison - Computer Generated vs. Manual.
Doctor

CMG

185
350
251
253
25 1

253

. Total

125 1
350 '

2.23
22.43
9.09
87.67
17.47
114.21
3.03
0.00
256.13

Volume
Under
Goal P,

1
1
1

2.40
21.60
6.00
91.33
19.05
112.34
3.03
0.00
255.75

Error

Mrnurlly Generrted

Cornputer Generrted
Volume
Under
Goal P,

Volume
Under
GoaIP,

Volume
Under
GoalP,

2.23

9.09

2.23
22.43
9.09

87.67
17.47
114.21
3.03
0.00
256.13

87.67
17.47
114.21
3.03
0.00
256.15

22.43

1
1

Volume

Volume
Under
GoalP,

Under
GoalP,

2-40

21.60
6.00
91.33
19.05
112.34
3.03
0.00
255.75

2.23
22.43
9.09
87.67
17.47
114.21
3.03
0.00
256.13

Goal P,

0.00
0.00
0.00
0.00
0.00
0.00
0.001
0.00 1
0.00 1

Goal P2

0.00
0.00
0.00
0.00
0.00

0.00
0.00
0.00
0.00 ,

Goal P,

0.00
0.00
0.00
0.00
0.00
O .O0
0.00

0.00
0.00

Table 7.8 Cost Mode1 Error Cornnarison Com~uterGenerated vs. Manual.

CMG

Dodor

Cornputer Generated

Direct

Goal P,,

Goal P,,

Gan
iually

Case
Time
Under
Goal pz1

Case
Time
Under
Goal P,,

Direct
Cost
Under
Goal P,

7.5 Predictive Validity

Schellenberger (1974) defines predictive validity as the ability of a model to produce

output that conforms to expected results in both theory and practice. Law and Kelton (1992)
argue for models having high face validity with systern experts. Because the resource allocation
model is normative rather than descriptive, there is no existing system output against which it can
be calibrated or compared. Tests of the model's predictive validity cannot, therefore, be
O bjectively

deterrnined. Instead, a subjective approach to model validation that engenders tests of

credibility with system experts was necessary.

A three stage process for subjectively assessing predictive validity was used. Results from

the model were presented to the Vice President of Nursing, who was asked to judge the model for
reasonableness, accuracy and practical importance. A number of model revisions were necessary
to support the Vice President's recommendations. These recornmendations form the basis of
Chapter 6. When model results were judged to be reasonable, i n t e ~ e w were
s
held with COleaders from each of the six clinical teams comprising the Surgical Council at Mount Sinai

Hospital. Results fiom the model were shown to each team CO-leader,who was asked to
comment on the practicaiity of the model's recomrnendations to his or her tearn. Through this
process hard and soft bounds for case preferences (p,')

were elicited, dong with information

relating to the reasonableness of the model's recommendations.

Upon completion of the clinical team interviews, the model was run with the team COleader's preference array, using data fiom the second half of 1994/95 to develop a hypothetical
resource allocation for the first half of fiscal 1995/96. Results of these runs were compared
against the actual resource consumption data for the first haif'of 1994/95 t o establish predictive
validity .
7.5.1 Establishing Face Validity

Law and Kelton (1992) argue for a practical approach to estimate conformance between
models and real world systems. It is, they note, generally impossible to perform a formal
statistical validation between model output data and the corresponding system output data (even
if it exists). Instead, they argue, validation should focus on producing models that have credibility
with people knowledgeable about the system under study. Models, they note, must have high
face validity, perform well in empirical tests or sensitivity anaiysis and produce results that are
representative of the system under study. When valid, verifiable models are accepted by decision
makers and put into practical use, they can be considered credible.
To test the face validity of the model, clinical team leaders fkom Mount Sinai were
i n t e ~ e w e dand asked to review the model's assumptions and comment on the reasonableness of
output produced under a test scenario (see Appendix F for the background material and scenario
description provided to the team leaders). Through a stmctured interview, tearn leaders were

asked to respond to the following queries about the model, its assumptions and its outputs:

1) Do the assumptions used to build the model seem reasonable to you?

2) Does the model capture ail of the factors you feel are important to resource
allocation?

Specrfic Comments:
1) Are there any specific recornmendations made by the model, with respect to
case mix, that do not make sense for your team? If so, how would you change

them?
2) Are there any specific recommendations made by the model, with respect to
practice change, that do not make sense for your team. If so, how would you
change t hem?
Acceptability:
1) If the specific changes to the recommendations you suggested were made,

could you live with this allocation of resources?

2) Ifyou had to choose between the recommendations for case m k changes and

the recommendations for practice changes, which would you select?

A total of nine clinical team CO-leaders(out of a possible 12) were interviewed. The

cohort consisted of five individuais with nursing backgrounds, one social worker and three
surgeons. One surgeon CO-leaderwas excluded from the analysis for organizational
considerations and two surgeon CO-leadersdeclined to participate, due to a lack of available time.
Results of the interviews were generally positive. Eight of the nine (88.9%) inte~ewees
expressed positive opinions of the model. The lone dissenter, the social worker, held no opinion
on the content of the model, but disagreed with its philosophical implications. "If there isn't
enough money to provide care, it is our responsibility as health professionals" she argued, "to
lobby the province to increase fnding."

7.5.1.1 Reasonableness of Mode1 Assumptions

Eight of the nine inte~eweesdescribed the mode1 assumptions as reasonable. Six of the
Rine respondents (66.6%) accepting the model's assumptions, however, expressed concem with
one or more assumption. Interestingly, al1 three of the surgeon respondents (100%) accepted the
model assumptions listed in Appendix F without condition. While five of the six CO-leaders
(83 -3%) drawn ffom the ranks of nursing or allied health professions expressed minor reservations

with the model assumptions, ody one felt the assumptions were severe enough to invalidate the
model.
Physicians were particularly drawn to the idea of maintainhg a stable income. One
physician noted, "If you're saying that the changes to case mix would be small, and my income
would be within 15% of what it is today, 1 would agree [to using this methodology]." Another
physician noted that while income maintenance is an important concem for him, many surgeons
who have years of sub-specialization training may be unwilling, or unable, to change their case

mix. Nevertheless, the surgeon stated a keen interest in the rnodel's potential to shift case rnix to
more "interesting" cases, while holding income stable.
Nursing and allied health care professionals questioned a number of model assumptions.
Assumptions regarding the economic motivation for changing case mix, in particular, raised
concerns for this group. One respondent wondered if it is reasonable to consider physician
income when selecting case rnix. Such decisions she felt, "should be based on the needs of
patients, their families and the hospital's ability to provide care." Two respondents argued against
small changes to case mix. Patient an-ivals, they suggest, are too random to be controlled at a
minute level. Case mix, they argued, can only be adequately controlled by eliminating staff
physicians or withdrawing admitting privileges. The respondent rejecting the model felt it was

171

unreasonable to use econornics as a buis for decision making in a publicly funded heaith care
system. "Everyone who needs care should be provided it. How can we tum someone away?
Equality of services, or at least equality of access to care must be maintained. How do you think
it would play in the [newspaper] if it was discovered that Mount Sinai was tuming away some

patients because they couldn't afYord to care for them? Ifthere isn't enough money to go around,
Mount Sinai should go into debt, or we should be lobbying the province to provide adequate
funding." Nevertheless, a CO-leaderwith a nursing background felt that it was important for the
hospital "to define what it was al1 about." In tirnes of decreasing resources, she argued, Mount

Sinai "can't be al1 things to ail people. We must decide what it is that we want to do and how we
should do it." The model, she felt, would help to bring issues regarding the importance of certain
clinical programs into sharper focus.
Another respondent who, in addition to her role as clinical tearn CO-leaderis a manager on
a nursing floor. questioned the assurnption of a direct relationship between nursing costs and
patient volume. Nursing floors, she notes, are staffed at specific levels, according to the number
of occupied beds. How could a change in the case mix, she argued, lead to a decrease in nursing
costs, unless beds are closed and staff laid off! This argument, while correct for the local
environment, is flawed when considered fiom the system wide context. If a unit has 5 FTE's
before and d e r a case mix change is implemented, the local cost has indeed not been changed by
case rnix revisions. However, this argument overlooks the fact that fewer nursing units may be
required to deliver care to the revised rnix of patients selected by the model.
7.5.1.2 Additional Model Factors

Respondents generally had few suggestions for additional model factors. Al1 nine
respondents (1 00%) discussed issues of pnority programs and priority cases, but were

172

cornfortable with the assumption that such priorities could be encapsulated within a vector of
stated preferences. One surgeon noted, "last year's volumes are a good place to start, as long as
there is roorn for give and take." In the absence of a stated preference, eight of the nine
respondents (88.9%) accepted the assumption that the previous year's case mix and volume
adequately represented tearn case m k and practice goals for the coming year.
Three respondents (33 -3%) suggested additional factors they thought might irnprove the
model's practical ability. One respondent, a nurse, suggested demand for services should be
included in the model's calculation of case mix. She expressed concem over the idea case rnix
could be increased indefinitely without regard to demand. Furthemore, she suggested, some
notion of demand is important when dealing with life-threatening, or time sensitive, cases such as
trauma or oncology and suggested that these be handled differently fiom purely elective cases.
Two respondents (22.2%), one a nurse, the other a surgeon, suggested the model include
provisions to eliminate low volume cases. Both indicated physicians and institutions have
difficulty maintaining clinical expertise when case volume is below a certain threshold. In
addition, they noted, such cases ofien tend to be expensive. The nurse co-leader suggested 8
cases as the cutoff volume; the surgeon suggested 10. Both acknowledged the literature on
minimum case volumes is conflicting.
One of the surgeon respondents, who has a research interest in economic assessment of
complex surgical procedures, discussed the inclusion of a measure of societal well-being in the
model. An ideal system he noted, "would use epidemiological methods to identify the incidence of
disease within the catchment area of an institution. Hospital resources could then be set up in a
way that we could produce the best possible outcomes for the population." The difficulty of
measunng trade-offs between different illnesses and modalities of treatment with probabilistic

173

outcome, he noted, would made such an extension unwieldy. Nevertheless, the surgeon
respondent noted, "it might be interesthg to look at the problem fiom the perspective of society,
rather than being lirnited by the walls of [any particular hospital]."
7.5.1.3 Specific Recommendations Regardiog Case Mir

AU respondents were able to identify model case rnix recomrnendations they felt were
undesirable. Specific recommendations for revisions to the model's selected case mix were
motivated by either clinical or strategic considerations. In some instances the model makes
recommendations that are not clinically reasonable. For exarnple, the model may recommend, for
economic reasons, a particular seMce increase the number of cases in CMG 80 1 (Wound
Debridement). Clinically, wound debridements are performed because a patient has developed a
post-operative infection. Since wound infections are considered an indication of poor quality of
care, institutions naturally want to eliminate, not increase, the number of debridements. In such
circumstances, model recommendations for increases clearly are not clinically reasonable. More

commody, respondents identified model recommendations not corresponding to their tearn7s

strategic initiatives. Table 7.9 lists the specific recomrnendations relating to case mix identified by
the clinical tearn CO-leaders.

1 1

Table 7.9 Case Bounds Identified.

Tearn

Description

CMG

251

Value

1 GASTROSTOMY & COLOSTOMY PROC

253

MAJOR INTESTINAL & RECTAL PROC CC

254

MAJOR INTESTINAL & RECTAL PROC NOCC

.

64

a86

115

r 154

41

r 55

289

1 INFLAMMATORY BOWEL DISEASE

103

r 106

801

1 WOUND DEBRIDEMENT

i l

- p p p p p

Description
-

Generai &
Speciaity Surgery

366

BACWNECK PROC NO FUSION c 65

19

226

437

OTH DERM PR NO MALlG ULC CEL

s2

438

OTH DERM PR FOR MALlG ULC CEL

s3

1
12 1

r 17

504
508

1 MAJOR URINARY TRACT PROC

1 MINOR UPP URINARY TRACT PROC

51O

1 TRANSUREM PROSTATECTOMY CC

554

1 M E C MALE REP SYS PROC (MNRH)

52

r 70

352

1 HIP REPLACEMENT CC

97

r 130

353

1 HfP REPLACEMENT NOCC

354

1 KNEE REPLACEMENT

101

r 135

357

1 FRACTURED FEMUR PROC NOCC

49

s 39

125

1 MAJOR CHEST PROCEDURES

12

r 27

--

Thoracic

184
Oncology

1 MAJOR RECONSTRUCTIVEVASC PROC

r 88

20

17

254

1 MAJOR INTESTRECTAL PROC CC

1 MAJOR INTESTJRECTAL PROC NOCC

1
32 1

361

( MUSCULOSKELETALBX FOR MALE

58

s46

362

1 MUSCULOSKELETAL BX NO MALE

59

s 47

727

1 ACUTE LEUK W/O MAJ PROC, CC

728

1 LYMPHWHRN LEUK, OTH PROC. NOCC

33

s 26

084

1 MlSC EAR. NOSE & THROAT PROC

61

s 48

253

42

7.5.1.4 Specific Recommendations Regarding Practice Change

None of the practice change recommendations made by the mode1 were identified as
unreasonable by the tearn leaders interviewed. (Practice bounds in the example scenario
presented to team leaders were set at *25% of historical average.)

r 57
z 43

7.5.1.5 Mode1 Acceptability

Eight of the nine (88.9%) CO-leaderrespondents accepted the model and its
recomrnendations. However, in response to the question, "Could you live with the model's
recomrnendations?" one nurse noted that the question may be more appropriately stated, "Can
patients Live with the model?" Since changes in case mix affect patients ability to access care, the
question, she felt, should be asked h m a patient's perspective.
Physicians, interestingly, were excited about organizational implications arising fiom the
rnodel. One surgeon noted that he would be happy with the allocation of resources indicated in
the model if, under a rate based system, budget authority and responsibility were transfemed to his
team. Managing case mix and product cost, he argued, would be a simple matter if tearns had
authority to allocate resources as they pleased and to "buy [hospital] seMces fiorn the lowest cost
provider, even if that means we do our surgery across the street." Another physician respondent
noted that since his team is so efficient, resource allocation based on efficiency and revenue
generation would produce favourable results for it. Under a revenue based system his tearn would
be seen as "favoured sons." The ability to generate revenue, and thus subsidize less efficient
seMces would, he noted, give his tearn enormous bargzining power with the hospital and other

clinical tearns. The third physician sirnply stated that the mode1 was acceptable, so long as there
remained some sense of give and take between the hospital and the tearn and that the need for
maintaining clinical expertise in a wide variety of cases was recognired.
7.5.1.6 Preferred View

Nine of the nine respondents (100.0%) indicated they preferred the view of resource
allocation based on practice change, rather than case rnix. Ail of the respondents indicated that
their first pnonty, when faced with diminishing resources, would be to improve the efficiency of

176
service delivery. Case mix adjustments should be made only when all avenues for efficiency gains
have been exhausted. Nevertheless, respondents noted that case mix changes created by moving
inpatient procedures to outpatient procedures are possible, and is a strategy actively pursued by
Mount Sinai. When pushed about the Limits of efficiency improvements, three of the six nonphysician respondents conceded that case mix adjustments rnay be necessary. One suggested
physicians be selectively elllninated from clinical teams, another suggested teams drop programs
that could be adequately provided by cornrnunity hospitals, the third indicated that specific, high
cost, iow revenue teams should be eliminated.
7.5.1.7 Face Validity Summary

InteMews with clinical team CO-leadersindicate the model, its assumptions and its outputs
enjoy face validity with context experts. Nevertheless, two important issues for resource
allocation models were identified during the interview process. These are:
1)

There is a dichotomy between physicians and non-physicians as to what an

institution's objectives should be and how it can best be managed.

2)

Autonomy and accessibility are important concepts for care providers. There
remains a reluctance among care providers to corne to grips with economic
realities and ethical quandaries inherent in resource allocation decisions that fit
the definition of "tragic choice".

Glouberman and Mintzberg (1994) argue hospitals are not cohesive organizations.
Hospitals exhibit organizational cleavages dong functiond lines and organizational membership.
Nurses and allied health professionals dong with physicians, the people who provide seMces to
patients, have a fundarnentally different set of pnonties than do a hospital's board of directors and
its administrators who are concerned with the ongoing management of the organization.
Furthemore, note Glouberman and Mintzberg, the loyalty of nurses and administrators, who are

employees of the hospital is different fiom that of physicians and the board of directors, who are
independent entrepreneurs not employed by or paid by the hospital (see Figure 7.12).
Figure 7.2 Glouberrnan and Mintzberg Four Faces Model.

Containment
Codition
r

,1

Board of Diredors

Administrators

" Community"

Status
Codit ion

" Control"

N ursng &
AlliedHealth

" Cure"

lnsder
1 Codition

ii

" Care"

Clinical
Codit ion
'1

Nursing and allied health care personnel, according to this model, perceive their role as
care givers. An important part of care, according to the model, is to act as patient advocate in a
systern where the objectives of administration and physicians may not always correspond to what
is best for patient care. The response of nursing personnel to the econornic motivation of the
model is entirely consistent with the Glouberman and Mintzberg model and is likely based on their
roie as patient advocate. There may also be a sense arnong nursing personnel that case mix is not
controllable, simply because it is generally beyond the scope of the hospital to control. (It should

178

be noted, however, that Mount Sinai has in some circumstances taken steps to control case mix

and volume. Of particular note is the hospital's decision to cap the number of newbom delivenes
in the Perinatal Planning Council.)

The reaction of physicians to the model is quite different fiom that of nursing personnel,
but is consistent with components of the "Four Faces" model. Doaors, note Glouberman and
Mintzberg, have a natural desire for managerial authority. Doctors, it is argued, are bright, highly
educated people who (rightly or wrongly) believe they know how the system should best be
managed. There is, according to this mode1 a desire among physicians to imitate the managerial
and entrepreneurial role of their status peers, the board of directors. The reaction of physicians,
who tend to accept the resource allocation model because of its implications for autonomy and
resource control is a natural extension of this afinity for managerial authority.
The dichotomy of opinion between physicians and non-physicians about organizational
objectives has practical implications for the resource allocation model. First of dl, it suggests that
acceptance of models of this nature rnay be more difficult among nursing and ailied health
personnel than among physicians. Secondly, a practical impiementation of model
recornmendations, particularly case mix adjustments, may not be palatable, uniess al1 efforts have
been taken to ensure the delivery of care is as efficient as is possible. As a result, greater
communication and coordination is required between physicians, administrators and nursing
personnel to eliminate inefficient service delivery and identie appropriate czse rnix and product
cost objectives.
A second theme ernerging fiom the team CO-leaderi n t e ~ e w is
s the reluctance of care
providers to recognize the econornic implications and ethical quandaries associated with resource
allocation decisions. Several inte~eweessuggested that "no one is tumed away fiom Mount

179
Sinai" or indicated that "everyone who needs care gets it." It is, they note, unethical to deny care
on the basis of economics. These opinions, while admirable, may not be valid.

The assertion that no-one is tumed away fiom Mount Sinai is essentially incorrect. The
hospital, for instance, does not provide cardiovascular surgeiy; neurosurgery and thoracic
surgery, while performed at Mount Sinai, are relatively small programs, in cornparison to other
teaching hospitals in Toronto. Individuals desing these services are generally referred elsewhere;
if they wish to receive them fiom Mount Sinai they must wait. Mount Sinai, like ail hospitals,

provides only those seMces for which it has specific and sufficient competence. Furthemore, it
dispenses oniy as much of its services as it is physically capable, given restrictions on physician
time, beds, operating room time, nuning staff and money. That nursing staffbelieve no one is
tumed away from Mount Sinai is likely due to the fact that demand for hospitai seMces is
buffered by the physician referral process.
The assertion that everyone who needs care gets it, is similarly incorrect. As Fleck (1994)
notes, demand for health care, in al1 societies, exceeds supply. Rationing, therefore, is a necessary
component of al1 health care systems, whether publicly or pnvately fnded. The real issue
underlying concerns raised by nursing personnel dunng interview sessions, therefore, is whether it
is ethicd to ration health care seMces explicitly. It is evident from the interviews that nursing
personnel - traditional patient advocates - are reluctant to accept rationing. Nevertheless, it
would be naive to assume that rationing does not currently happen and wiil not continue to do so.

The rnismatch between supply and demand make allocation decisions - tragic choices - necessary.
Because there is simply not enough supply to meet demand, rationing necessarily occurs.
The shortage of a social good creates profound difficulties for societies founded on
humanistic principles espousing equality between individuals, since it implies a monetary value for

180
human Me. Rationing, the byproduct of shortage, implies aU individuals are not equal, since some
individuals receive care and others do not. Because society is not able to resolve the confiia
created by tragic decisions, it simply chooses not to make a choice. Societies therefore delude
themselves, arguing that everyone who needs a tragic good should get it, although it is clear that
this is neither true, nor desirable. (Every dollar spent on health care is a dollar that cannot be
spent on education or housing or other social goods.) Choosing not to decide, however, dlows
society to avoid the difficult problem of rnaking tragic choices and thus preserves the ideal that
life is invaluable and that al1 persons are equal.
Health care is a tragic good. Since al1 methods for allocating tragic goods are imperfect
(Calabresi and Bobbit 1974), there is no method for allocating health care resources that is
without flaws. Any amount of fnds dedicated to the provision of health care will ultimately fail
to adequately address al1 possible demand. Similarly, any allocation of available funds will only
partially satis@ demands for care; rationing is an inevitable component of al1 heaith care delivery
systems. Suffenng will, therefore, continue to exist and will be unequally distributed arnong
individuals. The chef difference between rationing as it is currently practiced and rationing under
a defined case mix is that the rationing is explicit rather than implicit. Fleck (1 994) argues that
implicit health care rationing is piecemeal, uncoordinated and therefore arbitrary and unjust. In al1
instances, explicit health care rationing is preferable, since it is systematic and allows for the moral
and ethical review of trade offs.

Rationing, in the context of the resource allocation model, however, does not necessarily
mean that any individual is refsed service. Rationing rnay simply mean that an individual does
not Nnmediately receive seMce from Mount Sinai, or it may mean that an individual must seek
care at another institution. (A more efficient patient referral network may be a necessity of a rate

181

based fnding systern. If institutions have lirnited capacity to produce specific cases, a province
wide referral network may be necessary to ensure individuals have access to care, while
preservhg the nght of hospitals - non-profit corporations - to economic s u ~ v a i . )
Another theme identified during the interview process concemed the ethics of selecting
case m k to satisS, economic objectives. Contrary to the opinion of the respondent who indicated
it is unethical in a publicly funded system to make case mix decisions on the basis of econornics, it
may be unethical in a publicly funded system not to make case mix decisions on the basis of
economics. Because treatment is provided at the public expense, the provision of health case is
not strictly a pnvate arrangement between an individual and a provider; the public, as the service
payer, has the legitimate nght to expect the care afforded to any individuai is provided in the most
efficient, least expensive manner consistent with good medical practice. Therefore, it may
actually be improper for an institution, in a publicly funded system, to provide seMces locally that
couid be delivered more cost effectively elsewhere within the sarne region.
7.5.2 Establishing Predictive Validity

Law and Kelton (1991) argue the most definitive statement of a model's validity is its
ability to produce representative output. Validity, they note, is most comrnonly established by
comparing model output to the performance of real world system. If the results compare
favourably, the model is assumed to be valid and is used to predict the behaviour of the proposed
system. Because reai world systems are non-tirne stationary (the underlying statistical
distributions describing the system change over time), and auto-correlated (the values of
successive outputs from the system are related to one another), classical statistical tests based on
the assumption of independent, identically distributed observations are not directly applicable

182

(Law and Kelton 1991). Law and Kelton therefore argue against hypothesis tests for models of
real-world systems, instead suggesting the use of Turing teas.
Turing, writing in the 1950's on the question of whether machines would ever be able to
think, argued that it is unirnponant whether or not machines actudy think (in the sense of selfactudization), so long as it appears to people who use them that they think (Turing 1950). In the
context of model validation, Turing tests are used to determine whether an individuai
knowledgeable with the process is able to distinguish between cornputer generated and real world
output. When system experts are unable to differentiate between the two, the model is considered
valid in the sense that it simulates the expert's understanding of the real-world process.
In the case of the resource allocation model, a departure from the standard Turing test is
necessary since the model is normative rather than prescriptive. While the rnodel is capable of
reproducing histoncal case mix and pncing decisions, relatively little insight into its predictive
capability is achieved by doing so, since the model is designed to produce "better" decisions than
those made manually. Given an initial set of case rnix preferences prices, and resource
availabilities fiom one period in time, the model should produce case rnix and pnce
recommendations superior to those actually observed in subsequent penods. (Superior in this
sense means the rnodel's recornmendations for case mix and price satise to a greater extent the
economic and clinical goals of the hospital and physicians.) A conventional Turing test is
therefore inappropriate since the objective of the model is to produce a result experts recognize as
significantly better than that of the real-world system. Accordingly, context experts from Mount
Sinai, the Vice-President of Nursing and the Program Director of Nursing, were asked to review
several possible scenarios for resource allocation, including one generated by the model, and to
choose from among them the resource allocation that they, as the decision maker, would select.

183

In this instance, if decision makers select the model's recommendations as their preference, the
model cm be judged to sirnulate their decision making process; to "think" like the decision maker.
7.5.2-1 Resource Ailocation Scenarios

To test the predictive powers of the allocation model, decision makers were presented
three different resource allocation decisions possible in response to a particular set of allocation
targets. Resource allocation targets were detennined by dividing the model database into two
halves (April 1, 1994 - October 4, 1994 and October 5, 1994 - March 3 1, 1995) and projecting
resource consumption on an annual basis. October 4 was selected as the division date rather than
Saturday, September 30, the mid-point of the fiscal year, since the median patient discharge
occurred on Tuesday, October 4. Median discharge date was selected as the division date rather
than calendar median, since hospitals typically reduce seMces during the sumrner months. It was
thus anticipated that a larger number of patients would be discharged during the second haif of the
fiscal year than the first. To ensure comparability between the two halves of 1994195, the median
discharge date of surgical patients was used to divide the data. (In fact the differences between
the two halves of the year were relatively minor; 7,O 1 1 surgical patients were discharged up to
September 30, 1994, while 7,249 were discharged in the period between October 1, 1994 and
March 3 1, 1995.) Data fiom the second half of fiscal 1994195 (October 5 , 1994 - March 3 1,
1995) was then assumed as the database for tests of predictive validity, while data from the first

half of the year (April 1, 1994 - October 4, 1994) was assumed to be actual resource
consumption.
Data f?om the second half of the year was used to predict behaviour in the first hale
because of difficulties encountered in the underlying dataset. The resource allocation model
fiinctions best under conditions where resources are scarce. However, in 1994/95 resource

consumption varied greatly from month-to-month (see Figure 7.13), with average resource
consumption increasing and profit decreasing throughout the year (see Table 7.10). This trend
continues into fiscal 1995/96, making 1995196 aiso unsuitable as a validation benchmark for the
allocation model. It was therefore decided to structure a scenario in which data fiom the second
half of 1994195 represents curent case mix preferences and prices, while data from the first half
of 1994/95 represents actual system performance, since both approximate points on the hospital's
production possibility curve. Resource availability for the validation mns was therefore drawn

from the total consumption of resources in the first hdf of 1994/95. The upper tirnit on revenue
set in the benchmark runs was, accordingly, $39,264,433, while the upper Iirnit on direct cost was
set at $38,19 1,144. A total of 42,888 bed days (approxirnately 1 18 beds) and 944,2 10 minutes of
operating room time (approximately 8 operating rooms) was assumed to be available for the

benchmark mns. Target profit levels were set at $1,073,289 for the validation mns.
Surgical Planning Councll
Resource Consu mption
By Month (Fiscal 1994195)
3,500.000 f

'

-Total

Cost

3.000.000

Pot Funding

-- -

Profit

Table 7.10 Annualized Resource Consumotion Profde - Fixa 1994/95.

1 Potential Revenue
1 Faed Cost

1st Half 94/95

(01 Apr 94 - 04 Oct 94)

$21,351,934.00

1 Direct Cost

$16,839,210.00

1 Bed Days

42,888.00

1 Total Cost

2nd Haif 94/95

(05 Oct 94 - 30 Mar 95)

1 Operating Room Minutes

Three separate case mix and pricing policies were then developed. The first policy, caiied
Model 'A', consisted of the case mix and pricing decision identified by the resource allocation
model using data frorn the second half of 1994/95 as the underlying database and the resource
availabilities listed in Table 7.10 as targets and/or constraints. (Bounds on CMGs identified by
CO-tearnleaders during the face validity i n t e ~ e w listed
s
in Table 7.7 were incorporated into the
revised model database for validation runs.) The second policy, Model 'B' was drawn directly
frorn the assumed "actual" consumption of resources approximated by the dataset denved from
the first half of 19941%. The third policy, Model 'C' was created by assuming an "across-theboard" cut to constrained resources. Thus, to create case mix recommendations for Model 'C',
the net reduction in al1 physical resources was calculated for the test database in relation to the
target levels listed in Table 7.10 (A,
= 0.993 1;

A,

= 38,191, IW38,968,862 =

0.9800; A,

= 42,888143,186

= 944'2 1011,009,524 = 0.9353). The most constrained physical resource

(operating roorn time) was identified, and physician preferences in the test database were scaled
by this factor (0.9353) to create recornmendations for case mix. To create cost model
recomrnendations, product prices were scaled by the appropriate factors (ie, direct cost was

186

multiplied by,,A

length of stay was multiplied by A,,;

case t h e was multiplied by A,d. A

sumrnary of the three options presented to context experts appears in Table 7.1 1.

Table 7.1 1 Mode1 Profdes bv Clinical Team.

Direct Cost

$4,472,232.06

$4319,962.04

$4,494,130.00

$4,182,885.21

Contribution

$4,560,375.58

$4,507,990.32

$4,617,417.67

$4,265,325.98

Physician Billings

$2,123,725.39

$2,123,725.39

$2,136,7 14.06

$1,986,323.48

OR Tirne

229,591-90

212,856.92

216,586.42

214,737.61

Bed Days

12.419.00

12,908.16

13,O 10.00

12,333.25

$1,990,677.48

$2,243,757.1 3

$1,898,176.94

Revenue

Diect Cost

$2,029,48 1-40

$903.338.19

$6840,022.86

$343,962.00

$844,893.54

1
I

OR Time

6 1,329.49

59,915.70

71,289.16

57,361 56

Bed Oays

1,881.41

1,766.79

1,996.00

1,759.69

Revenue

$7,616,922.15

$7,539.740.47

$7,106,887.21

$7,124.118.46

Direct Cat

$2,579,363.19

$2,543.006.79

$2.328.078.00

$2.412.482.17

Contribution

$5,037.558.96

$4,996,733.68

$4,778,809.21

$4,711,636.29

Physician Billings

$1,113,142.49

$1,113,142.49

$1,024,133.42

$1,041,123.81

OR Time

125,589.62

124.288.41

111,229.54

117.464.15

Bed Days

6,987.1 5

6,664.69

6,290.00

6,535.09

Revenue

$1,616,474.91

$1,631,183.49

$1 ,092,254.16

$1,51 1,891 -36

Direct Cost

$769,424.69

$743,417.75

$547,476 .O0

$719,644.04

Conribution

$847,050.22

$887,765.74

$544,778.16

$792,247.32

Physician Billlngs

$379,216.63

$379,21633

$268,008.16

$354,68 1.87

41,947.79

39,406.76

28,270.10

39,233.83

OR r i e

Bed Days

..

1,513.32

. . ..

1,272.00

1,456.06

Revenue

$14,382,146.58

$14,146,002.91

$14,047,36513

$13.451.642.82

Direct Cost

$6,869,950.20

$6,752,908.52

$6,852,218.00

$6,425,474.5 1

Conbibution

$7,512,196.38

$7,393.094.46

$7,195,147.13

$7,026,168.31

Physician Billings

$2,842,091 .O8

$2,842,091 .O8

$2,788,980.69

$2,658,211.96

352,542.65

317,288.38

302,702.18

329,733.60

1
$5,063,033.63 1
$1,43OI?34.43 1

1
S,662.621-72 1
$ls673,346.0 1

$4,732.334.78

OR Time

Bed Days

Revenue
irect Cost

1
1
1

18,051.56
$5,059.689.25
$1,455.490.57

1
1
1

17,717.89

17,476.00

16,883.65

$ t , W .322.47

Conbibution

$3,604.198.68

$3,32,29920

$3,989,275.73

$3,371,012.31

Physician Biliings

$1,488,701 -83

$1,488,701 -83

$1,6f 5.69.41

$1,392,385.01

OR Time

198,52348

190,453.82

214,132.74

185,679.31

Bed Days

2.290.29

2.317.15

2.844.00

2,142.1 1

Revenue

$39,737,321 .93

$5Sl264,2M.0O

$39,264,433.00

337.166.375.60

Total Cost

$38.968.863.29

S38.134158.00

$38,191. 144.00

$37.298.636.60

Profit

$768,458.64

$1,073.289.00

($132.261-00)

Physicien Billings

OR Tme
Bed Days

t111291923.14

$8,534,278.10

W1534,278.09

$8,514,426.00

$7,982,122.88

1,009,525.00

944,120.00

944,120.00

944,120.00

43,186.00

4J88.00

42,888.00

40,392.10

As can be seen from Table 7.1 1, Model 'A', the resource allocation recommended by the

model, is superior to the allocation decisions represented by Model 'B' and Model 'C', thus
s a t i s h g the first condition of the revised Turing test. In al1 instances, physicians are better able
under Model 'A' to achieve their target level of income. Under Model 'B' some physicians
receive more income than they would like, while others are forced to rnake do with less than they
prefer, with overall physician income decreasing under Model 'B'. Model 'C',in al1 cases,
provides physicians with less than their preferred income. From an administrative standpoint

Model 'A' provides improved profitability, at a lower cost than does Mode1 'B'. The
recommendations made by Model 'C',while costing less than either Model 'A' or Model 'B',
yield a negative potentiai profit and is, therefore, inferior to A or B. Clnically, Model 'A' is able
to make case cost recommendations without disturbing any of the key clinical programs of the
hospital. Of the 28 restrictions on case mix identified in the face validity process, Model 'A'
satisfies 28, Mode1 'B' satisfies 10, and Model 'C'satisfies 12.
The recommendations were presented to the two context experts, who were asked to
select, from the three models presented, the one which they, as decision makers, would prefer.
Both context experts selected Model 'A', the resource allocation model. The Vice President of
Nursing selected Model 'A' because it was within the global budget, it provided the best income
for physicians, and it had Iittle impact on clinically important cases. The PDN for Nursing
Informatics selected Model 'A' because its economic and clinical recommendations most
consistently achieved the hospital's targets and those of its medical staff In particular, she noted
Model 'A' optimized institutional profitability, while preserving clinically important prograrns.
7.6 Validation Summary

The resource allocation model presents unique validation challenges since it is a

normative, rather than predictive model. As with al1 normative models, validation by cornparison
with histoncal performance is not meaningfl. Instead, validation efforts focused on ensunng the
technical veracity of the model' s components. Adopting the Schelienberger framework, the
theoretical validity of the model was venfied through ongoing review with both context and
content experts. Data validity was assured through the selection of credible, pnmary data
sources. Data parameters with uncertain values were tested via sensitivity analysis and the impact

of error in these parameters was detedned to be acceptable. The logical validity of the mode1

189
was tested by comparing a sirnplified version of the computer generated model against one

produced by hand. Predictive validity was proven through the use of a modified T u ~ test
g in
which decision makers from the hospital were asked to choose, fiom arnong three resource
allocation scenarios, the one they preferred. The decision makers both selected the
recommendations produced by the model. Because the components of technical validity
described by Schellenberger and the face validity requirements described by Law and Kelton, are
met by the resource allocation model, it was judged to be a valid model of resource allocation.

Having proven the veracity of the allocation model and confirmed its ability to produce
desirable recommendations, a series of tests were carried out to simulate the current decision
making process facing Mount Sinai Hospital. In 1995, the govenunent of the Province of Ontario
announced that funding to hospitals would be reduced by 18% in a three-stage reduction over the
penod of fiscal l995/96 through fiscal 1997/98. The province's plan called for reductions in
funding of 5, 6, and 7 percent over the three years of the prograrn (Eves 1995). Accordingly, a
scenario was structured for the allocation model, in which fnding levels were reduced by 5, 6 and
7 percent, relative to the 1994/95 budget, in three stages. This chapter describes the decision

making scenario posed to the allocation model, lists the model's recommendations and defines the
quality of the solutions. From the recommendations made by the model, several conclusions
about the impact of funding changes on physicians, patients, and the hospital are apparent. The
chapter concludes with a discussion of the strategies needed to cope with large scale budget
reductions.
8.1 Decision Making Scenario - Phased Budget Reductions

To simulate the province's planned reduction in hospital funding, three distinct sets of mns
were made with the allocation model. Reductions in total funding of 5, 6, and 7 percent, resulting

in nrmzrla~ivereductions of 5 , 1 1 and 18 percent, were assumed for the three years of the
provincial restructuring program. In each year of the program, it was assumed Mount Sinai
Hospital would reduce fixed overhead by an amount equivalent to the provincially imposed
reductions. The availability of beds and operating room time was, however, assumed to Vary.
One set of runs was made at each budget level under the assumption that beds days and operating
room time would be reduced in proportion to the provincially imposed cuts. Another set of runs

was made under the assumption that productive, physical resources could be maintained at
1994/95 levels. In each period, however, desired physician billhgs were assumed to remain at
their base (1994195) level. The target for institutional profit was aiso assumed to remain at
19941% levels: $98 1,957 under a rate-based funding model and $0 under a global budget
mechanism. Targets for institutional profit were extrapolated from actual resource consumption
by the Surgical and Perioperative Planning CounciI in 1994195. At that tirne, the Surgical
Planning Council would have generated a potential profit of $98 1,957, if a rate-based funding
model was in place. However, in 1994,Mount Sinai, like dl hospitals in Ontario, was fnded

under a global budget. In that year, the Surgical Planning Council at Mount Sinai Hospital
consumed $38.5 million. If one assumes that the total resources granted to the Surgical Planning
Council in 1994/95 represented a global budget, or envelope of resources that could be expended,
a target income of $0 can be postulated. See Table 8.1 for a list of assumed resource
availabilities. During ail mns it was assumed that the minimum and maximum clinical bounds
defined by CO-teamleaders during the face validity interviews remained in effea. See Table 7.7
for a listing of identified cases.

Table 8.1 Resource AvaiIabiiities - Phased Budeet Reductions

Fiscal 1995186
Cumulative Budget
Reduction
- -

Revenue
Total Cost

Fffed Cost

Fiscal 1998197

Fiscal l99l98

1
1
1

Desired Profit

$39.481.894.$m
$38.499.936.OOb
$21.589.532.00
$981,957.00

(Global Budget)
Bed Days (High)

42,963.00

Bed Days ( L w )

42,963.00

--

OR Minutes (nig;
OR Minutes ( L w )

Desired Physician
Billings

1
1

975,702.00
975,702.00

" When run under a rate-based funding mechuiism G,the revenue cap, is equal to this parameter.
b

When run under a global fiinding rnechanism G,the global budget cap, is equai to this parameter.

To gauge the impact of different operating possibilities, a number of runs were made
within each scenario. Runs were made with varying settings of hospital funding mechanism
(global fnding or rate-based funding), and maximum change in case mix. A total of 2% runs
were defined and executed. In each run, the soft bounds on physician billings were assumed to be
*5% of base billings, while the hard bounds on physician billings were set at *10% of base

billings. Soft bounds on case rnix deviations in each penod were set equal to the planned
cumulative reduction in provincial funding (ie 5, 11 or 18%). Soft bounds on physician practice
parameters (LOS,OR time, direct case cost) were aiso assumed to be equal to provincially
irnposed funding reductions. Al1 remaining mode1 parameters (penalty weights) were held fixed at
the values Iisted in Tables 8.2 and 8.3.

Table 8.2 Peadtv Weiehts Economic Parameters

Deviation Beiow Target

Parameter
Description

Deviation Above Target

Modd
Designation

Parameter
Value

Model
Designation

Parameter
Value

Break Even

WF'

1 00,000.00'

w,*

0.00

Global Budget

0.00

W;

100,000.00'

Physician Billings

w0

1 .O0

WE

1 .O0

'Penalty applicable to a rate-based method of hospital funding only. Penalty weight set to 0.0 for
global funding method.

Table 8.3 Penalty Weights Physician Practice Parameters

Deviation > Soft Bound

Deviation s Soft Bound

Parameter
Description

Model
Designation

Parameter
Value

Model
Designation

Cases

WC

0.1000

WCC

10,000.0000

LOS

WL

0.0001

WLL

10.000.0000

OR Time

Wo

0.0001

WOO

10,000.0000

WM

10,000.0000

Direct Case Cost

WK

0.0001

Parameter
Value

194

8.2.1 Phased Budget Reductions 5% Reduction

Runs of the ailocation mode1 indicate the economic goals of the hospital and its associated
physicians c m be achieved through targeted change to case mix after a 5% budget reduction.
This result holds under both global and rate-based methods of fnding. Furthemore, results also
indicate both sets of economic goals can be achieved, regardless of the availability of productive
resources (ie beds and operating room time). Nevertheless, mode1 results indicate that while the
number of inpatient beds available has tittle impact on mode1 recommendations, operating room
time availability is a significant factor in the satisfaction of economic objectives.

8.2.1.1 Institutional Profit 5% Reduction

Mode1 results show administrative desire for institutional profit cm be easily satisfied
through targeted case rnix change at a 5% budget reduction. Desired institutional profit was
achieved in al1 runs of the mode1 at the 5% budget reduction level. Results were independent of
the method of hospital fnding (rate-based or global budget), the availability of productive
resources (beds and operating room time) and the hard bounds on physician case rnix change. See

Table 8.4.
Table 8.4 Institutional Profit at the 5% Budeet Reduction Level
Maximum
Change to
Physician Case
Beds High'
(100%)

MDc

40%

Global
Budget
Target $0

Rate-Based
Target: $981,957

$981,957.86

8eds Low
(95%)

OR High
(100%)

OR Low
(95%)

$981,957.86

$981,957.86

$981, 9 ~ . 8 6

$0.00
'~esources
mark& as "High are assurneci to remain at 100% of 1994/95 levels. Resources marked as "Low" are
assumeci to bave been reduced f?om thei 1994/95 by an amount egual to the reductions imposeci by the provincial
governrnent. in this case "hw"represents a 5% reduction h m 1994/95 levels.

8.2.1.2 Aggregate Physician Biings - 5% Reduction

Mode1 resuits suggest it is possible to satisQ aggregate physician desire for income at the
5% funding reduction level. Nevertheless, constraints on the maximum aliowable change to case

mix is a crucial determinant of the model's ability to identify a case mix satisfying ail physician
income preferences. The availability of physical resources also affects the model's ability to
satisfy physician income preferences. Of particular note is the fact that while bed availability has
little influence on physician income, constraints on operating room time affect physician incorne
significantly. Interestingly, the funding method in place has no practical infiuence on the model's
ability to satis@ physician billing preferences.
Table 8.5 Total Physician Income at the 5% Budget Reduction Level

Maximum
Change to
Physician
Case Mix

'Runs d

Rate-Based
Target $8,505,610

Global*
Budget
Target
$8,505,610

Beds High
(100%)

Beds Low
(95%)

OR High
(100%)

OR Low
(95%)

Beds Low

10%

$8,379,167

58,379,167

$8,379,167

$8,279,446

$8,278,827

$8,399,548

20%

$8,489.278

$8,489,278

$8,489,278

$8,392,001

$8,388,357

$8,505,173

30%

$8,505,61O

$8,505,610

$8,505,610

$8,487,690

$8,483,749

$8,505,61O

40%

$8,505,610

$8,505,610

$8,505,610

$8,505.610

$8,505,61 O

$8,505,610

OR Low

e undm a global budgeting rnethodology assume bais and OR availability art hi&

From the figures listed in Table 8.5, it can be seen that the most significant factors
affecting the model's ability to recommend case mix changes satisfjmg physician billing
preferences are the nurnber of operating room minutes available and maximum allowable change
to case mix. As physicians are willing to entertain larger changes to case mix targets, the ability
of the rnodel to meet target biiiings increases. It cm also be seen fiom Table 8.5 that the
availability of operating roorn time significantly influences the ability of physicians to generate
preferred levels of income. This observation is consistent with the regression analysis of Section
6.2 in which it was noted that a large proportion of physician billings are fixed. Thus, physicians

196
within the Surgical and Perioperative Planning Council at Mount Sinai gain more income from an
admission than they do from an extended length of stay. Therefore, the greater the amount of
available operating room time, the easier it is for physicians to generate a given level of income.
A second conclusion that can be drawn bom Table 8.5 is the insignificance of hospital

fnding methodology on physician billings. In Table 8.5 it can be seen that physician billings tend
to be slightly higher under global funding than under rate-based funding. Table 8.5, however, is
based on the assumption of proportional reducions in G under both a global budget and ratebased funding mechanism. Because the rate-based revenue limit ($3 9-48l million) upon which
reductions are calculated in Table 8.5 is larger than the global budget ($38.500 million), the
absolute value of a 5% reduction in provincial funding is more onerous, in this instance, under
rate-based funding than a global budget. The total reduction in fnding created by a 5%
provincial cut on a $39.481 million base, under rate-based funding, is % 1.974 million. The total
reduction in funding created by a 5% cut to a base of $38.500 million under global budgeting is
only $1.925 million. Therefore, under the assumption of a global budget funding mechanism, an
extra $50,000 is available. If, however, the reduction in the global assumed in Table 8.5 is
replaced with an absolute reduction equal in total dollars to the reduction imposed by a decrease

in the rate-based fnding cap (ie, the global budget is also reduced by a total of $1.974 million),
identical mode1 results are observed. Compare Table 8.5 to Table 8.6.

Table 8.6 Total Physician Income at the 5% Budget Reduction Level (Global Budget)
Maximum
Change to
Physician
Case Mix

Global Budget
Target $8,505,610

Beds High
(100%)

Beds Low
(95%)

OR High
( 3 OO0h)

OR Low
(95%)

Beds Low
OR Low

This observation has important practical implications for institutions as the funding
payment rnethodology changes fiom global budgeting to rate-based funding. Under the
assumption that the policy lever available to the province is G, the base upon which changes are
calculated will affect the total funding an institution will receive. Since the base arnount upon
which reductions (or increases) are calculated varies between fnding mechanisrns, the mechanism
in place influences the magnitude of reductions. Perversely, institutions currently funded under a
global budget, who would be "profitable" under a rate-based funding methodology, rnay suffer
greater absolute budget reductions when the funding methodology changes.
To illustrate this point, imagine three institutions, each with global budgets of $10 million
and total expenses of $10 million. Under a 5% reduction in the global budget, each institution
would have the following budget profile:

rable 8.7 Example reduction profile under global budget

Institution A

Institution B

Institution C

Original Budget
(G)
Total Cost

Required Savings
(5% of G)
Revised Budget
(G - Required Savings)
Reduction in Total Cost
Necessary to Maintain
Profitability.
(Total Cost (Revised
Budget Profit))

Assume institution A's current case mix would generate $1 1 million, institution B's
current case mix would generate $10 million and institution C's current case mix wodd generate
$9 million, if a rate-based funding modei was in place. Under rate-based funding, a 5% budget

reduction, based on an institution's revenue rather than its cost, would yield the following
reduction profile:
Table 8.8 Example reduction profile under rate-based funding

1 Total Cost

1$

1 profit

Required Savings
(5% of G)

Revised Budget
(G - Required Savings)
Reduction in Total Cost
Necessary to Maintain
Profit
(Total Cost (Revised
Budget Profit))

Institution A

I
1

10.00 $

Institution B

I
1

10.00 $

Institution c

10.00

The results of Tables 8.7 and 8.8 indicate that disparities in the absolute arnount of

legislated reductions exist between the two fnding methodologies, due to the base upon which
the reductions are calculated. Furthermore, Table 8.8 shows the absolute value of reductions
under a rate-based fnding mode1 is typicdy greatest for those institutions that are currently
"profitable". Combined, the results of the two tables suggest a transition strategy for institutions
faced with a move from a global budget to a rate-based funding methodology. In the penod
imrnediately prier to the change in funding methodology, institutions should seek to break even,
rather than generating a profit, under the forthcoming rate-based funding method. This strategy
will ensure that the absolute value of future budget reductions is minimized, under the assumption
that reductions are based on G.
If identical reductions are imposed on the model (ie the total reduction is equal in dollar,
rather than a proportion of dissimilar base amounts), solution results are identicai, regardless of
the method of funding assumed. In Table 8.9 an analysis of the differences between model
recommendations under rate-based funding and global budgeting at the 5% reduction level is
given. The figures in Table 8.9 were determined by summing the absolute difference between
model case mix recomrnendations under the assumption of global budgeting and rate-based
fnding for each combination of physician and CMG (

xyEn15G'0bd

- x,-

1). As can be seen

BPId

fiom the table, the model results are identical; the total difference between the two solutions is
dways less than 1/100th of a case and in many cases is less than l l l 000th of a case. Thus, it may
be concluded, for the purposes of case mix selection, that the choice of fnding methodology is of
no practical significance if the magnitude of reductions under both methods is identical.

Table 8.9 Difference in Mode1 Recommendations Global vs. Rate-Based Funding

Maximum
Change to
Physician
Case MD(
10%

Absolute Oifference in Modd Case Recommendations

Beds High
(100%)

Beds Low
(95%)

0.0001

OR High
(100%)

OR Low
(95%)

0.0001

0.0016

Beds Low
OR Low

0.0003

0.0016

The explanation for the equality of the two models, when identical reductions are
irnposed, is straightforward. Figure 8.1 presents a hypothetical solution space for an institution
faced with a reduction in funding. Assume total funding will be reduced by an amount (R-R') if a
rate-based funding rnethodology is in place or by (G-G') if the institution is funded under a global
budget methodology, and that (R-R')

= (G-G').

Assume, furthemore that the institution and its

medical staff wish to deliver x*cases, where x' equais 6

En%. If the relative proportions of x,

are fixed (ie, the ratio of cases to be delivered cannot be rnodified), the institution must reduce its
volume in response to changes in provincial fnding. This is s h o w in Figure 8.1 as the volume of
cases at R' and G'. As can be seen frorn the figure, if the relative proportion of cases comprising
the hospitalyscase rnix does not change, the impact of budget reductions on case rnix and volume
varies with funding methodology. In Figure 8.1, the value of %. is less than x,., indicating that
funding reductions should be more onerous under global budgeting. in reality, of course, the
relative proportion of cases within a case mix

be varied. Thus, the slopes of the lines in

Figure 8.1 can be adjusted in response to changes in institutional funding;

Figure 8.1 Hypothetical solution space for a hospitd facing budget

reductions, assuming case mix proportions must be maintained.
I

Dollars
R
R

G
G'

Xo.

Xfr

X'

Cases

In Figure 8.2. the identical situation is presented under the assumption, employed in the
allocation model, that physicians are willing to change the relative proportion of cases within their
case mix, so long as billings from the new case mix are equal to the billings from the original case

mix (ie, 1,En>i, does not necessarily e q u a l ~ , ~x', , but z,b,x, = 1 b, x', for ail y). The
geometric impact of this assumption in Figure 8.1 is that the dopes of the cost line and the
revenue line c m change. Figure 8.2 presents a hypothetical solution space for the scenario
described for Figure 8.1 in which case rnix has been adjusted to meet changes in total funding,
under both fnding rnethodologies. Under rate-based fnding, the case mix must be adjusted so
that the sum of the total revenue consumed by al1 physicians is less than or equal to R'. Under a
global budgeting rnethodology the cost line must be adjusted such that the total cost of
production is less than or equal to G'. Each solution allows the hospital's medical staff to achieve

202

their target biiiings through a case mix x". However, an additional goal is present under ratebased fnding, which States that the selected case mix (x"), should generate a target profit equal
to P. There is, therefore, an implicit constraint on total cost in each mode1 nin under rate-based
funding, since R - P = G. Because R - P is equal to G, RI-P must equai G'. Therefore, the two
problems are identical, since the problem under rate-based funding implicitly defines the problem
under a global budget methodology.
It should be noted, however, that the equivalency of hnding methodologies holds only
under a limited set of conditions. For the two problems to be identical, the desired institutional
profit under rate-based funding must be equal to the daerence between G under a global budget
and the rate-based funding revenue cap. In addition, weights on the goal of achieving profitability
(w,' and w,') must be large, relative to the weight attached to achieving physician billings. If

either of these conditions does not hold, or the province uses changes in the funding vector (R) as
a policy lever to reduce fnding to hospitals (as is the case in the United States), the two problems
are not guaranteed to be identical.

Figure 8.2 Hypothetical solution space for hospitai facing budget

reductions assuming case mix can be varied.

Xo.

Xw

X"

Cases

8.2.1.3 Specific Physician Billings 5% Reduction Level

As noted in Section 8.2.1.2,aggregate desire for physician billings can be satisfied at the

5% budget reduction level. The ability of the mode1 to identiQ a solution satisfjmg aggregate

physician desire for billings, however, is dependent on the hard bounds on case mix deviation and
the amount of operating room time available. If maximum case deviation is less than 30% and
operating room time availability is high, or maximum case deviation is less than 40% and
operating room time availability is low, the model is unable to idente a solution satisfling the
income requirements of al1 physicians. This observation, however, does not apply equally to ail
clinical teams. The ability of the model to idente solutions satisfjmg the income requirements of
individual tearns varies. Table 8.10 and Figure 8.4 list physician billings, broken down by clinicai
team, against maximum case bounds. As c m be seen from the table, when operating room

availability is hi& the mode1 is able to identify a case rnix for al1 clinical teams except Oncology
at case bounds of 10%.

Table 8.10 Ph7 ician Income at the 5% Budget Reduction Level

Target
Billings

General &
Speciatty

Musculoskeletal

Thoracic

Oncology

Dent/Eye/ENT

Maximum
Change to
Physician Case
Mix

Beds High
OR High
(100%)

Beds Low
(95%)

OR Low
(95%)

Figure 8.3 The impact of case rnix constraints on physician biiliigs.

Results shown here are based on the assumption of high OR tirne
availability.

Physcian Billings vs Caie Bounds

Riysaai Bllings
3,000,000

,,

2.500.000

,\

Maximum Change to Case Mix

/
i

--GI

Gen M s k

Tho Onc Den

-31t--++--

The results in Figure 8.3, however, are based on the assumption of high operating room
availability (available operating room time is equal to 1994195 levels). Figure 8.4 shows a
breakdown of physician billings, by clinical team, under the assumption of low OR availability.
The results s h o w in Figure 8.4 are similar to those seen in Figure 8.3: The model is able to

identify a mix of cases for most clinical teams meeting billing objectives without requiring large

changes to case mix. However, a rnix for the Oncology team cannot be identified without
changes in case mix of at Ieast 40%. From this result, it rnay be concluded that operating room

availability has a significant impact on the ability of the model to identiQ a solution, based only on
case mix, acceptable to the hospital and its associated medical staff

Figure 8.4 The impact of case mix constraints on physician b i h g s . Results

shown here are based on the assumption of low OR tirne availability.

niyscian Billingsvs Case Bounds

Ryaam 811ings

Maximum Change to Case Mix

I

GI

!- - .
I

Gen M& Tho Onc Den

-+*-

8.2.1.4 Case Mix Recommendations 5% Reduction Level

Mode1 recommendations for changes to case mix affect clinical teams differently. In
addition, mode1 recommendations for case mix change Vary with the amount of operating room
time available and hard bounds on maximum case mix deviation. Tables 8.1 1 and 8.12 sumrnarize
model recommendations for changes to case mix. The figures in Table 8.11 show the sum of the
absolute value of the difference between model recommendations and physician preferences for
case mix ( Enlx, - p,'l)

under the assumption that both inpatient bed and operating room time

availability is high. Table 8.12 shows an identical analysis, under the assumption that operating
room time availability is low. Both analysis assume a 5% reduaion in institutional fnding and a
rate-based funding method for hospital funding.

Table 8.11 Total Case Mix Changes at the 5% Reduction Level - Beds High, ORS High
Maximum
Change to
Physician Case
Mi

Total Recommendationsfor Case Mi Change

GI

General8

Speciaity

MusculoskeletaI

Thoracic

Oncology

DenEyd
ENT

10%

348.65

180.00

98.34

56.15

418.63

355.03

20%

703.16

303.97

1715 8

97.84

709.83

625.06

30%

351.O6

84.23

109.60

79.14

739.01

232.69

40%

321-72

97.39

118.64

70.27

660.45

234.44

Table 8.12 Total Case Mix Changes at the 5% Reduction Level- Beds High, ORS Low
Total Recommendations for Case Mix Change

Maximum
Change ta
Physician Case
MD(

Gl

General &
Specialty

Musculoskeletal

Thoracic

Dent/Eye/
ENT

Oncology

10%

424.08

168.27

110.00

57.00

427.45

319.90

20%

615.73

331-99

176.93

101.O2

705.43

593.25

30%

692.75

492.54

237.54

150.11

1040.78

875.55

108.07

1 3 2 . 2 1 1 , 4 2 ~ 8 ~ 1

40%

386.74

The results listed in Table 8.1 1 are displayed graphically in Figure 8.5.

1-

401.39

Figure 8.5 Total change to case mix versus hard bounds on case mix.
Results presented assume high bed and high OR availability at a 5%
reduction in institutional fiinding.

Totd Case M ix Change vs Case Bounds

Gaies

800

i---

GI

Gen Msl< Tho Onc Den

Jrt

u--+-

The pattern of recommendations Iisted in Table 8.1 1 and illustrated in Figure 8.5 provides

insight into volume model recomrnendations. The volume model is a hierarchical goal program,

with the primary objective of jointly satisfjnng the economic goals of the hospital and its medical
staff. At case mix bounds of 10%, the model is unable to achieve the billings desired by the
Oncology and Dent/Eye/ENT tearns. Therefore, the model's ability to select a case mix similar to
a stated target is severely limited by the primary mode1 goal of satis@ng the econornic desires of
physicians and the hospital. When the hard bounds on case mix are set at 20%, the model is
better able to identiQ a solution meeting the billing preferences of physicians. Nevertheless, the
solution quality of the second and third goals of the volume model, assunng a minimum slate of
resources for each clinical team and rninimizing the change to case mix, degrades when the hard
bounds on case recomrnendations are set at 20%. This occurs because of the hierarchical

structure of the model; higher order goals are always satisfied fira and are never aiiowed to
degrade in response to lower order goals. Thus, because the primary goal of the volume mode1 satis@ng econornic desires - is met, poorer lower order goals are tolerated.
At the 30% and 40% bounds on case mix recommendations, the model is able to identi&

several solutions that jointly satis.

the economic objectives of both physicians and the hospital.

Because several solutions are available at the higher order goals, the model has greater flexibility
to satisQ the lower order goal of minimizing change to case mix. Thus, at the 30% and 40%
bounds on case mix, the model is able to idente solutions in which the total recornmended
change in clinical tearn case mix is less severe than at the 20% level.
Table 8.13 surnmarizes mode1 case mix change recommendations for each clinical tearn at
case mix bounds of 3O%, under the assumption of high operating room availability. The figures
listed in Table 8.13, and displayed graphically in Figure 8.6, show the total number of
recommendations for case mix volume in each of the five possible ranges available to the model:

[Hard Lower Bound, Sofi Lower Bound), [Soft Lower Bound, Preferred Volume) preferred
Volume, Preferred Volume], (Preferred Volume, Soft Upper Volume], (Sot? Upper Volume,
Hard Upper Volume].

Table 8.13 Vol me Model Recommendation Summarv 5% Reduction

# CMGs with
Volume
Between Hard
Lower Bound
and Soft Lower
Bound

# CMGs with
Volume
Between
Preferred and
Soft Lower
Bound

# CMGs wiai
Volume at
Preferred

'

# CMGs with
Volume
Between
Preferred and
Soft Upper
Bound

# CMGs wiai
Volume
Setween Hard
Upper Bound
and Soft Upper
Bound

1 Gen 8 Spec
Thoracic

1 Oncology
Figure 8.6 Mode1 case mix recommendations - 5% reduction.

Case M ix Recomrnendation Didribution

# of Cses Within Rage
Cases

GI

Gen

Mdc

Tho

Onc

Den

Clinical Team

As is evident fiom Figure 8.6, mode1 recommendations generally result in a net increase in

the number of cases performed by clinical teams. The case mix change recommended by the
volume mode1 is a simultaneous increase in "profitable" cases and reduction in "unprofitable"
cases. As shown in Figure 8.6, the significance of mode1 case mix recommendations varies by
clinical team. In most instances, mode1 recommendations at the 5% level result in a large volume
reduction for only a minor number of cases, coupled with recommendations for smder decreases

to a larger number of services. However, the pattern of changes recommended for the GI team
and the Oncology tearn d s e r fiom those of the remainder of the clinical teams comprising the
Surgical and Perioperative Planning Council at Mount Sinai Hospital. The model, as shown in
Figure 8.6,suggests large reductions to a significant number of GI cases. In addition, model
recornmendations for the Oncology tearn suggest both large volume increases and decreases.
8.2.1.5 Case Mix Recommendations - 5 % Reduction

Volume model results can, in aggregate terms, be viewed as strategic recommendations

for product growth or decline. Table 8.14 summarizes mode1 recornrnendations for product
expansion or contraction identified as necessary for Mount Sinai and its medical staff to achieve
their econornic objectives after a 5% budget reduction. Table 8.14 lists a maximum of six
significant recornmendations for each team as measured by percentage and absolute change in

volume, under model mns in which case constraints were set at a maximum of 30% and OR
availability was hi&. Recommendations are sorted by percentage change in volume and filtered
so that al1 CMGs with absolute volume change of less than 5 cases are excluded.

Table 8.14 Xecomm nded Changes to Case Mix

Preferred
Cases

Description

GI PROCEDURES (ENDOSCOPY)
OTHER DEFINED BY MODEL
OTHER DEFINED BY MODEL
OTHER HEPATOBlLtARYAND PANCREAS
OTHER G.I. DIAGNOSES AGE > 18
LESS EXTENSIVE ESOPHAGEAL, STOM
BLADDER & URETHRAL PROCEDURES
OTHER LOWER EXTREMITY PROCEDURE
OTHER DEFINED BY MODEL
OTHER MUSCULOSKELETAL PROCEDURE
MULTIPLE OR BILATERAL JOINT RE

L
E
l
Gen & S ec
MUSCU~Oskeletal

035

tf
Thoracic

FRACTURED FEMUR PROCEDURES WIT

OTHER DEFINED BY MODU,
GI PROCEDURES (ENDOSCOPY)
OTHER VASCUIAR PROCEDURES

1
1

2286.00
135.00
387.00
17.00
49.00
69.00
1088.00
58.00
245.00
35.00
15.00
48.00
168.00
105.00
f 5.00

Assigned
Cases

1
1

2401.O0
176.00
418.06
11.O0
34.00
48.00
1143.00
76.00
258.00
46.00
10.00

Change

39.00
219.00
111.00 ;
10.00 1

115.00
41.O0
31.06 ]
-6.00
-15.00
-21.00
55.00
18.00
13.00
11.O0
-5.00

-9.00
51.O0
6.00
-5.00

CMG

Description

Oncology

DO0
736
063
383
076
000
125

Dent 1 Eye /
ENI

DO5

OTHER DEFINED BY MOOEL

CHEMOTHERAPY
TRANSFUSIONS
JOINT REPLACEMENT FOR WIGNANT
MAJOR HEA AND NECK PROCEDURES
OTHER DEFINED BY MODEL
MAJOR CHEST PROCEDURES
LENS PROCEDURES
DENTAL SURGERY
OTHER DEFINED BY MODEL
OTHER OPHTHALMlC PROCEDURES (M
DENTAL U(TRACTION/RESTORAnON
M R O I D PROCEDURES

060
000
057
089
479

Prefened
Cases
322.00
198.00
169.00
31.O0
34.00
406.00
102.00
521.00
423.00
330.00
192.00
56.00
1f 9.00

Assigned
Cases

Change

419.00
258.00
217.1 1
21.O0
23.00
385.00
71.00
548.00
445.00
347.00
182.00
43.88
83.00

97.00
60.00
48.1 1
-10.00
-1 1.00
-21-00
-31.00
27.00
22.00
17.00
-10.00
-12.12
-36.00

As can be seen fiom the table, volume model recommendations generally suggest an

increase in outpatient procedures and a corresponding decrease in the number of inpatient

procedures.

8.2.1.6 Practice Parameter Recommendations 5% Reduction

Al1 mns of the model made at the 5% budget reduction levei assume soft bounds on
physician practice (length of stay, operating roorn time, direct cost) equal to 5%. Hard bounds on
physician practice were assumed to be 18%. At these bounds, al1 volume model solutions could
be translated into cornmensurate cost model recommendations under the assumption of
cooperation between physicians. A surnmary of CO^ model recommendations for practice
changes appears in Tables 8.15 to 8.17. Cost model recommendations assume soft bounds of 5%
and hard bounds of 18% on case mix. See Appendix G for a complete list of volume model and
cost model recomrnendations after a 5% budget reduction.

Table 8.15 Cost Model LOS Distribution - 5% Reduction

Total Bed Days
Assigned by Cost
Model

Thoracic

1
1
1

Oncology

--

General & Speciatty

Musculoskeletal

# of Cases LOS Set

Between Hard and
Soft Lower Bound

# of Cases LOS Set

Between Prefened
Value and Soft
Lower Bound

# of Cases LOS Set

at Prefened Value

.- -

1,823.0
1,194.0
517.0
3,535.0

Table 8.16 Cost Model OR Time Distribution 5% Reduction

Total OR Minutes
Assigned by Cost
Model

1 General & Specialty

MusculoskeletaI
Thoracic

1 Total

67.142.0

# of Cases OR Tirne
Set Between Hard
and Soft Lower
Bound

# of Cases OR Tirne
Set Between
Preferred Value and
Soft Lower Bound

120,938.7

32,799.9

975,701.8

# of Cases OR Tirne
Set at Preferred
Value

Table 8.17 Cost Model Direct Cost I istribution 5% Reduction

Assigned by Cost
Model

1 General& Speciaity 1
1 Muculoskeletal 1

$868,564

$2,443,628

1 Thoracic

$544,856

1 Total

$16,015,786

# of Cases Direct
Cost Set BeWeen
Hard and Soft Lower
Bound

# of Cases Direct
Cost Set Between
Preferred Value and
Soft Lower Bound

# of Cases Direct

Cost Set at
Preferred Value

8.2.2 The 11% Budget Reduction

To simulate the second year of the provincial restructuring program, a set of mns was
canied out on the model under the assumption the hospital's base budget would be reduced by a
total of 1 1%. Runs of the model were made only under the assumption of rate-based funding
since, as shown in Section 8.2.1.2, model recommendations are insensitive to the method of
hospital funding in place, so long as the absolute value of the reduction is equal. Results of the
model at the 11% reduction level echo model results made at the 5% reduction level. Specifically,
runs made of the model at the 1 1% reduction level indicate it is possible to satis@the economic
aspirations of both hospitals and physicians. However, some physician flexibility with regard to
case mix is required for the model to renirn a solution via case mix change alone. Results of the
model at the 1 1% reduction level highiight, furthemore, that the availability of operating room
time is a critical determinant of economic and clinicai aspirations.
8.2.2.1 Institutionai Profit

- 11% Reduction

Desired institutional profit was achieved in al1 runs of the mode1 made under the
assumption of an 1 1% budget reduction. Institutional profit achieved was independent of the
availability of productive resources available (beds and operating room tirne) and hard bounds on
physician case mix change.

Table 8.18 Institutional Profit at the 11% Budget Reduction Level

Maximum
Change to
Physician Case Ma

Rate-Based Funding
Target $981.95'
l

Beds High
OR High
(100%)

Beds Low
(89%)

OR Low
(89Oh)

8.2.2.2 Aggregate Physician Billings 11% Reduction

Mode1 results show it is possible to satisQ aggregate physician desire for income at the
1 1% funding reduction level. However, constraints on the maximum allowable change tu case

mix rnust be relaxed significantly (t60%) before a solution meeting the aggregate income desire
of al1 physicians can be identified. The rnodel's ability to identiQ a solution acceptable to al1
physicians is, furthemore, confounded by the availability of physical resources. When operating
room availability is low, physician desires for billings c m o t be satisfied uniess hard bounds on
case mix are set to *80% of target. The availability of inpatient beds has no impact on the
model's ability to satis@ physician billing desires. See Table 8.19.

Table 8.19 Aggregate Physician Billing 11% Budget Reduction Level

Maximum
Change to
Physician Case Mix

Total Physician Billings

Target $8,505,610

Beds Low
(89%)

Beds High
OR High
(100%)

OR Low
(8g0r6)

8.2.2.3 Specific Physician Billings I l % Reduction

As was noted at the 5% budget reduction level, the ability of the mode1 to identifL a

solution satisfjing individual physician desires for billings varies by team and is dependent upon
hard bounds on case mix deviation and the amount of operating room time available. This
observation holds tnie at the I l % budget reduction level, but is more pronounced. See Table

rable 8.20 Phy ician Income r

Clinical
Team

Target
Billings

Maximum
Change to
Physician Case
Mix

OR High
(100%)

(89%)

(89%)

Target
Billings

Thoracic

Maximum
Change to
Physician Case
Mix

Be& High
OR High

(100%)

Beds Low
(89%)

OR Law

Target
Billings

Maximum
Change to
Physiaan Case
Mix

Beds High
OR High

Beds Low
(89%)

OR Low
(8gor6)

(1 00%)

The results listed in Table 8.20 are displayed graphically in Figures 8.7and 8.8.
Figure 8.7 Physician billings versus case bounds - 11% budget
reduction. Operating room tirne availability assumed to be high.
I

PhyScian Billingsvs Case Bounds

Fhyaaa Bllirigs

M aimum Change to Case Mix

Figure 8.8 Physician billings versus case bounds - 1 1% budget

reduction. Operating mom time availability assumed to be low.
i

Physaan illingsv s Case B o u m

Riy3aa7 Bllings

20%

30%

40%

50%

60%

70%

80%

Maimum Change to Case Mix

8.2.2.4 Case Mix Recommendations - 11% Budget Reduction

Table 8.2 1 shows the sum of the absolute value of the difference between mode1
recoinmendations and physician preferences for case mix at the 1 1% budget reduction level, under
the assumption that both inpatient bed and operating room time availability is high. Table 8.22
shows an identical analysis, under the assumption that operating room time availability is low.
Resuits of this analysis suggest that significant case mix changes are required to jointly achieve the
economic goals of both physicians and the hospital after an 1 1% budget reduction. Case mix
changes are particularly severe for the Oncology tearn.

Table 8.21 Total Case Mk Changes at the 11% Reduction Level - OR Avaiiabilitv Hieh

Maximum
Change to
Physiaan Case
MD<

Totai Recommendationsfor Case MD<Change

GI

General8
Speciaity

Musculoskdetal

Thoracic

Oncology

Dent/Eye/
NT

20%

724.0

316.6

188.0

100.0

750.8

675.0

30%

1,001-2

478.8

235.6

131.3

1,046.6

948.0

40%

1,333.3

588.7

319.6

194.6

1,371-4

1,225.6

50%

1,759.4

731.1

321.2

237.2

1,626.4

1,533.3

60%

1,580.8

285.4

338.9

285.6

1,938.0

1,338.8

Table 8.22 Total Case Mix Changes at the 11% Reduction Level - OR Availability Low

Maximum
Change to
,
Physician Case

MD<

Total Recommendations for Case Mix Change

GI

General &
SpeciaRy

Musculoskeletal

Thoracic

Oncology

DentlEyel
ENT

20%

708.9

224.1

180.0

100.4

739.1

675.0

30%

854.8

492.5

253.0

151 .O

1,069.5

875.6

40%

734.1

659.5

324.0

196.0

1.369.3

1,163.8

50%

880.9

817.5

317.2

237.2

1.650.4

1,442.3

60%

1,104.1

979.6

370.9

285.6

2,016.3

1,729.9

70%

1,107.9

1,146.5

406.1

333.7

2,334.5

2,029.3

80%

830.8

1,006.8

449.7

359.7

2,714.1

2,311.1

Table 8.21 and 8.22 results are displayed graphically in Figures 8.9and 8.10.

Figure 8.9 Total case mix change versus maximum case bounds, 1 1%
budget reduction. Operating room availability high.
I

1-

Totd Case Mix Change vs Case Bounds

20%

50%

40%

30%

60%

M aimm Change to C;rse Mix

!!
1

--Gl

Gen Msk Tho Onc Den

**--+-

Figure 8.10 Total case mix change versus maximum case bounds, 1 1%
budget reduction. Operating room availability low.
4

Tot ai Case M ix Change vs Case Bounds

Cases

30%

40%

50%

60%

70%

80%

Mpmum Change to Cse Mix

If_-

GI

Gen M&

Tho Onc Den

+u+

As can be seen in Figures 8.9 and 8.10, mode1 recomrnendations tend to result in case rnix

changes for clinicai teams that increase proportionately with the hard bound on maximum case

deviation. The total case mVr change recommended by the model increases until al1 physicians are
able to achieve their desired income, after which model recommendations for case change tend to
decrease. This pattern is consistent with the results seen at the 5% reduction level. Nevertheless,

the distribution of case mix changes recommended by the model is more severe at the 11%
reduction level than at the 5% level; model recomrnendations result in a significantly greater
number of cases whose volume is set within the range P a r d Lower, Soft Lower) at the 1 1%
level. In addition, fewer cases were set at their preferred volume. See Table 8.23 and Figure
8.11.

Table 8.23 Volume Mode1 Recommendation Summarv - L 1% Reduction

# CMGs with
Volume
Between Hard
Lower Bound
and Soft Lower
Bound

Gen & Spec

Thoracic

# CMGs with
Volume
Between
Preferred and
Soit Lower
Bound

# CMGs with
Volume at
Preferred

# CMGs with

Volume
Between
Preferred and
Soft Upper
Bound

# CMGs with
Volume
Between Hard
Upper Bound
and Soft Upper
Bound

Figure 8.1 1 Distribution of mode1 case mix recommendations - 1 1 %

budget reduction. Operating room time avaiiability assumed hi&;
maximum case deviation *60% of ~referred.

GI

M*

Gen

Tho

On:

Den

Clinical Team

8.2.2.5 Case Mix Recommendations I l % Reduction

Mode1 recommendations for case rnix change provide insight into an econornically rational
case mix strategy for physicians and hospitals. In Table 8.24 mode1 recomrnendations resulting in
large changes to case mix are given for each clinical tearn.

Table 8.24 Recomrnended Changes to Case Mix 11% Reduction

CMG

Description

Gen & Spec

MUSCU~Oskeletal

Thoracic

Oncology

Dent / Eye /
ENT

GI PROCEDURES (ENDOSCOPY)
OTHER DEFINED BY MODEL
BlOPSY
OTHER G.I. DIAGNOSES AGE > 18
LESS EXTENSIVE ESOPHAGEAL, STOM
OTHER DEFINED BY MOOEL
BLADDER & URETHRAL PROCEDURES
OTHER DEFINED BY MODEL
POST-OPERATNEAND POST-TRAUMA
OTHER OR PROCEDURES OF THE B
BIOPSY
OTHER DEFINED BY MODEL

, OTHER OEFINED BY MODEL

OTHER LOWER EXTREMITY PROCEDURE

OTHER MUSCULOSKELETAL PROCEDURE
MULTIPLE OR BllATERAL JOINT RE
OTHER PROCEDURES FOR DlAGNOSlS
OTHER DEFINED BY MODEL
OTHER DEFINED BY MODEL
GIPROCEDURES(ENDOSC0PY)
OTHER DEFINED BY MODEL
OTHER OR PROCEDURES OF THE B
OTHER VASCULAR PROCEDURES
VElN LIGATION AND STRIPPING
GI PROCEDURES (ENDOSCOPY)
OTHER DEFINED BY MODEL
CHEMOTHERAPY
MAJOR HEAD AND NECK PROCEDURES
MAJOR CHEST PROCEDURES
OTHER DEFINED BY MODEL
LENS PROCEDURES
OTHER DEFINED BY MODEL
ORBITAL AND OTHER EYE PROCEDURE
THYROID PROCEDURES
NASAL PROCEDURES (MNRH)
OTHER OPHTHALMIC PROCEDURES (M

Preferred
Cases

86.00
135.00
115.00
- 49.00
69.00
387.00
1088.00
223.00
11.O0
8.00
40.00
265.00

1
1

1
1

- --

1
1

1
1
1

- -

245.00
58.00
35.00
15.00
17.00
174.00
168.00
105.00
48.00
9.00
15.00
94.00
428.00
322.00
198.00
34.00
102.00
406.00
521.O0
330.00
223.00
119.00
144.00
192.00

1
1
1
1

1
1
1

Assigned
Cases

3418.53
216.00
184.00
24.00

. .. . .

1
1

1
1

Change

,.

n.00

1
1

162.00
834.00
528.00
357.00
47.00
57.00
76.00

1132.53
81.00 ]
69.00 1
-25.00

344.00 1
1251.36 1
248.00 1
'i8.00
13.00
45.00
206.96
392.00 (
93.00
56.00
6.00
6.00
i54.00
269.0
168.00
3.00
6.00
37.00
685.00
516.00
317.00
13.00

1
1

-43.00
163.36
25.00
7.00
5.00
5.00
-58.04
147.00

35.00
21.O0
-9.00
-1 1.O0
-20.00
101.O0
63.00
29.00
-6.00
-9.00
-57.00
257.00
194.00
119.00
-2100

1
1

-244.001
313.00
198.00
134.00
-72.00
-87.00
-116.00

A general trend emerges fiom the model recommendations listed in Table 8.2 1 that

supports observations made at the 5% level. In al1 instances, model recommendations suggest
large increases in day surgery or outpatient procedures (day surgery procedures have a CMG

number prefked by the letter "D"),coupled with large decreases in inpatient procedures. These
results correspond to current practices in health care which emphasize outpatient procedures over

225

inpatient procedures. The mode1 aiso recommends that al1 teams, except DentaVEyeENT, reduce
the reliance on inpatient cases with historicaliy low volumes ("000 Other Defined by Modei").
These cases, designated as Type "C" in the distribution by d u e analysis, tend to be inefficient
because teams do not produce a large enough volume to develop s t r e d n e d treatment processes
or take advantage of economies of scde.
8-2.2.6 Practice Parameter Recommendations

- I l % Reduction

Runs of the model indicate that cornmensurate volume and cost model solutions can be
identified d e r an 1 1% budget reduction. Model results, however, also indicate constraints on

maximum deviation to practice parameters must be relaxed when operating room time is tightly
constrained.

A surnmaq of cost mode1 recornmendations for practice changes after an I 1% budget

reduction appears in Tables 8.25 to 8.27. Soft bounds on practice parameters equal to 1 I % and
hard bounds on practice parameters equal to 30% are also assumed in the results appearing in

Tables 8.25 to 8.27. See Appendix H for a complete listing of volume and cost model
recommendations at the 1 1% budget reduction level.
Table 8.25 Cost Model LOS Distribution - 11% Reduction
Team

Total Bed Days

Assigned by Cost
Model

GI

1 Thoracic

# of Cases LOS Set

Between Hard and

Soft Lower Bound

10,865.7

1.254.2

DentaEyeiENT

1 Total

# of Cases LOS Set

at Preferred Value

19

24

0
-

Oncology

# of Cases LOS Set

Between Preferred
Value and Soft
Lower Bound

7
-

--

--

11

15,700.4

51

46

2,296.1

17

18

38,236.9

110

141

Table 8.26 Cost Model OR Time Dis ribution 11% R duction

# of Cases OR Time
Set Between Hard

Assigned by Cost
Model

1 Total

# of Cases OR Time
Set at Preferred
Value

# of Cases OR Time
Set Between
Preferred Value and
Soft Lower Bound

and Soft Lower

Bound

975.701.8

Table 8.27 Cost Model Direct Cost Distribution 11% Reduction

Team

Total Direct Cost

Assigned by Cost
Model

1 General & Speciaity 1

Thoracic
Oncology

$868,564

1-

# of Cases Direct
Cost Set at
Preferred Value

# of Cases Direct
Cost Set Between
Preferred Value and
Soft Lower Bound

# of Cases Direct
Cost Set Between
Hard and Soft Lower
Bound

26

$544,856

18

$6,335,392

96

169

81

Total

$16,015,786

8.2.3 The 18% Budget Reduction

To simulate the third year of provincial restructuring, runs of the model were carred out
under the assumption the hospital's base budget would be reduced by a total of 18%. As was
assumed for the 1 1% reduction, runs were made only under the assumption of rate-based fnding.
Runs made at the 18% reduction level indicate it is not possible to jointly satisQ the economic
aspirations of both hospitals and physicians through case n?ix changes alone. The model indicates.
furthemore, that physician income and institutional profit are very sensitive to the availability of
operating room tirne. Mode1 results highlight the need for sigruficant changes to the case mix of
the Oncology team, if the joint economic desires of doctors and the hospital are to be satisfied.
8-2-3.1 Institutional Profit

- 18% Reduction

Desired institutional profit could not be achieved in al1 runs of the mode1 made under the
assumption of an 18% budget reduction. The ability of the model to idente a solution satisfying
institutionai requirements for profitability is ptima-ily infiuenced by the hard bound on case rnix
and the availability of operating room time. Hard bounds rnust be at least *60% of target values
before the model is able to identiQ a solution achieving desired institutional profitability.

Table 8.28 Institution: I Profit at the 18% Budget Reduction Level

Maximum
Change to
Physician Case Mix

Institutional Profit
Target $981,957'
Beds High
OR High
(1OO0h)

Beds Low
(82Oh)

OR Low
(82%)

8.2.3.2 Aggregate Physician Billings - 18% Reduction

Mode1 results suggest it is not possible to satis.

aggregate physician desire for income,

after an 18% funding reduction, through case mix adjustment alone. Even when hard bounds on
case constraints are set to + 100% of target value, a solution acceptable to dl physicians could not
be identified. Furthemore, the model's ability to identiG an acceptable solution is confounded by

the availability of operating room time. Like runs at the 5% and 11% budget reduction levels,
inpatient bed availability has little influence on physician billings at the 18% reduction level.
Operating room availability, however, was observed to have a significant impact on physician
billings. See Table 8.29.

Table 8.29 Aggregate hysician Biiiing 18% Budget Reduction Level

Maximum
Change to
Physician Case Mix

Total Physician Billings

Target $8,505,610
Beds High
OR High

Beds Low
(82Or6)

OR Low
(82%)

(1 00%)

--

--

--

8.2.3.3 Specilic Physician Biilings 18% Reduction

The mode1 was unable to identiQ a solution satisfj4ng the billing desires of al1 physicians at
the 18% reduction level. Nevertheless, solutions acceptable to al1 teams, except Oncology, can be
identified if constraints on case mix are sufficiently relaxed and operating room avaiiability is high.
At case bounds equal to &30% of target, the billing desires of two clinical teams (General &

Specialty Surgery and Thoracic Surgery) cm be achieved. At 70%, five teams (GI, General&
Specialty Surgery, Thoracic, Dental/Eye/ENT) can be satisfied.

Table 8.30 Physician

Clinical

Target
Biilings

--

--

at the
Maximum
Change to
Physician Case
Mix

Reduction Level
Beds High
OR High
(100%)

Beds Law
(82%)

OR Low
(82%)

Target
Billings

Maximum
Change to
Physician Case
MD(

Beds High
OR High

Beds Low
(82Oh)

OR Low
(82Oh)

(100%)

The results listed in Table 8.30 are displayed graphically in Figures 8.12 and 8.13.

Figure 8.12 Physician bdings versus case bounds 18% budget

reduction. Operating room time availabiity assumed to be hi&
I

Physaan Billings vs Case Bounds

Fhyaaan Blllings

Maximum Change to Case Mix

--GI

en

an
++-

M*

T ~ O Onc

Figure S. 13 Physician billings versus case bounds - 18% budget

reduction. Operatinrr room time availabilitv assumed to be low.
-

Physaan Billin- vs Case Bounds

Fhyga'ai Rllim

30%

40%

50%

60%

70%

80%

90%

Madmum Change to Case Mix

1
I

GI
--e----

Ge" M*

Tho Onc De"

-iIC-++

100%

8.2.3.4 Case Mis Recomrnendations 18% Budget Reduction

Table 8.3 1 shows the sum of the absolute value of the dserence between mode1
recommendations and physician preferences for case m k at the 18% budget reduction level, under
the assumption operating room tirne availability is high. Table 8.32 shows an identical analysis,
under the assumption that operating room time availability is low. Results of this analysis confirm
the results seen under the 1 1% budget reduction; significant case mix changes are necessary,

particularly in the case of the Oncology team, to jointly achieve the economic goals of both
physicians and the hospitai. Changes to case mDc are confounded by the availability of operating
room time.
Table 8.31 Total Case Mix Changes at the L8% Reduction Level - OR Availability Eigh

Maximum
Change to
Physician Case
Mix

1
GI

Generaf &
Speciaity

Musculoskeletal

Thoracic

Total Recommendationsfor Case MD<Change

Oncology

DenUEyel
ENT

30%

1,067.0

458.2

255.3

150.2

1,056.7

982.0

40%

1,402.0

584.3

318.8

193.4

1,348.6

1,289.0

50%

1,736.0

729.3

402.7

239.0

1,603.7

1,587.0

60%

2.095.0

894.9

459.7

238.0

1.944.8

1,899.3

70%

2,353.5

1,037.5

519.6

333.7

2,271.8

2,135.4

80%

2,770.0

1,187.2

581.4

381,2

2,609.5

2,451.9

90%

3,155.2

1,337.3

541.6

429.2

2,884.8

2,752.7

100%

3,471.3

1,482.1

593.2

464.3

3,125.4

3,037.6

Table 8.32 To al Case Mir Changes at the 18 Reduction Level OR Avaabiiity Low

Total Recommendations for Case MD<Channe

d$i:kY&e

GI

Mix

General &
Speciaity

1,839.3

Muscul*

Thoracic

Oncology

DenEyd
NT

skdetal

417.9

195.5

322.4

i,237.9

Table 8.28 and 8.29 results are displayed graphically in Figures 8.14 and 8.1 5.
Figure 8.14 Total case mUc change versus maximum case bounds, 18%
budget reduction. Operating room availabiiity assumed to be high.
4

Totai Case M ix Change vs Case Bounds

30%

50%

40%

60%

70%

80%

90%

Maximum Chaige to Case M ix

1

GI
1

Gen M&

Tho Onc Den

+*+

100%

1,167.1

Figure 8.15 Total case rnix change versus maximum case bounds, 18%
budget reduction. Operating room availability assumed to be low.

Total Case Mix Change vs Case Bounds

CasPs

2,500

40%

50%

60%

70%

80%

90%

100%

Maimum Change to Case Mix

--GI

Gen Msk Tho Onc Den

**--i-

As noted at the 5 and 1 1% reduction level, model recommendations for case rnix change

tend to increase proportionately with the hard bound on maximum case deviation until ail
physicians are abIe to achieve their desired income, &er which model recomrnendations for case
change decrease. When operating room availability is low, however, model recomrnendations for
GI team case mix change fluctuate with case mix bounds. SpecificaUy, model recommendations
regrding the number of CMG O00 (Other Inpatient - Defined by Model) and CMG DO89
(Endoscopy) were observed to Vary greatly with bounds on case mix. See Figure 8.16.

Figure 8.16 Recommended change in volume for CMG 000 (Other

Inpatient) and D28 (Endoscopy). 18% budget reduction, OR
avaiiabili~assumed low.

Recommended Volume Change

~dosnpy& Other
Chaigeinm
200

(200)
(400)

(1.000)

'

40%

50%

60%

70%

80%

90%

100%

Maximum Chaige to Case M ix

In addition to the magnitude of volume change, the distribution of case mix

recomrnendations at the 18% budget reduction level is more severe than at the 5% or 1 1% level.
Table 8.33 and Figure 8.17 describe the distribution of case mix recomrnendations made by the

mode1 under the assumption of high OR availability and hard bounds on case mix equal to hl 00%

of preferred volume. As can be seen, recornmended changes result in a large number of cases
being set in undesirable ranges - m d Lower, Soft Lower) and (Sofi Upper, Hard Upper].
Interestingly, the distribution of volume mix changes at 18% is sirnilar to the distribution of
volume rnix changes at 1 1%.

Table 8.33 Volume Modei Rec ~mmendationDistribution 18% Reduction

Volume
Between Hard
Lower Bound
and Soft Lower
Bound

1 Gen 8 Spec

# CMGs with
Volume
Between
Preferred and
Soft Lower
Bound

# CMGs with
Volume at
Prefened

# CMGs with
Volume
Between
Preferred and
Soft Upper
Bound

Figure 8.17 Distribution of case rnix recornmendations - 18% budget

reduction. Operating room time avaiiability assumed high; maximum
case deviation 100% of preferred.

--

Case M ix Recommendation Ditribution

# of CasGs Within Rmge
Casss

GI

M*
Tho
Clinicai Team

Onc

# CMGs with
Volume
Between Hard
Upper Bound
and Soft Uppet
Bound

8.2.3.5 Case M i . Recommendations 18% Reduction

In Table 8.34 mode1 recommendations resulting in large changes to case mix are given for
each clinical team. As can be seen fiom the figures in the table, the trend towards day surgery
procedures, identified at the 5 and 11% reduction levels, continues at the 18% level.

Team

CMG

028
DO0

D61

1
Gen & Spec

Musculoskeletal

256
267
000
D35
703
821

1 OTHER DEFINED BY MODEL

, DOO

OTHER DEFINED BY-MO=

OTHER LOWER EXTREMITY PROCEDURE
- OTHER MUSCULOSKELETAL PROCEDURE
[ OTHER PROCEDURES FOR DiAGNOSlS
1 W O R LOWER EXTREMI~YPROCEDURE
1 OTHER DEFINED BY MODEL
OTHER DEFINED BY MODEL
GI PROCEDURES (ENDOSCOPY)
OTHER DEFfNED 0Y MOOEL
OTHER OR PROCEDURES OF THE 8
OTHER V A S C U M PROCEDURES
VElN LlGATlONAND STRIPPING
GI PROCEDURES (ENDOSCOPY)
OTHER DEFINED BY MODEL
CHEMOTHERAPY
OTHER MUSCULOSKELETAL PROCEDURE
1 MAJOR CHEST PROCEDURES
1 OTHER DEFINEDBY MODEL
LENS PROCEDURES
DENTAL SURGERY
OTHE3 DEFINED BY MODEL
OTHER OEFINED BY MODEL
1 OWER OPHTiiALMfC PROCEDURES (M
LENS INSERTIONWITHOUT CC (MNRH

380
379
803
371
1 O00
DO0
028
000
703
187
192
028
DO0
736
379
125

Oncology

GI PROCEDURES (ENDOSCOPY)
OTHER DEFINED BY MODEL
BIOPSY
LESS EXTENSIVE ESOPHAGEAL. STOM
ANUS AND STOMAL PROCEDURES, AG
OTHER DEFINED BY MODEL
BLADDER & URETHRAi PROCEDURES
OTHER OR PROCEDURES OF THE B
COMPLICATIONS OF TREATMENT AGE

1 O00

Thoraac

Description

1 ooo

DO5
D60
,000
DO0
O57
056

1
1

1
1

Preferred
Cases

2286.00
135.00
115.00
69.00
66-00
387.00
1088.00
8.00
6.00

265.00
245.00
58.00
35.00
17.00
60.00
174.00
168.00
105.00
48.00
9.00
15.00
94.00
428.00
322.00
198.00
46.00
102.00
406.00
521.O0
423.00
330.00
158.00
102.00
278.00

1
,

1
1

1
1

Asigned
Cases

4572.00
270.00
230.00
7.00
3.00
0.00
2176.00
16.00
12.00

0.00
490.00
116.00
70.00
0.00
39.00
102.80
336.00
210.00
96.00
0.00
2.00
200
856.00
644.00
396.00
0.00
0.00
0.00
1042.00
846.00
660.00
0.00
1.00
1-00

1
1

Change

2286.00
135.00
115.00
-62.00
-63.00
-387.00
1088.00
8.00
6.00

-265.00
245.00
58.00
35.00
-17.00
-21.00
-71.20
168.00
105.00
48.00
-9.00
-13.00
-92.00
428.00
322.00
198.00
-46.00
-102.00
-406.00
521.O0
423.00
330.00
-158.00
-191.00
-277.00

8.2.3.6 Practice Parameter Recommendations - 18% Reduction

Runs of the mode1 made at an 18% budget reduction indicate that cornmensurate cost
mode1 solutions can be identified for the resource allocation problem. Commensurate cost model
solutions c m be identified with sofl bounds on practice equal to *18% of current practice and
hard bounds of 160% of current practice. A surnrnary of cost model recommendations for
practice changes after an 18% budget reduction appears in Tables 8.35 to 8.37. See Appendk 1
for a complete list of volume and cost model recomrnendations d e r an 18% budget reduction.

Table 8.35 Cost Model LOS Distribi tion 18% Reduc

# of Cases LOS Set
Between Hard and
Soft Lower Bound

Assigned by Cost
Model

General & SpeUalty

1
1
1

Musculoskeletal
Thoracic

Value and Soft

Lower Bound

1,620.8
5,855.8
1,090.1

Oncotogy

14,584.3

Total

35,229.4

Table 8.36 Cost Model OR Time Dis :ribution 18% R duction

# of Cases OR Time
Set Between Hard
and Soft Lower

Assigned by Cost
Model

Bound
GI

222,869.5

General8 Speciaky

67,142.0

Set Between
Prefened Value and
Soft Lower Bound

Set at Prefened
Value

43

26

32

18

Musculoskeletal

120,938.7

Thoracic
- -

Oncology

Total

32,799.9

O
-

325,784.8

97

975,701.8

251

Table 8.37 Cost Model Direct Cost 1 stribution 18% Reduction

Assigned by Cost
Model

1 General & Speciafty 1

1 Total

$704,809

S13,689,772

# of Cases Direct
Cost Set Between
Hard and Soft Lower
Bound

# of Cases Direct
Cost Set Between
Preferred Value and
Soft Lower Bound

# of Cases Direct
Cost Set at
Preferred Value

8.3 Solution Quality

In runs of the allocation model made at cumulative budget reductions of 5 and 11%,
solutions to the resource allocation problem s a t i s w g the economic desires of both physicians
and the hospital can be identsed through changes to case mix aione. At the 18% budget

reduction level, however, the model is unable to identify a solution satisfying the economic
objectives of physicians through changes to case mix alone. These results suggest a number of
practical implications for case mix seleaion algorithms.
The most obvious conclusion that can be drawn fiom the application of the resource
allocation model to the budget reduction scenarios facing Mount Sinai Hospital is that a strategy
of pure case mix adjustment, while practical for small to medium sized budget reductions. is not,

of itself. strictly feasible for large scale budget reductions. Furthemore, this observation suggests
that improvements to clinical efficiency will continue to be necessary, even if case mix adjustment
is employed, when large scale budget reductions are required.
Nevertheless, if the results of the model are compared against a benchrnark of across the
board (ATB) reductions in case volume, the resource allocation model, even though imperfect,
can be seen to produce excellent results. Table 8.38 compares aggregate physician income
suggested by the model at each of the three budget reduction levels with hard case bounds set at
50%. high operating room time availability, and low inpatient bed availability against aggregate
physician income achievable under an equivalent, across-the-board reduction in case volume.
Table 8.39 compares institutional income recommended by the model against income expected
fiom an across the board reduction in volume. As is evident fiom the tables, the ailocation mode1
is aiways able to outperform an across-the-board reduction in case volume from the standpoint of
physician billings and institutional income. Clearly then, if physician practice parameters cannot

be changed in response to perturbations in extemal funding, the resource allocation model

provides solutions that dorninate traditional cost-cutting techniques. From this standpoint, the
rnodel produces high quality solutions.
Table 8.38 Physician Biing Comparison: Model vs. Across the Board Cuts
Budget
Reduction

Target
lncome

Model
lncome

Across the Board

lncome

ATB Scale
Factor

Table 8.39 Institutional Profit Comparison: Model vs. Across the Board Cuts
Budget
Reduction

Target
lncome

Model
lncome

Across the Board

lncome

ATB Scale
Factor

From the results listed in Tables 8.38 and 8.39, the quality of mode1 recomrnendations,
based solely on volume change is apparent. The resource model, however, provides two sets of
recommendations: a volume adjustment in which prices are assurned to be held fixed and a cost
adjustment, in which volume is assumed to be fixed. Tables 8.40 and 8.41 provide an indication
of the quality of model recommendations based on change to physician practice, rather than case
mix. The two tables list cost mode1 recornmendations for each clinical team at each of the three

budget reductions, under the assumption of hard bounds on cases equal to

* 50% of target,

high

operating room time availability, and low inpatient bed availability. The results s h o w in Tables
8.40 and 8.41 also assume hard bounds on physician practice equal to &50% of preferred practice

and cooperation between physicians.

I'able 8-40 LOS Cornparison: Model vs, Across the Board

Team

Budget Reduction
11%

Budget Reducon
5%

1
General8 Specialty
Musculoskeietal

Modei
BedDays

/
1

1,714.40

1
1

6,463.101

ATB
BedDays

1,851.45

Sedi$

ATB
BedDays

Modei
Bed Days

1,734.51

6,444.301

6.039.42

1,743.30

6.446.601

Budget Reduction
18Oh

Thoracic

1,218.90

1.232.15

1.155.20

1.154.32

Oncology

16,617.60

16.799.78

15,542.70

15.738.68

Total

M.814.60

40,814.60

38,236.70

38,236.70

ATB
Bed Days

Table 8.41 Direct Cost Com~arison:Model vs. Across the Board

Budget Reducon
11%

Budget Reduction
5%

AT6
Direct Cost

Model
Direct Cost

G1

$4,347,701

$4,240,969

Musculoskeletal

$2.423.731

Thoracic

$570.230

$566.892

S16,0.15.785

$16,015,784

$2.383.643

Budget Reduction
18%

Model
Direct Cost

ATB
Direct Cost

Model
Direct Cost

$4,059,423

$3,956,693

$3,748,584

92,388,644

52,223,865

$2.289.880

$518.795

$528,893

$458,438

AT6
Direct Cost
$3,625,044

$2,037,461

$484,561

1 f 14,902,232 1 $14,942.231 1 f 13,889,778 1

513,689,778

Oncology

Total

When the results of Tables 8.40 and 8.41 are analyzed and graphed, a number of
interesting trends emerge. Figure 8.18 presents a graph showing the bed days assigned by the
mode1 as a percentage of the bed days that would be assigned under an across the board reduction

in length of stay (ie Bed Day%,, / Bed Days,

* 100).

A value equal to 1000/0indicates the

model's recornmendations are equivalent to an across the board reduction in physician practice.
Values below LOO% indicate that the mode1 recornrnends bed days be reduced at a rate exceeding
that of an across the board change to practice. A vaiue above 100%' conversely, indicates the

244

model recommends a team's allocation be reduced at a rate less than that of an ATB cut.
RoughIy translated, dinical tearns with a model to ATB ratio greater than 100% have less
flexibility to deal with budget reductions through case mix change than do clinical teams with
ATB ratios less than 100%. Clinical tearns lacking case mix flexibility (ie not able to substitute

outpatient procedures for high cost or resource intensive inpatient cases), must maintain a greater
share of scarce resources to achieve their economic goals. Conversely, teams with a model to
ATB ratio less than 100% possess inherent flexibility to deal with budget reductions through

changes in case mix and are therefore less susceptible to fluctuations in resource availabilityFigure 8.18 The ratio of model recommendations to an across the board (ATB) reduction in total
bed days assigned to ctinical teams.

120h

110%

--

100%

--

90h

--

80h

--

Total Bed Days Assigned

Model vs. ATB

U)

Og

Ci

f a

gs
23

aa

B
GI

Gen

Msk

Tho

Onc

Den

245

No definitive concIusion, however, can be drawn from Figure 8.18 with regards to cost
model recomrnendations in cornparison with ATB cuts, since some clinical teams receive less
severe cuts under the model's recommendations (model to ATB ratio greater than 100%), while
others gamer recornrnendations for more severe restrictions (model to ATB ratio less than 100%).
It is thus impossible to Say that model recornmendations are better or worse than an ATB
reduction in bed days. It may merely be noted that the model recommendations are different from
an ATB cut. Nevertheless, model recommendations imply an important relationship between

economic desires and resource allocation. Model recomrnendations are therefore an important
tool that can be used by the hospital to plan resource allocation. Teams assigned more bed days
by the model than would be suggested by an ATB cut have less ability to ded with resource

reductions through case mix change alone. If these teams are to achieve their economic
aspirations, their access to these resources must be protected, or their clinical efficiency must
improve. Thus, from Figure 8.18, it c m be seen that as beds become more constrained. an
optimal allocation of resources would transfer beds fkom the GI and Oncology teams to the
Musculoskeletal, Thoracic, and DentaVEyeENT teams.
A similar pattern can be observed in the model's recomrnendations for direct cost dollars.

Model recornmendations for direct cost dollars differ from those expected under an ATB budget
reduction: Some dinical teams receive less severe recornrnendations than would be due under an
ATB cut, others receive more severe recomrnendations. Thus, it cm be seen that the model

c a ~ obe
t judged as producing cost recommendations that are better or worse than an ATB cut;
the model's recornrnendations are simply different. However, the relationship associated with
model to ATB ratios for LOS recommendations cm be extended to direct cost. Clinicai teams
with a mode1 to ATB ratio less than 100% have greater case mix flexibility than their counterparts

246

with a model to ATB ratio greater than, or equal to, 100%. Thus, 5om Figure 8.19, it c m be

concluded, as budgets shrink, an optimal policy would transfer funds fiom the General and
Specialty, Thoracic, Oncology, and Dent/Eye/ENT teams, to the GI and Musculoskeletal teams.
Figure 8.19 The ratio of model recommendations to an across the board (Am)reduciion in total

Total Direct Cost Dollars

Assigned
Model vs. AT6

GI Gen Msk Tho Onc Den

8.4 Practical Extensions

At the 18% budget reduction, the resource allocation model was unable to identi& a

solution satisfying the econornic goals of the hospital and al1 physicians. Even when hard bounds

on case rnix were relaxed to impractical levels (*100% of preferred values), a solution generating
sufficient billings for the Oncology tearn could not be found by modifjing only case rnix.
Therefore practice reforrns must be implemented, in addition to case mix changes, to ensure that

247

an acceptable solution can be found. There are three broad strategies that may be employed to
produce conditions in which a solution to the resource allocation problem based only on volume
changes can be identified. These strategies are:
1. Physicians may elect to accept a lower income than they enjoyed in the base

year of the scenario.

2. The hospital may elect to lower the h e d costs of production.
3. The hospital and its associated medical staff may decide through joint changes

to hospital processes or physician practices, to lower the variable costs of

production.
In the following sections, the impact of these mixed strategies is described. In each nin of
the mode1 it is assumed that a budget reduction of 18% is in efFect and that, as a practical policy.
the maximum allowable bounds on any case mix change is *40% of target, unless othenvise
noted.
8.4.1 Decreasing Acceptable Physician Income

Although al1 physician billing requirements cannot be achieved through changes to case
mix alone after an 18% reduction in budget, a solution satisfjmg the billing desires of some

physicians can be identified. Using an ATB reduction in volume as a benchmark, it is clear that a
solution to the resource allocation problern can be found if al1 physicians are willing to entertain a
reduction in income equal to the provincially imposed budget cut (ie, al1 case volumes are reduced
by 18%). Thus, as a lower bound on the problem, a reduction in billings equal to 18% will always
allow a solution to the resource allocation problem to be attained via case mix changes done. As
was shown in Section 8.2.3.1, a solution in which billings are reduced by 1O%, cannot be found,
given the preferred case rnix of the physicians in the Surgical and Penoperative Planning Council
at Mount Sinai Hospital. However, ifphysicians are willing to accept billings in the range 1018% of base year billings, a solution to the resource allocation problem should exist. Table 8.42

presents a list of billings for al1 physicians, Oncology tearn physicians, and non-Oncology tearn
physiaans at varying hard bounds on billings. These results are displayed graphically in Figure

Table 8.42 Physician Biiiings versus Hard Bounds on Biilings

I

Physician Billings

Hard
Boundson
Billings

Al1 Physicians

5%'

$8,080,329

10%'

Oncology

% of Target

Non-Oncology

All Physcians

$2,658,630

$5,421,699

95.00h

95.00%

95.00%

$7,655,049

$2.518,703

$5.136,346

90.00%

90.00%

90.00%

25%

$7,657,163

$2,104,852

$5,570,311

90.23%

75.20%

97.60%

30%

$7.657.163

$2,104,852

$5,570,311

90.23%

75.20%

97.60%

Oncology

Non-Oncology

Solutions at these bounds did not saris@ institutional income nquiremcnts

Figure 8.20 The impact of hard bounds on physician billinas.

Physcian Billings
% of M r e d BiIlings

Hard Bounds on niys'cian Wlings

O ncology Non-O ncology Totd
-1FInditutional lnoome Not Stided
-

---

Table 8.42 confims the hypothesis that a solution to the resource allocation problem
based only on case mix changes can be found, within case constraints of 140% of target, if
physicians are willing to accept a reduction in income of between 10 and 18%. In fact, Table 8.4 1

249

indicates that hard bounds on billings of 15% allows a solution based only on case mix change to
be identified. It may, therefore, be concluded that reducing physician desire for income is a
feasible rnixed strategy for the resource allocation problem. However, it is evident that such a
strategy violates the assumption that physicians are constrained profit satisficers. It is therefore
unlikely that reduction in physician income is a practical rnixed strategy for the resource allocation
problem.
Nevertheless, an interesting pattern ernerges when constraints on physician billings are
relaxed beyond the 18% level. In Figure 8.20 it can be seen that as bounds on billings are relaxed.
the total income of Oncology physicians decreases, whiie the income of non-Oncology physicians
increases. At bounds of *25% of target income, the biiiings of non-Oncology physicians, as a
group, is equal to 98% of target, with 3 of the 5 non-Oncology tearns achieving their preferred
billings; the income of Oncology physicians, however, decreases to 75% of its target value. The
observations in Table 8.42 and Figure 8.20 are important fiom a practical standpoint, since they
suggest Oncology may be a less efficient user of resources than other clinical tearns.
8.4.2 Lowering the Fixed Costs of Production

An 18% reduction in the budget of the Surgical & Penoperative Planning Council is equal

to a decrease in base budget of $7,106,741. In runs of the mode1 made in Section 8.2.3, it was
assumed fixed costs could be reduced in proportion to the decrease in provincial funding. At an
18% budget reduction, this corresponds to a decrease in fixed expenses of $3,886,116. Thus,

changes in case rnix contributing savings of $3,220,625 are necessary to make up the remainder of
the !7,106,741 savings required. If the fixed costs of production are reduced through concerted
efforts by management below 82% of their base year amount, however, the onus to achieve
additional costs savings through case rnix changes cm be eased. It is obvious that if the fixed

costs of production are reduced fiom base levels by a total of %7,106,741 (32.92%), no change is
required to physician case mix to meet the provincial budget cuts. Therefore, within the range of
fixed costs equal to $17,703,416 and $14,482,791, a solution must exist that will satisS>the
econornic aspirations of both physicians and the hospital.
Table 8.43 provides a cornparison of aggregate physician biliings versus reductions in
fixed costs of production, under the assumption of high inpatient bed availability (42,963 bed
days) and a maximum bound on case deviations equai to

* 40% of target; Table 8.44 lias

aggregate physician billings versus k e d cost reductions under the assumption of low inpatient
bed availability (35,229 bed days). Results from Table 8.43 and 8.44 are displayed graphically in
Figures 8.21 and 8 -22.

Table 8.43 Physician Biiiings versus Fixed Costs Inpatient Beds High (42,963)
1

Fixed Costs
Reduced

Al Physicians

by

Oncology

% of Target

Physician Billings
Non-Oncology

Ali Physicians

Oncology

Non-Oncology

18%

$7,655,049

$2.51 8,703

$5,136,346

90.00%

90.00%

90.00%

23%

$8,247,361

$2,540:711

$5,707,052

96.96%

90.78%

100.0%

28%

$8,505,610

$2,798,559

$5,707,052

100.00%

700.00%

100.00%

32%

$8,505,61O

$2,798,559

$5,707,052

100.00%

100.00%

100.00%

Table 8.44 Physician Billings versus Fixed Costs - Inpatient Beds Low (35,229)
Fiied Costs
Reduced
by

% of Target

Physician Billings

AH

Physicians

Oncology

Non-Oncology

AI Physicians

Oncology

Non-Oncology

18%

$7,655,049

$2,518,703

$5,136,346

90.00%

90.00%

90.00%

23%

$8,206,561

$2,562.167

$5,644,394

96.48%

91-55%

98.90%

Figure 8.21 The impact of fixed cost on physician billings after an

18% budget reduction. Inpatient bed availability is high (42,963).

Physaan BTllings
% of Des'red BSllins
1 05%

85%

- - - - - - - - - - - - - - - - - - - - - - - a - - - - - - - - - - - - - - - - - - - - - -

80%

----------------------------------------------

75%

-'

28%

23%

18%

32%

Fixed C a M u d i o n
I

Oncoloq Non-Oncolog Total

\
1

-t-

Figure 8.22 The impact of fixed cost on physician billings after an

Physician Billings
% of Desred Billings

252

The results of the analysis support the hypothesis that a solution to the resource allocation
problem c m be identified by varying only case rnix after an 18% reduction in base budget if the
fixed coas of production are reduced sufficiently. In Tables 8.43 and 8.44 it can be seen that a
reduction in fked costs of production equal to approximately 25% of the base year amount
($5,397,383) is necessary before a solution satisfjmg the econornic desires of d physicians
through case rnix changes less than, or equal to, 40% of target cm be identified. However, the
results in Table 8.43 are based on the assumption of high inpatient bed availability (ie, the number
of bed days available is unchanged from 1993194 levels). It is unlikely that Mount Sinai would be
able to achieve significant reductions in overhead costs without a corresponding decrease in

inpatient beds. Table 8.44 and Figure 8.22, therefore, present the results of a similar analysis
comparing physician billings against reductions in fixed overhead under the assumption that total
bed days are reduced by 18% fiom their base year totals and case mix changes rnay be no greater
than *40% of target values. The results of this analysis show that a solution to the resource
allocation acceptable to ali physicians, in which oniy case mix is varied, cannot be found when bed
availability is low. It may therefore be concluded that bed availability constrains case mix
selection and, by extension, physician billings, at an 18% budget reduction when maximum
bounds on case mix are defined to be *40% of target income. Therefore, reductions in fixed costs
coupled with case mix change is an impractical rnixed strategy for resource allocation decisions
necessitated by large reductions in institutional funding. From the results of Table 8.44, it is
obvious that improvement in clinical efficiency, specifically related to reductions in direct cost and
length of stay are a necessary cornponent of any rnixed strategy for the resource allocation ansing
as a result of a large budget reduction.

253
8.4.3 Improving Clinical Efficiency

Perhaps the most obvious mked strategy to employ in instances where a solution to the
resource allocation problem, based solely on case rnix adjustments, cannot be identified is to
conjointly improve the clinical efficiency with which services are delivered, while making
adjustments to case mix. This strategy, because it relies on improvements to clinicai efficiency, is
closely aligned with the current methodologies used by Mount Sinai to respond to changes in
provincial funding and was identitied by CO-teamleaders during face validity interviews as the
preferred method for dealing with a shrinking resource base.
As was the case for physician billings bounds, a lower bound on clinical efficiency afler an

18% budget reduction may be inferred by comparison to an ATB benchmark; if clinical efficiency
improves by 18%. a solution to the resource allocation problem can be found that does not
require any change to case rnix. As has been demonstrated, a solution based only on case mix
change, without improvements in clinical efficiency, cannot be found. Therefore, within the range
of clinical efficiency improvement equal to 0% and 18%, a solution should exist in which a rnixed
strategy is feasible. Table 8.45 lists billings for al1 physicians as clinical efficiency improvernent
ranges between O and 12%. The results in Table 8.45 are displayed graphically in Figure 8.23. In
each run of the mode1 used to develop Table 8.45 and Figure 8.23, a maximum bound on case mix
changes equal to &40% of 1993/94 targets was assumed. The availability of operating room time
was assumed to be high while inpatient bed availability was assumed to be low.

Table 8.45 Phvsician Biiiines versus Ciinicai Efficiencv Im~rovements

Change in
Clinicat
Eniciency

Physician Billings
Ail Physicians

Oncology

O
h of Target
Non-Oncology

AI Physicians

Oncology

Non-Oncology

0%"

$7,655,049

$2,518,703

$5.1 36,346

90.00%

90.00%

90.00%

2%

$7,663,378

$2.51 8,703

$5, 144,675

90.10%

90.00%

90.15%

4%

$7,951,232

$2,518,703

$5,432,529

93.48%

90.00%

95.19%

6%

$8,142,557

$2.51 8,703

$5,623,854

95.73%

90.00%

98.54%

8%

$8,292,698

$2,585,647

$5,707,051

97.50%

92.39%

100.00%

10%

$8,417,731

$2,710,680

$5,707,051

98.97%

96.86%

100.00%

12%

$8.505.61 0

$2,798,559

$5,707,052

100.00%

10O.OO0h

100.00%

Solutions at these bounds did not satise institutional incarne rcquirtmcnts.

Figure 8.23 The impact of clinical efficiency improvement on

physician billings after an 18% budget reduction. Inpatient bed
availability is assumed to be low (35,229).

Physiaan Billings
% of M r e d Billings
105%

From the results of Table 8.42 it can be concluded that a rnixed strategy involving
improvements in clinicai efficiency, combined with adjustrnents to case rnix, is feasible for budget
reductions equal to 18%. From Figure 8.23, it is also apparent that a solution acceptable to al1
non-Oncology physicians, can be found under a rnixed strategy in which clinicd efficiency is

255

improved by 8% over the base year. A solution acceptable to d physicians c m be identified

through a mixed strategy when the clinicai efficiency of d providers is improved by 12%. This
solution provides an intermediate method for addressing resource allocation problems after large
reductions in provincial funding. It requires a Iess onerous change in physician practice than
would be necessary under an across the board reduction in volume or ciinical practice. However,
as is evident fiom Figure 8.23, the vast majority of physician desires can be achieved through a
rnixed strategy with oniy small improvements in ciinical eficiency. The economic desires of
Oncology physicians are more difficult to satisfjr; large changes in the practice of al1 physicians are
necessary before the economic desires of Oncology physicians c m be satisfied. This presents an
interesting problem in motivation. Non-Oncology physicians, who require an improvement in
efficiency of only 8% to support their economic desires under a mixed strategy, must improve
their efficiency by a further 4% so that Oncology physicians may achieve their econornic goals as

well. In Table 8.46, results of the mode1 are listed under the assumption that non-Oncology teams
improve their ciinical efficiency by a maximum of 8%; the remainder of the clinical efficiencies
necessary to support a mixed strategy are assumed to corne fiom improvements in the efficiency

of the Oncology team. As was assumed in the results of the global improvement in clinical
efficiency, results listed in Table 8.46 assume a maximum bounds on case mix changes equal to
WO% of target, high operating room availability, and low inpatient bed availability.

- --

Change in
Oncology
Team
Clinical
Efficiency*

Physician Biiiings versus Oncology Clinil

P hysician Billings

A l Physicians

Oncology

Oh

Non-Oncology

Al Physicians

of Target

Oncology

Non-ncology

--

1 8%
$5,707,052
100.00%
Non-Oncology tearns are assumai to have achicvcd an 8% improverncnt i: clinical eficiency.

100.00%

100.00%

Figure 8.24 The effect of improvements in Oncology tearn clinical

efficiency. Al1 other tearns are assumed to have improved clinical
efficiency by 8%.

Phyadan Billin% of Des'red Billings

From the results of Table 8.43, it can be seen that targeting the clinical efficiency
with which Oncology seMces are delivered at Mount Sinai produces a feasible rnixed strategy
solution to the resource allocation. Because Oncology generates a Iarger share of cost savings,
the onus for increased efficiency fiom tearns for which there is no economic benefit, is eliminated.

257

8.5 Using the Mode1 to Determine Mixed Strategies

As has been noted, the resource allocation mode1 provides two points on the production

possibility fiontier of a hospital: One point, in which the costs of production are held fixed and the
volume of products is allowed to Vary and a second point, in which the volume of production is
held fixed while the costs of production are allowed to vary. From results of practical tests of the
model it was observed that, in some circumstances, the point on the production possibility curve
involving only change in volume f ~ l to
s satisq the economic desires of both the hospital and its
medical s t a . See Figure 8.25 for a hypothetical view of the production possibility fiontier.
Figure 8.25 Theoretical production possibility cunie.

Theoreticai Rodudion PogSbility Curve

Mount ana Hospitd
% Change Due to Volume
1.2 1

O -2

0.4

0.6

O .8

1.2

% Change Due to Rioes

In Figure 8.25, point V represents the volume model solution, in which the total savings

necessary to achieve a feasible solution come entirely fiom changes in volume. Point P is the
point on the production possibiiity curve where al1 necessary cost savings come fiom reductions in
price. Points A and B represent intermediate points on the curve, where prerequisite savings
come fiom a combination of price change and volume change. Points A and B can, therefore, be

258

thought of as representing the two points on the production possibility curve identified in Section
8.4.3.

While points on the production possibility curve can be generated rnanuaily, the model can
also be used to investigate the production possibitity curve automatically. When the model is
solved for any data set, the points P and V are identified. Point V is the solution obtained by
adjusting ody volume, P is the solution obtained by adjusting only price. At large budget
reduction levels, V may not fulfili the economic desires of aU physicians. If the two solutions are
presented to decision makers, it is obvious that P represents a more desirable solution since it
satisfies all economic objectives. P,however, also represents a new set of prices. If these new
production pnces (or a portion of the irnprovement implied by them) are assumed over the
original costs of production, the mode1 may be re-iterated, to provide new recommendations for
both case mix and practice pararneters. This process can be repeated until an acceptable mixed
strategy involving both price and volume changes is identified. Therefore, an algorithm for
automatically investigating the production possibiiity fiontier using the resource allocation may be
postdated:
1.

Solve the resource allocation mode1 retaining user specified

bounds on case and practice variation.

lf the economic desires of all physicians are met by the volume

model solution, stop. Othenvise proceed to step 3.
IdentifLthe "price" of production (LOS, case time, direct case
cost) recomrnended by the cost mode1 for each doctor/CMG pair,
in which the physician is unable to achieve desired billings
through case rnix change alone. Define A,, the implied
irnprovement in efficiency, to be the dserence between the
current and recornmended pnce for each doctor/CMG pair:

A,,,,yn(LOS) -Ap(o~)

, /, -

- /
- Op -

Am y n ( ~ t m i ) = k,

l,,,, for al! dissatisfied y, n;

Om

- ,k

for dissatisfied ail y, n;

for dissatz.@ed ail y, n;

Where:

I, is the

model recommended Iength of stuy.

$, is the cuvent le@

O
Yn
O,

of s w .

is the rnodel recommended case time.

is the m e n t case tirne.

kP is the modei recommemied direct cost.

k; is the m e n t direct cost.

4.

Update the current cost of production for each doctodCMG

combination by subtracting a &action, s, of each recommended
improvement in efficiency fiom the current cost of production.

5. Go to Step 1.

The algorithm outlined above simulates the decision process entered into with decision
makers when negotiating a price and volume contract. At each iteration of the revised model, the
recommendations of the volume and cost models are compared. If both models satisfi the
economic desires of the hospitai and its associated physicians, the algorithm halts, since there is
no clear rationaie for accepting one solution over the other. If, however, the volume model
solution does not satisQ a particular physician's economic desires, the cost model solution is

260
assumed to be preferable. Given that decision makers when confionted with the two, unequal
views have an economically rational reason for preferring the cost rnodel recommendations, it cm
be assumed that any fraction (fiom O to 100%) of the cost rnodel's recomrnendation for
improvements to clinical efficiency will also be acceptable. Therefore, a new pnce for clinicd
services, acceptable to dissatisfied physicians can be determined. Using this price, the model may
be re-iterated to determine if a mixed strategy c m be identified for the resource allocation
problem. The algorithm continues until a mted strategy acceptable to al1 physicians is found.
This algorithm has a number of interesting feahires. It allows a solution (if one exists) to

be automatically identified to any resource allocation problem, regardless of the ievel of fnding
reductions or the hard bounds on case rnix changes in effiect. Furthemore, because the resource
allocation model determines the "best" allotment of resources at each step of the iteration, model
recommended changes in practice and volume are deterrnined normatively; the model's
recommendations are based on what is appropnate for al1 providen, given the resources globally
available, the efficiency with which tearns use these resources, and their case mix flexibility. Thus,
inefficient services will be given recommendations for larger reductions in price, length of stay, or
case time, than efficient providers. The model, furthemore, functions within the parameters
onginally designated by the user at mn time. Therefore, no recomrnendations retumed by the
mode1 violate specified bounds on case m k change or practice parameters.
The algorithm uses a fraction, S. of the recommendation implied by the cost mode1 at each

iteration, rather than the complete recomrnendations for improved clinical efficiency, to maxirnize
the proportion of savings due to volume changes. If the value of s is set too large, the model
tends to assume an aggressively determined improvement in efficiency at each step of the
iteration. In the lirnit, when s=I, the savings necessary to meet the provincially imposed budget

cut are alrnost entirely achieved through pnce reductions. Kfintermediate values of s are assumed
at

each iteration of the model, a greater proportion of savings attributable to volume changes is

observed. See Table 8.47 and Figure 8.26.

Table 8.47 The im9act of s on price & volume

Total Savings Due to
Price Change

Total Savings Due to

Volume Change

20%

$1,757,461

$1,286,411

57.74%

42.26%

40%

$1,484.594

$1,559,278

48.77%

51.23%

60%

$1,623,982

$1,419,890

53+35%

46.65%

80%

$2,165,312

$878,560

71.14%

28.86%

100%

$2,706,643

$337,229

88.92%

11.08%

Price Model
Recomrnendation
Fraction

% Savings Due to
Price Change

% Savings Due to
Volume Change

Figure 8.26 The impact of step size on priceholume savings.

PriceNolume Tradeoffs

0%

!
0%

20%

40%

60%

80%

100%

Step Se (s)
-

- -

From Figure 8.26 it is apparent that values of s between 40% and 60% yield results in
which case mix plays a substantially larger part in producing the total cost savings necessary to

meet reductions in funding imposed by the provincial governrnent. In Tables 8.48-8.5 1 a

262

summary of recommendations arising fiom an iterative application of the resource ailocation

mode1 in which the value of s was set at 60% are presented. The results in this table assume sofi
bounds on case mix changes of 18% and hard bounds of 60%; bed availability is assumed to be
low, but operating room avaiiability is assumed to be high. The results also assume soft bounds
on practice change recomrnendations of 1800 and hard bounds of 60%. Total hospital income
was set equal to the target value of $98 1,958.
Table 8.48 Team Billing Sumrnary, Iterative Model, Step Size 60%
Team

Target
Billings

Oncology

DentaEyeENT

Cost Model
Bilings

$2,798,559

$2,798,559

$2,798,559

$1,562,516

$1,562,516

$1,562,516

$8,518,918

$8,518,918

Total

Volume Mode1
Billings

-- - - -

$8,518,918

Tables 8.48-8.5 1 show that the iterative algorithm is able to identify a feasible mixed
strategy solution to the resource allocation problem after an 18% budget reduction without
demanding change in case mix greater than

* 60% of target. In Table 8.48, which lists physician

billings, it can be seen that the iterative algorithm produces a solution meeting the economic
desires of al1 physicians, as well as the hospital. This solution is achieved through a combination
of changes to physician practice and case mix, which are surnmaized in Tables 8.49-8.52.

rable 8.49 Team Bed Day Summary, Iterative Model, Step Size 60%
Team

Target
Bed Oays

Revised Costs for

Volume Model
Total Bed
Da~s

Musculoskeletal

6.786

6.130

Volume Model
Recommendations
Total Bed

Savings Due
to Cost
Change

9.67%

Savings Due
ta Volume
Change

Da-

5,978

2.24%

--

Oncology

I
1

17.684

15.914

10.02%

13,9521

11.08%

Total

42,963

38.879

9.50%

35,256

8.43%

Thoracic

1,118
T -13.84%

Table 8.49 lists model recomrnendations for changes in the total bed days assigned to each

clhicai team. From the table it can be seen, for example, that model recommendations suggest a
decrease in the bed days assigned to the GI team fiom 12,684 to 10,203. Of this amount, 1,656

bed days ( 12,684- 1 1,O28),or 13.OS%, are due to increases in clinical efficiency, while 825 bed
days (1 1,028-10,203), or 6.05%, are due to changes in case mix. Sirnilar results are listed for
each team. By comparing team results, it can be seen that the model suggests large LOS
reductions for the GI and Oncology teams.

Table 8-50 Team OR Tim Summary, Iterative Model, Step Size 60%
Team

Revised Costs for

Volume Model

Target
OR Minutes

Total OR
Minutes

Thoracic

1
1

Oncology

325.785

325.785

Total

975.702

975.702

Musculoskeletal

1
32.800 1

120,939

Volume Modd
Recommendations

1
32,800 1

Savings Due
to Volume
Change

TotalOR
Minutes

SavingsDue
to Cost
Change

0.00%

1
1

O.OOOh

325.785

0.00%

0.00%

975.702

O.0O0h

120,939

0.00%

0.00%

32,800 (

0.00%

120,939

Table 8.50 gives model recommendations for changes in the total operating room time
assigned to each clinical team. Because the model was run under the assumption of hi&
operating room time availability, no changes to case time were necessary
Table 8.51 Team Direct Cost Summary, Iterative Model, Step Size 60%
Team

Revised Cos& for

Volume Model

Target
Direct Cost

Total Direct
Cost

% Savings

Due to Cost
Change

Volume Model
Recommendations
Total Direct
Cost

% Savings
Due to
Volume

Change

1 Musculoskeletal
1 Thoracic

1 Total

$598.558

$598.558

S16.910.443 1 2 1 5 , 4 2 6

Table 8.5 1 provides a view of model recommendations for changes in direct cost after an
application of the iterative algorithm to the resource allocation problem. The table lists

recommendations for reductions in direct case cost through both increases in clinical efficiency

and changes to team case rnix. Particularly large improvements in the clinical efficiency of the
Oncology tearn are indicated; smaller reductions in the variable case cost of the GI and
Musculoskeletal teams are necessary. No changes in variable case cost are required of the
Generai and Speciaity, Thoracic, and DentaVEyeENT teams.
Recornmendations for improvements to clinical efficiency listed in Table 8.5 1 imply an
interesting relationship between physician practice and economic aspirations. If the results of
Table 8.5 1 are compared to Table 8.52, which lists the average physician billings achieved per
direct cost dollar expended, a correlation can be seen between the size of clinical efficiency
recommendations and marginal billings per direct cost dollar.

Table 8.52 Marginal billings per direct cost dollar cornparison with model results
Physician Bi1lings
Per Direct Cost Dollar

Rank - Physician
Billings per Direct Cost
Dollar
(1 is best)

Rank - Model
Recornmendations for
Reductions in Direct
Cost
(1 is best)

GI

$0-47

General & Specialty

$0.69

Musculoskeletal

$0.43

Thoracic

$0.50

1'

Oncology

$0.41

Team

From the results listed in Table 8.52, it is apparent that iterative model recomrnendations
for improvements in clinical efficiency are closely related to physicians' marginal billings per direct
cost dollar. The rank ordering of the mode1 recommendations listed in Table 8.52 is nearly
identical to the rank ordering of marginal physician billings (only the ordering of the GI and
Musculoskeletal teams are interchanged.) If, however, restrictions on the minimum number of

266

cases that must be completed either because of restrictions placed on case rnix by clinical team
leaders or non-deterministic demand are relaxed the rank o r d e ~ of
g model recornmendations is
identical to the rank ordenng of marginal physician biliings per direct cost dollar (see Table 8.53).
The reason for this is straightfonvard. When maximum change to case mix is lirnited to *60%,
physician desires for billings cannot be satisfied through case rnix change alone. This shonage
arises because of restrictions on available resources in general and direct cost dollars in particular.
(In linear prograrnrning parlance, the constraint on direct cost dollars is said to be "tight".) Thus,
when the model decides on case mix, the optimal solution to the problem is to increase the
volume of CMG/physician pairs with high billing to direct cost ratios, while simuitaneously
decreasing CMG/physician pairs with low billing to direct cost ratios.

Table 8.53 Marginal billings per direct cost douar cornparison with model results after
minimum case bounds are eliminated.
1

Physician Billings
Per Direct Cost Dollar

Rank - Physician
Billings per Direct Cost
Dollar
(1 is best)

General & Specialty

$0.69

Musculoskeletal

$0.43

$0.50

$0.99

1 Thoracic
DentiEye/ENT

Rank - Model
Recommendations for
Reductions in Direct
cost
(1 is best)

'Tied.
From the results shown in Tables 8.5 1-8.53, it can be seen that model recommendations

for improvements in clinical efficiency are normatively based; the model makes recommendations
for improvements in efficiency in instances where tearns are less efficient users of direct cost
dollars relative to their peers.

267

It should be noted, however, that the normative efficiency described above refers to the
ability of physicians to use hospital resources to generate income, rather than clinical efficiency
with which hospital inputs are converted to patient outcornes. Thus, we may note that the mode1
normatively determines the "biliing efficiency" of clinical teams. This result is a function of the
model's assumption that physicians are constrained profit satisficers; decisions on case rnix, when

resources are tightly constrained, are based on the ability of physicians to convert resources into
income and generate hospital revenue. This observation has a number of interesting implications.
For example, clinicd tearns with high billing to direct cost ratios may be econornicdy better off if
they pay less efficient seMces not to provide patient s e ~ c e s .By so doing they could secure a
larger share of necessary scarce resources, and hence retain a greater proportion of their desired
incornes.
Table 8.54 Team Case Mir Change Summary, Iterativ
1

General & SpeciaHy Surgery

Musculoskeletal

Total Case MD(

Changes
Recommended by
Volume Model*

1,823.0

517.0

Oncology

Recornmended by
Volume Model

1.194.0

Thoracic

Total

Target Cases

3,535.0

14,260.0

Total case mu< changes are determincd by rurnming the difference k m prcferred
volume and recommendcd volume for cach physicianXMG pair.

In Table 8.54, mode1 recommendations for case mix changes in response to an 18%
budget reduction are listed. The first column of the table gives the target volume for each team.
The second column lists the total number of cases suggested by the volume component of the

resource allocation mode1 at the termination of the iterative solution process. The third colurnn
provides a surnmary of the total case changes suggested by the mode1 (lr,

1%. - n, 1 for al1 y).

As is evident fiom the table, a large change in Oncology tearn case mix is necessary to satise the

economic desires of al1 physicians d e r an 18% reduction in the provincial budget. A summary of
mode1 recornrnendations involving large case mix changes appears in Table 8.55.

Table 8.55 Recommended Changes to Case Mix Iterative Solution

Team

CMG

Prefened Assigned
Cases
Cases

Description

387
135-

472.21

85.2

216.01

81-0,

231

9-01

-14.01

27.0 1
1284.0
264.0

-4201
196.0
41.O

691
1088
223
40
265245 (

581
35 /

93.0
56.0

151

6.0-

000
DO0

OTHER DEFINED BY MODEL

OTHER DEFiNED BY MODEL

1265

IA~~DOMINAL
LAPAROSCOPY

1256
D35
DO0
~ 6 1
000
Musculoskeletal 000
380
379
350

ILESS
EXTENSIVE ESOPHAGEAL, STOM

BIOPSY
OTHER DEFINED BY MODEL
OTHER DEFINED BY MODEL
OTHER LOWER XlREMIfY PROCEDURE
OTHER MUSCULOSKELETAL PROCEDURE
MULTIPLE OR BILATERAL JOINT RE

IWOR LQWER EXTREMrrY PROCEDURE 1

1
General
& Speciatty

Thoracic

I
Oncology

1370
,000
DO0
028
000
703
187
192
028
DO0
736
383
076
125

loenVEyUENT
+

060

POO
084
056

BIADDER & URETHRALPROCEDURES

OTHER DEFINED BY MODEL

OTHER DEFINED BY MODEL

OTHER DEFINED BY MODEL
GI PROCEDURES(ENDOSCOPY)
OTHER DEFINED BY MODEL
OTHER O.R. PROCEDURES OF THE 6
OTHER VASCULAR PROCEDURES
VElN LlGATlON AND STRIPPING
GI PROCEDURES (ENDOSCOPY)
OTHER DEFINED BY MODEL
CHEMOTHERAPY
JOINT REPiACEMEM FOR MAUGNAN
MAJOR HEA AND NECK PROCEDURES
[MAJOR CHEST PROCEDURES
LENS PROCEDURES
DENTAL suRGERY
~OTHERDEFINED BY MODEL
MISCELLANEOUS EAR, NOSE AND TH
LENS INSERTIONWITHOUT CC [MNR

Change

15
94,
428
322
198
31
34
102
521
423

14.21
142.0
269.0
124.0
57.0
3.0
6.0
77.0
631-2
516.0
317.0
12.0
13.0
40.0
792.1
500.0

3301
48
278

390.0 1
0.0
227.0

251
174
t 68
105
48
9

1
1

32.O
210.7
392.0

-8.0
-54.3

147.0
35.0
21O
.

-9.01
-10.81
-32.0
101.O
19.0
9.0
-6.0
-9.0
-17.0

203.2
194.0
119.0
-19.0
-21.O

-62.0
271.1

n.o
60.01
-48.0
-51.O

269

While the iterative algonthm is able to i d e n t e a mixed strategy solution to the resource
allocation problem satisfjmg the economic aspirations of the hospital and al1 its associated
medicd staff, a strategy based purely on price change is also feasible. Tables 8.56 - 8.58 contrast
the economic efficiency necessary to idente a solution to the resource allocation problem under a
mixed strategy and a strategy based solely on price changes. The results in the tables are based on

sofl bounds on practice &OS, case time, direct case cost) equal to 18% and hard bounds on
practice equal to

* 60% of target.
-

Table 8.56 LOS Recommendation Distribution Iterative Solution

Team

# of Cases
LOS Set
Between
Hard and
Soft Lower
Bound

1 General & Speeiaty

Thoracic

# of Cases
LOS Set
Between
Preferred
Value and
Soft Lower
Bound

# of Cases
LOS Set at
Preferred
Value

Cost Model Recommendations

- -

1
1
1

# of Cases
# of Cases
L O S S ~ ~ LOS Set
Between
Between
Hard and
Preferred
Soft Lower
Value and
Bound
Soft Lower
Bound

# of Cases
LOS Set at
Preferred
Value

Musculoskeletal

1 Total

Volume Model Recommendations

11

Table 8.56 presents a view of the distribution of model recomrnendations for changes in
length of stay at the termination of the iterative algonthm; the number of cases in which model
recommendations are set in the ranges defined by [Hard Lower Bound on LOS, Sofi Lower
Bound on LOS), [Sofi Lower Bound on LOS, Preferred LOS), and [Preferred LOS, Preferred

LOS] are listed for both the volume and cost models. Volume model recomrnendations
surnmarize the nature and the impact of required changes in length of stay, if a rnixed strategy is
accepted. Cost model recommendations descnbe the impact of LOS changes if a purely price

based strategy is accepted. From the results listed in Table 8.56, it c m be seen that a price based
strategy requires greater reductions in length of stay. In particular, 20 additional cases, of which
13 are GI or Oncology cases, must undergo significant reductions in LOS under a pure pnce

strategy.

Table 8.57 Case Time Recommendation Distribution Iterative Solution

Volume Model Recommendations

# of Cases
Case Time
Set
Between
Hard and
Soft Lower

Bound

# of Cases
Case Time
Set
Between
Preferred
Value and
Soft Lower
Bound

# of Cases
Case Xme
Set at
Preferred
Value

Cost Model Recommendations

# of Cases
Case Time
Set
Between
Hard and
Soft Lower
Bound

# of Cases
Case fime
Set
Between
Preferred
Value and
Soft Lower
Bound

# of Cases
Case Time
Set at
Preferred
Value

General & Speciaity

1 Total
Table 8.57 presents a view of the distribution of mode1 recornmendations for changes in

case time at the termination of the iterative algorithm. As can be seen from the table, no changes
in case time are required under either strategy.

Table 8.58 Direct Cost Recommendation Dist ibution II rrative Solution

Volume Model Recomn endations

# of Cases
Direct Cost

# of Cases
Direct Cost

Between
Hard and
Soft Lower
Bound

Between

Set

Prefened
Value and
Soft Lower
Bound

1 General & Speciaky

1 Total

15

Set at
Preferred
Value

.-

Cost Model Recomme

# of Cases
Direct Cost
Set
Between
Hard and
Soft Lower
Bound

# of Cases
Direct Cost
Set
Between
Preferred
Value and
Soft Lower
Bound

# o f Case.
Direct Cost
Preferred
Value

106

Table 8.58 presents the distribution of model recommendations for changes in direct case
cost at the termination of the iterative aigorithm. From the table, it is evident that a larger number
of cases must have substantial reductions in direct case cost under a pure price strategy than

under a mixed strategy. In total, 61 additional cases, of which 45 are Oncology cases, rnust have
reductions greater than the soft bound on practice change (&18% of target) to suppon a strategy
based purely on price change.
From the results of Table 8.56 through 8.58 it can be seen that the mixed strategy requires
l e s onerous changes in physician practice than does a strategy based purely on pnce. Therefore,
we may conclude that in an environment where ciinical efficiency improvements are increasingly
harder to achieve, a rnixed strategy approach to the resource allocation problem is appropriate.

See Appendix J for a sumrnary of mixed strategy model recomrnendations.

9.0 Conclusions
Case mix management, as has been demonstrated, is a feasible technique that cm be
ernployed by acute care hospitals facing budget reductions. Case rnix, when appropriately set, can
be used to jointly achieve the economic goals of acute care institutions and their associated
medical staff, while preserving patient access to care in an environment of dirninishing physical
and financial resources.
The results of a practical application of the model to the decision problem facing the
Surgical and Perioperative Planning Council at Mount Sinai Hospital show that goal programming
models can be used to help institutions adjust to reductions in hnding available fiom extemal
sources. Using the goal programrning models defined in Chapter 5, solutions to the resource
allocation problem satisfjmg the econornic goals of the hospital and al1 its surgical teams were
identified for budget reductions of 5, 11, and 18%. At the 5 and 1 1% budget reduction levels,
acceptable solutions employing only changes to case mix were identified. At an 18% budget
reduction, a mixed strategy consisting of changes in case mix coupied with improvernents in
clinical efficiency was necessary to achieve al1 economic goals. However, by employing the
resource allocation model's ability to generate conunensurate costing policies f?om a case rnix
assignment. the mode1 can be used to automatically identiQ both the cost and volume changes
necessary to achieve desired savings, preserve clinicaily important programs, and support the
economic desires of physicians and the hospital.
A number of important conclusions may be drawn fiom the model results. A synopsis of

conclusions drawn fiom the model appears in this chapter. These observations can be separated
into conclusions related to policy, institutional operations and technical modeiling issues.

273
9.1 Policy Conclusions

In addition to generating recommendations regarding case rnix and physician practice in

response to changes in extemal fnding, the resource allocation model may be used as a test
platform to investigate policy issues related to hospital andor physician reimbursement. Of
particular interest to policy analysts, physicians and hospitai management is the conclusion that
hospitai funding mechanism has no impact on the resource allocation problem, when identical (ie
dollar-for-dollar) reductions in funding is assumed.
While there is an empirical basis for believing that physician and institutional motivation
will change when funding mechanisms convert from global funding to a rate-based system,
practical results indicate that solutions to the resource allocation problem are identical under both
methodologies, if the province empioys retains controi of the totolfunds transfewed to institution

as a policy mechanism under rde-basedfunding. This observation has significant implications

for hospitals and physicians in the transition between the two fnding systerns. Because the net
effect of reductions under both funding methodologies are identical, institutions and physicians,
who are faced with an uncertain tirnetable for the full implementation of a rate-based funding
system need not be concerned with the actual rnechanism in place. Results from the model
indicate that the magnitude of the fnding reduction, rather than the mechanism in place,
determines the response necessary. Thus, planning for case mix adjustments and clinical
efficiency improvernents can be made independently of the funding mechanism in place. Results
fi-om the model also show that if'budget reductions are anticipated after the transition from global
funding to a constrained, rate-based fnding mechanism a transition strategy that achieves breakeven in the period irnmediately pnor to the imposition of the new fnding mechanism is optimal.
This strategy will rninimize the absolute value of funding reductions in subsequent periods.

274
A second policy conclusion that can be drawn fkom the model is that improvement to

ciinicai eEciency remains an important and necessary strategy for institutions functioning under a
rate-based funding model. Of some concem to policy analysts is the fear that, under a rate-based
model, institutions may abandon inefficient or unprofitable services in order to seek out more
lucrative or less expensive services. The existence of a perverse incentive to dump necessary
care, however, is not supported by nins of the model or discussions with care providers
undertaken as part of model validation. When reviewing the model with chical team leaders,
surgeons, and hospital administrators, a cornmon theme of presewing important cases. while
rninimizing case mix changes to less important procedures was noted. This observation supports
model assumptions that providers' preference is to maintain, as much as possible, a consistent
case mix and volume. Another ovenvhelming theme in interviews with team leaders was the
necessity of exhausting improvements in clinical efficiency before undertaking case mUr
adjustments as a cost cutting strategy. Therefore, it may be concluded that a perverse incentive to
shift care fiom high cost, high intensity or Iow profit cases to low cost, low intensity or lucrative
cases, does not exist in practice. Furthermore, results of the model suggest that such a policy is
unlikely to produce case mixes that satisQ the economic desires of al1 physicians. Therefore, the
importance of improvements to clinical efficiency is not necessarily compromised by a constrained
rate-based method of funding.
9.2 Institutional Conclusions

A number of practical conclusions related to the operation of Mount Sinai Hospital c m be

denved from the model results. Of note are the model results related to the availability of
productive resources and the effect of lowered availability on the capabiiity of the institution and
its medicai staff to achieve their economic and clinicai goals.

275

Tests conducted at the 5, 11, and 18% budget reduction levels dl indicate that physician
b i h g s are highly dependent on the avdabiiity of operating room tirne; as operating room time is
reduced, the ability of physicians within the Surgical and Perioperative Planning Council at Mount
Sinai Hospital to satisfj. their econornic aspirations, as weU as those of the hospital, are seriously
compromised. At an 18% budget reduction, aggregate physician billings are, on average, 4.14%
lower when operating room availability is reduced to a level cornmensurate with the provinciaily
imposed cut; institutionai income is reduced, on average by 0.2%. From these results it may be
concluded that policies of resource reductions undertaken as a cost savings measure should avoid
reducing, as much as possible, the total amount of operating room tirne available to surgeons.

Physician billings and institutional income were not, however, sigruficantly infiuenced by
the availability of inpatient bed days. In most scenarios tested, inpatient bed availability was not
observed to influence physician incorne. This is an important practical conclusion for the research
site. The model results confirm the hospital's policy of reducing fixed and variable costs through
inpatient bed reductions. The model indicates physicians are able to make substitutions within
their case mix in response to inpatient bed reductions, predorninately by shifting inpatient
procedures to outpatient cases. At an 18% budget reduction, the resource allocation mode1
recommends a 28% increase in the number of outpatient procedures and a 4.18% decrease in
inpatient procedures as part of a mixed strategy for responding to a decrease in hnding. It should
be noted, however, that the model assumes an inpatient occupancy rate of 100%. In practice, it
may be necessary to add ont0 the bed days recommended by the model a stock of beds to
accommodate fluctuations in census due to emergency patients and random variation in length of
stay.

The significance of OR time availability on physician income can be determined by

contrasting it to inpatient bed avaitability. Because there are fewer low OR intensity cases that
surgeons can switch to in response to changes in OR tirne, physician billings are sensitive to the
arnount of operating room time available. Thus? it may be concluded that policies designed to
reduce fixed costs should focus on reductions to inpatient beds, rather than surgical time.
An additional conclusion of practical interest to Mount Sinai Hospital is the normative

nature of recomrnendations for pradice changes made by the model, when mn iteratively. In
instances where case mix alone does not satisfy the economic desires of physicians, the iterative
model adopts a series of recommendations for changes to the practices of dissatisfied physicians.
M e r a number of iterations, a solution to the problem c m be found in which al1 physicians can be
satisfied via case mix andlor practice changes. Tearns for which the iterative model recommends
large improvements in physician practice may be thought of as Iess efficient users of hospital
resources. Iterative model recomrnendations are, therefore, normative suggestions for equitable
resource allocations, or benchmarks, against which the efficiency of care providers cm be
compared. Of particular interest to the administration of Mount Sinai Hospital is the conclusion
that the efficiency of the Oncology team is significantly lower than that of the remaining clinical
teams comprising the Surgical and Perioperative Planning Council. (Efficiency in this context,
however, refers to the ability of physicians/clinical teams to convert hospital resources into
billings, rather than the efficiency with which resources are used to provide patient outcornes.)

This conclusion suggests that concerted effortsshould be made to increase the clinical efficiency
of this team or to increase physician billing rates for Oncology services. Sub-specialists with
clinical expertise in Oncology rnay wish to use the results of the model when negotiating fee-forservice arrangements with the province.

9.3 Mode1 Conclusions

The resource allocation rnodel clearly demonstrates that case mix management can be an
effective technique for acute care hospitals to reduce costs, while p r e s e ~ n ginstitutional income
and physician biliings. The model is particularly effective when the total reduction in budget is
small, or can be combined with increased clinicai efnciency. At runs completed under the
assumption of an 18% budget reduction, a total savings of % 1-9 million was identified for the
Surgical and Perioperative Planning Council at Mount Sinai Hospital through case mix
adjustrnents. These savings, when combined with improvements in clinical efficiency enable the
income requirements of al1 physicians to be satisfied, while ensuring that al1 cases are within *60%

of target values.
When cornpared to a benchmark of across the board cuts in either volume or practice
parameters, the resource allocation model was observed to produce superior results. The
resource allocation model provides both the institution and physicians with better income than
does an across the board cut, when clinical efficiency cannot be improved. When clinicai
efficiency c m be improved, a mixed strategy involving changes in both volume and pnce (if one
exists) cm always be identified by the model. The results of the mode1 show that the clinical
efficiencies necessary to satisfy institutional income and physician billing desires are less severe
under a mixed policy of volume and cost change than under a pure cost reduction policy.
In many ways, the resource allocation model is a significant improvement on existing case

mix models listed in the literature. Uniike the uni-dimensional models of Dowling (1976) and
Feldstein (1 967), the resource alIocation model addresses the duai nature of decision making in
hospitals, by modelling the economic desires of both physicians and the institution. The resource
allocation model is therefore able to make better decisions without undue reliance on, and

278

sensitivity to, ad hoc case bounds. The resource allocation model i s furthemore, a more
rigorous and complete application of multi-cntena decision making to the problem of case mix
than is evident in the works of Rifai and Pecenka (1989) or Robbins and Tuntiwongpiboon
(1989); the resource allocation model described here is also more appropriate to an acute care

setting than that described by Panitz (1988).

In practical terms. the resource allocation model provides proof of concept. Case rnix
management, either aione, or Ui conjunction with improvements to clinical efficiency, can be used
by acute care hospitals facing reductions in extemal fnding. The model's rapid run times and

implementation as a desktop based PC application, using a simple, easy to understand Windows

front end, indicates that case rnix management tools can be built, designed and delivered to
hospitals for the purposes of strategic planning. We rnay therefore conclude that this model is of
theoretical interest as a model of decision making behaviour in hospitals, and of significant
practical value,
9.4 Future Research

Case mix management is an under reported area in the literature. Because of the ethical
implications involved in suggesting case mix changes, researchers have been reluctant to apply
this industrial technique to a health care setting. Nevertheiess, it is a feasible and practical
strategy for deaiing with reductions in extemal funding. However, the ethical questions created
by this approach must be addressed. In Chapter 7, it was noted that reductions in the volume of

services delivered does not necessarily imply that individuals are denied service. It does,
however, suggest that individuals may have to wait, or may have to seek seMces at institutions
other than Mount Sinai. Because the hospital revenues used in this model are based on average
costs for dl institutions in Canada, there are, by definition, institutions able to produce at a profit

279

the seMces that Mount Sinai wishes to decrease. There is, therefore, an obvious economic
rationale for these profitable, but unknown, institutions to increase their volume by taking on the
cases Mount Sinai wishes to decrease. Within the confines of the Canadian system, al1 seMces
can be efficiently and effectively delivered by institutions through case mix adjustment alone.
However, the institutions that wish to increase their volume may not be conveniently CO-located
to institutions wishing to shed cases. That an institution in Vancouver has an econornic desire to
increase the volume of a particular procedure is uniikely to be of much comfort to a patient in
Toronto, who needs the procedure but is unable to access it because local institutions cannot
deliver it at a cost effective pnce.
To alleviate the ethical problems related to under and over supply of services, case mix
planning models should be applied at a regional level. Case mr planning, while feasible at the
hospital level, rnight be better addreued by considering a closed system of several hospitals in
which demand for seMces must be entirely satisfied by institutions within a given region. We
therefore conclude that regional models for case mix planning, encompassing al1 institutions
within a given geographic region are necessary. However, it must be noted that regional case mix
planning requires a more formalized and coordinated referral network than is currently in place.

Therefore, hrther research, of a technical and clinical nature, into systems to support timely,
equitable, and cost-efficient referral patterns should also be encouraged.
On a technical level, two major lines of research are suggested by the resource allocation
model. In Chapter 8, it was noted that a strategy for resource allocation based solely on case mix
adjustment, whiie feasible, will not satisfy the economic desires of al1 physicians. It was therefore
determined that a mixed strategy based on simultaneous case mix and product cost adjustment is
necessary. The iterative algorithrn, defined to resolve this problem using the model's ability to

280

recornmend both cost and case mix changes, is able to idente a solution to the resource
allocation problem (if one exists) within any defined bounds on case mix deviation. However, the
model makes recomrnendations on the basis of broad assumptions regarding the acceptability of
practice change to key decision makers. Decision maker preferences for cost/volume tradeoffs

may, however, be more complex than has been assumed in the iterative algorithm. An area for
frther research, therefore, is the relationship between cost and volume tradeoffs for the decision
making groups involved in the resource allocation problem. A large volume of literature in the
area of multi-criteria decision making (MCDM), applicable to this problem, exists. Future
research could focus on adapting, or extending, existing MCDM techniques to the issue of
volume and cost trade-off
Another line of research suggested by decision makers, when reviewing the mode1 and its
recomrnendations, is the area of operationaiizing model recornrnendations. Although the model
assumes a conservative view of k e d and variable product costs, it is not entirely clear ifhodei
recommendations for case mix change translate completely into savings at an operational level.
Product costs used in the model are based on information maintained by Mount Sinai in its clinical
costing database. However the standards used to identiq fixed, variable, direct and indirect
expenses in this database have been the subject of concern at the study site and have been revised
substantially over time. There are, as might be expected in a system as large and as new as Mount
Sinai's case costing system, some questions of data accuracy and quality related to both product
definitions and the processes used to maintain and update the database. An area of rich research
potential therefore is the detailed study of institutional costs in an acute care hospital over a wide
range of outputs. An interesting research question posed by this dissertation is whether the cost
savings identified through the model's manipulation of case costing information c m be validated

28 1

against actual practice. This question, because it deals with input data quality, rather than model
results, is essential, but beyond the scope of this project.
Areas for further investigation also include a more detailed study of the relationship
between physician billings and hospital billings/revenues. In particular, a more detailed study of
the factors innuencing billings for outpatient procedures, should be undertaken.
Finally, an operational application of the model in an ongoing decision making process is
indicated. This dissertation has proven the concept of case rnix management to be a feasible
technique for acute care hospitals to accommodate reductions in funding, while p r e s e ~ n g
institutional income and physician billings. A logical extension for this research is, therefore, to
move the concepts described in the model into practice. A full scale application of the model to
the problern of case rnix management for an acute care hospitai requires redevelopment of the
user front end and database engine for the model. A faster, more powerfil database engine is
particularly necessary before the model can be used in practice; the vast majonty of the time
required to solve an instance of the model is due to the database queries required to build, format
and write the model. In addition, access to current, online costing information is an absolute

necessity for a model of this type. The method of off-line batch processing used in this study to
obtain the data required by the model, resulted in a dataset becorning available eighteen months
after the close of the defined fiscal periods. The lag between data collection and its reiease for
analysis greatly Iessens the ability of the model to be an active management tool. Thus, a full
scaie application of the resource allocation model requires unfettered, odine access to detailed
management data.

9.5 Summary

In this dissertation, we have developed a rnodel of decision maker behaviour in an acute

care hospital, using goal programming to represent the differing and sometimes conflicting desires
of hospitals and their associated medical staff. This model, a significant improvement over case

mix models defined in the literature, provides a more detailed and accurate representation of
decision maker motivation. A number of practical additions were included to extend the model
beyond the theoretical domain, into the practice of decision making in the Surgical and
Penoperative Planning Council at Mount Sinai Hospital in Toronto. The model and its associated
database were critically analyzed using a four stage approach in which theoretical, data, logicai
and predictive validity were tested and found to be acceptable. The model was then applied to the

practical decision making problern facing Mount Sinai Hospital. Three sets of scenarios were run
in which institutional funding was reduced cumulatively by 5, 1 1 and 18%, to simulate cuts

imposed by the provincial government. Runs of the model indicated that a solution to the
resource allocation problem based strictly on case rnix adjustments could be identified at the 5 and
1 1% budget reduction levels. At the 18% budget reduction level, a rnixed strategy involving

conjoint volume and cost changes was found to be necessary. Using the model's ability to set
both product costs and case mix, an extension to the model was developed to automatically
identi& normative benchmarks for cost and volume. It was therefore concluded that case mix
management is a valid, feasible tool for acute care hospitals to reduce costs while p r e s e ~ n g
income, maintaining physician billings, and protecting access to care.

10.0 References
Baker, K.R.(1992). Elements of Se uencine and Schedulinq. Hanover, NH, Amos Tuck
School of Business Administration, Dartmouth CoUege.
Baligh, H. H. and D.I. Laughhunn (1969). An Econornic and Linear Model of the Hospital.
Health Services Research 4(4): 293-303.
Blake, Carter, O'Brien-Pallas and McGiiiis-Hall(1995). A Surgical Process Management Tool.
In Proceedinns of the 8th World Conmess on Medicd Idonnatics Edited by Robert Greenes.
Vancouver BC: International MedicaI Monmatics Association.
Blake, I.T. (1994). A Concentual Framework for SurPical Suite Management Working Paper,
Department of Industrial Engineering, University of Toronto.
Brown, D.L. and M. Hayward (1994). Key Factors in Improving Surgical Resources. Healthcare
Management Fomm 7(3): 20-26.
Butler, 1. ( 1995). Management Science/Operations Research Projects in Health Care: The
Administrator's Perspective. Health Care Manaeement Review 20(1): 19-25.
Calabresi, G. and P. Bobbit (1978). Traeic Choices. New York, Norton.

Chae, Y.M., J.D. Suver, et al. (1985). Goal Programming as a Capital Inveament Tool for
Teaching Hospitals. Health Care Mananement Review lO(1): 27-3 5.
Charnes, A., W.W. Cooper, et al. (1968). A Goal Programming Model for Media Planning.
Manaeement Science 14(8): BQ3 -B43O.
Coutts, J. (1996). Ontario Doctors Want Probe of Secret Data Use. The Toronto Globe and
Mail, December 9, 1996.
Coyte, P., Bronskill, S. et al. (1 995). An Economic Evaluation of Two Treatments for
Femoral Shaft Fractures. Working Paper, Department of Cornmunity Heaith, University of
Toronto.
Dowling, W.A. (1976). Hospital Production. Lexington, Mq Heath and Company.
Dmmmond, M.F., Stoddart, G.L. and G.W. Torrance (1987). Methods for the Economi
Evaluation of Health Care P r o m s . Odord, Oflord University Press.

Evans, R.(1 984). Strained Mercy: The Economiwf Canadian Health Care. Toronto,
Butterworths.
Eves, E. (1 995). 1995 Fiscal and Economic Statement. Toronto, Government of Ontario
Publications.

Farley, P.J. (1986). Theories of the Pnce and Quantity of Physician SeMces. Journal of Health
Economics 544): 3 15-333.
Feldstein, M.S. (1967). Economic Analvsis for Health SeMce Efficiency. Amsterdam, NorthHolland.
Finlay, P.N and J.M. Wilson (1990). Orders of Validation in Mathematical Modelling. Journal
of the Ooerational Research Society 41(2): 103-109.
Fleck, L.M. (1994). Just Caring: Oregon, Health Care Rationing, and Informed Democratic
Deliberation. The Journal of Medicine and Philosophv 19(5): 367-38 8.
Garrison, R.H., Noreen, E.W., Chesley, G.R., Carroll, R.F. (1996). Mananenal Accountinn:
Conceots for Plannine. Control and Decision Making. Toronto, Irwin.
Gass, S. and B.W. Thompson (1980). Guidelines for Mode1 Evaluation: An Abridged Version
of the US General Accounting Office Exposure D r a . Q erations Research 28(2):43 1-439.
Gass, S. (1983). Decision Making Models: Validation, Assessment and Related Issues for
Policy Analysis. Operations Research 31(4):603 -63 1.
Galbraith J.K. (1985). The New Industrial State. Boston, Houghton Mifflin Company
Gloubeman, S. and H. Mintzberg (1994). Manamne the Care of Health and the Cure of Disease.
Working Paper, Department of Comrnunity HealthJniversity of Toronto.
Hancock, W.M and M. Isken (1992). Patient Scheduling Methodologies. Journal of the
Society for Health Svstems 3(4):83-94.
Harris, J.E. (1977). The Internai Organization of Hospitals: Some Economic Implications. Bell
Journal of Economic~8(4): 467-482.
Hay, J. and M.J. Leahy (1982). Physician-Induced Demand: An Empincal Analysis of the
Consumer Information Gap. Journal of Health Economics 1(1):23 1-244.
Health Canada (1994). National Health Exgenditures in Canada: 1975-1993. Ottawa, Policy and
ConsuItation Branch.
Health SeMces Restmcturing Commission (1 997). Metropditan Toronto Health Services
Restnicturiner R e ~ o n .Toronto, ON, Health SeMces Restmcturing Commission.
Hughes, W.L. and S.Y. Soliman (1985). Short-Term Case Mix Management with Linear
Programrning. Hos itds and Health Service Administration 30(1): 52-60.
Ignizio, J.P. (1 976). Goal Promamming and Extensions. Lexington,

Heath and Company.

Ignizio, J.P.and T.M.Cavalier (1994). Linear Proeramrninq. New York, Prentice-Hall.

Jackman, H.W. (1973). Financing Public Hospitais in Ontario: A Case Study in Rationing of
Capital Budgets. Mangeement Science 29(4): 645-65 5 .
Law, A.M. and W.D. Kelton (199 1). Simulation Modeline and Analysis. New York:
McGraw-Hill.
Lave, J., P. Jacobs, et al. (1992). Transitional Funding: Changing Ontario's Global Budgeting
System. Health Care Financin Review 13(3): 77-84.
Lee, S.M. (1973). An Aggregative Resource Allocation Model for Hospital Administration.
Socio-Economic Plannine Science 7: 38 1-3 95.
Masud, AS. and C.L. Hwang (198 1). Interactive Sequential Goal Prograrnming. Journal of the
Qperational Research Society 32(5): 39 1-400.
McGlone, T.A. and R-J. Calatone (1992). A Goai Prograrnming Model for Effective Segment
Determination: A Comment and Application. Decision Sciences 23(5): 1231- 1239.
Morey, R.C.and D.A. Dittman (1984). Hospital Profit Planning Under Medicare Reimbursement.
Management Science 32(2): 250-269.
, Peereboom, E., Murray, M., Ho, R., Beed, J., and 1. Young
O'Brien-Pallas, L.L., I ~ n eD.,
(1992). Factors that Influence Vanability in Nursing Workload at the Hospital for Sick
Children: Final Report. Toronto, The Hospital for Sick Children Foundation.
Nelson, C.A. and J.R.Wolch (1985). Intrarnetropolitan Planning for Community-Based
Residential Care: A Goals Programming Approach. Socio-Economic Planning 19(3): 205-2 12.
Mitha, A. and K. Reid (1996). Tum Around in the Operatine Room at Mount Sinai Hosoital.
Bachelor's Thesis, Department of Mechanical and Industrial Engineering, University of
Toronto.
Mount Sinai Hospital (1994, Winter). Celebrating 70 Years of Health Care: Mount Sinai
Hospital Then and Now.
p. 2-3.

Panitz, E.(1988). The SeMces Mix Decision in Not-For-Profit Organizations: A Math

Prograrnming Approach to Comrnunity Mental Health Service M x Selection. Doctoral
Dissertation, University of Kentucky.
Rice, T. (1983). The Impact of Changing Medicare Reimbursement Rates on PhysicianInduced Demand. Medical Carc 2 l(8): 803-81 5.
Rifai, A.K. and J.O.Pecenka (1989). An Application of Goal Programming in Healthcare
Planning. International Journal of Production Manggement lO(3): 28-37.

Robbins, W.A. and N. Tuntiwongpiboon (1989). Lmear Prograrnming is a Usefbl Tool in

Case-Mix Management. Healthcare Financial Management 43(6): 1 14- 1 16.
Salkin, H.M. and K. Mathur (1989). Foundations of Intener Promamminq. New York,
North-HoIIand.
Sandier, S. (1989). Health SeMces Utihtion and Physician Income Trends. OECD Social
SeMces#7: Health Care Systems in Transition - The Search for Efficiencv. Paris, OECD Press.
Saunders, C. ( 1985). Model Validation: The Miuing Process. Journal of Systems
Management 36(3): 26-29.
Schellenberger, R.E. (1974). Cnteria for Assessing Model Validity for Managerial Purposes.
Decision Science. S(5): 644-653.
Schniedejans, M.J. (1995). Goal Prosammine: Methodology and Application%Boston, Kluwer
Acadernic.
Schwartz, M. and M.L. Lenard (1994). Prospective Payrnent Systems Through Equilibrium
Prices. Management Science 40(6): 774-787.
Silver, E.A. and R. Peterson (1985). Decision %stem for Inventory Management and
Production Planning. New York, Wiley and Sons.
Simon, H.A. (1960). The New Science of Management Decision. New York, Harper & Row
Spencer, G. and A. Topp (1993). The History of Ontario's Hospital F u n d i n e s t e m . Toronto,
Joint Policy and Planning Cornmittee.

Stewart, J. (197 1). Surgical Specialties Affect Scheduling. Journal of the Amencan Hospital
Association 45(1): 132-136.
Tengs, T., Meyer, G., Seigal, J., Pliskin, J., Graham, I., and M. Weinstein (1996). Oregon's
Medicaid Ranking and Cost-Effectiveness: 1s There Any Relationship? Medical Decision
Making l6(2):99- 107.

Turing, A.M. ( 1950). Computing Machinery and Intelligence. Mind 59(9):43 3-460.
Ubel, P.A., Loewenstein, G., Scanlon, D., and M. Karnlet (1996). Individual Utilities are
Inconsistent with Rationing Choices: A Partial Explanation of Why Oregon's CostEffectiveness List Failed. Medical Decision Making 16(2): 108-116.
Wagner, H.M. (1 975). Pnnciples of Operations Research. Englewood ClEs, NJ, Prentice-Hall.
Winston, W. (1 99 1). Qperations Research: Aoolications and Aleorithms. Boston, PWS-Kent
Publishing Company.

Appendix A: Example Volume Model

\\****t**f***t~*******tt*t******t**********************\\

\\
\\
\\
\\
\\
\\

Volume Model Version 5.0: This file contains an

LP listing of the example problem described in
Chapter 5. The mode1 is in standard CPLEX LP
format. Super & sub scripts have not been
retained.

\\
\\
\\
\\
\\
\\

\\****~***tt*t****t***t*****tt*************************\\

MINIMIZE

"*
+
+
+

+
+

Preemptive Goal # 1 - Economic Goals * * *

lOOOOOOdFn
OdEn
10 OOdBAn
1 O O OdBBn
1 O O OdBCn
1 O O OdBDn

* * * Preemptive Goal

+
+

+
+
+

+
#2

OdFp
1OOOOOOdEp
lOOOdl3Ap
lOOOdBBp
1 O O OdBCp
10OOdBDp 1

- Minimum Resources

* * * Preemptive Goal # 2 - Case Mix Goals * * *

\\

* * * Constraint 1 :
* * * Hospital must generate a "sufficientn profit

\\
\\
Beven

\\
\\

***
***

\\
\\

***

\\

* * * Constraint 2:
* * * Costs cannot exceed global budget

Global :
+

9411.000xA-185
+ 12006.6OOxB-251
+ 11229.000xC-251

\\

5862.4OOxA-350
+ 7618.800xB-253
+ 9314.200xC-253

* * + Constraint 3: Doctor

A ' s preferred billing

PBillA:
+ 10746.000xA-185

\\
\\

2225.000xA-350

ldBAp + ldBAn = 7 3 8 5 0 . 4

* * * Constraint 3a: Upper bounds on cases based

* * * on maximum preferred billing, Doctor A

***
***

\\
\\

UBillA:
+ 10746.000xA_185 + 2225.000xA-350

\\
\\

<= 1 2 9 2 3 8 . 2

* + * Constraint 3b: Lower bounds on cases based

* * * on minimum preferred billing, Doctor A

***
t

\\
\\

LBillA:
+ 10746.00Ox.A-185

\\

2 2 2 5 . 0 0 0 x A _ 3 5 0 >= 1 8 4 6 2 . 6

***

* * * Constraints 3-3c: Doctor B

\\

PBillB:
+

\\

6051.000xB-251

* * * Constraints

5116.000xB-253

6051.000xC-251

5116.000xC-253

ldBBp= 5 0 3 5 4 1 . 6

*** \\

- 3c: Doctor C

PBi11C:
+

+ ldBBn

ldBCp + ldBCn = 6 8 9 9 6 1 . 6

* * * Constraints 3 & 3b: Doctor D

PBillD:
+
8420.000xD-125 + 2225.000xD-350 + ldBDn

\\

\\
\\

8.900xA-350 + dLBeds-An

5862.440x.A-350 + dLDir-An

***

Constraint 46

+ dLBeds-Bn

***
=**

\\
\\

dLBeds-Bp = 345.66

*** \\
+ dLDir-Bn - dLDir-Bp = 207238.230

*** \\

Doctor B

Doctor C

= 37303.776

d L O R B p = 5295

+ 96.41xB-251 + 2613.695~8-253 + dUDir-Bn

Constraint 4a

\\
\\

*** \\

UDirB :

***

= 903.00

***
***

dLDir-Ap

\\
\\

*** \\

13.700xB-253

***
***

dLOR-Ap

Doccor B

\\
* * * Constraint 4c - Doctor B
LDirB:
+ 12006.550xB-251 + 7618.750xB-253

\\
\\

dLBeds-Ap= 65.46

* * * Constraint 4d: Upper bounds on cases based

* * * on maximum preferred contribution - D o c ~ o rA

\\
* * * Constraint 4b - Doctox B
LTimeB :
+ 255xB-251 + 215x8-253 + dLOR-Bn

\\

159.000xA-350 + dLOR-An

\\
* * * Constraint 4a
LBedsB :
+
17.500~8-251 +

\\

***
***

* * * Constraint 4c: Lower bounds on cases based

* * * on minimum preferred direct cost - Doctor A

LDirA:
+ 9411.000x.A-185 +

\\
\\

ldBDp = 25548

* * * Constraint 4b: Lower bounds on cases based

* * * on minimum preferred case time - Doctor A

LTirneA:
+
74.000xA-185

\\
\\

\\

* * * Constraint 4a: Lower bounds on cases based

* * * on minimum preferred LOS - Doctor A

LBedsA :
+
29.000xA-185

\\
\\

***

dUDir-Bp = 799900.218

*** \\

LBedsC :
+
23.100xC-251

16.100xC-253

\\
* * * Constraint 4b - Doctor C
LTimeC:
+
222.000xC-251 +
216.000xC-253
\\
* * * Constraint 4c
LDirC:
+ 11229.000xC_251 +
\\

+ dLBeds-Cn - dLBeds-Cp

= 572.04

*** \\
+ &OR-Cn

- &OR-Cp

7041.6

*** \\

Doctor C
9314 -190xC-253

+ mir-Cn

mir-Cp

= 315566 -3875

*** \\

* * * Constraint 4d - Doctor C

UDirC:
+

35l6.735xC-25l

\\
* * * Constraint 4a
LBedsD :
+
11.000xD-125 +

***

Doctor D
2.000xD-350

\\
* + * Constraint 4 b - Doctor D
LTirneD :
+
l83.OOOxD-l25 +
368.0OOxD-350
\\
* * * Constraint 4 c
LDirD :
+
3937.000xD-125 +

- dUI3ir-c~

+ dUDirCn

1690.515xC-253

***
+ &OR-Dn

4785.000xD-350

***
+ dLDir-Dn

\\
* * * Constraint 4d - Doctor D
UDirD :
+ 4545.200xD-125 335.010x.D-350 + dUDir-Dn

= 330.6

\\

- dLDir-Dp
***

= 7.8

\\

&OR-Dp

- Doctor D

1218385.044

\\

- dLE3eds-Dp

+ dLBeds-Dn

= 5233.200

\\

dUDir-Dp

= 25804.056

\\

* * * Constraint 5: Doctor A's preferred # of CMG 185

PA-185

l.OOOxA_185 + ldCA-185n

IdCA-185p

***

= 2.4

\\
* * * Constraint S a : Dr A's soft upper bound on CMG 185
SUA-1 8 5 :
+ ldCA-185p + 1dCCPA-185n - 1dCCPA-185p= .6
\\
"* Const 5b: Dr A's soft lower bound on CMG 185
SLA-185 :
+ ldCA-185n + ldCCNA-laSn - IdCCNA-I85p= .4

***

***\\

\\

* * * Const 5 - Sb Case Preferences Doctor A,CMG 350 * * * \ \

PA-350 :
+ 1xA-350
+ ldCA-350n - ldCA-350p = 21.6
SUA-350 :
+ ldCA-350p + IdCCPA-350n - 1dCCPA-350p = 2.4
SLA-350 :
+ ldCA-350n 4 IdCCNA-350n - IdCCNA-350p n = 2.6

\\

\\
* * * Const 5 - Sb Case Preferences Doctor B, CMG 251
PB-251:
+ 1xB-2 51 + ldCB-25111 - IdCB-2 S lp = 2 6 - 4
SUB-2 5 1 :
+ ldCB-251p + IdCCPB-25ln - IdCCPB-2Slp = 3.6
SLB-251:
+ ldCB-251n + IdCCNB-25111 - 1dCCNB-25lp = 3.4

"*

\\

"* Const 5 - Sb Case Prefences Doctor B, CMG 253 * * * \ \

PB-253 :
+ 1xB-253 t ldCB-253n - ldCB-253p = 67.2
SUB-2 53 :
+ ldCB-253p + ldCCPB-253n - IdCCPB-253p
= 6 -8
SLB-253 :
+ ldCB-253n + IdCCNB-253x1 - IdCCNB-253p = 7.2
\\

"* Const 5 - Sb Case Preferences Doctor C, CMG 251

PC-251:
+ 1xC-251 + IdCC-25ln - ldCC-251p = 28.8
SUC-251 :
+ idCC-251p + IdCCPC-25ln - IdCCPC-2Slp = 3 - 2
SLC-251:
+ IdCCNC-25ln - IdCCNC-25lp + ldCC-251x1 = 3.8

\\

*** \ \

\\

* * * Const
PC-253 :
+ idCC-253n SUC-2 53 :
+ IdCC-253p +
SLC-253 :
+ ldCC-253n +
\\

- 5b Case Preferences Doctor C, CMG

253

*** \\

* * * Const 5 - 5b Case Preferences Doctor D I CMG 125

PD-125 :
+ 1xD-125 - ldCD-125p + ldCD-125n = 2.4
SUD-125 :
+ ldCD-125p + IdCCPD-125n - IdCCPD-12Sp = . 6
SLD-125 :
+ ldCD-125n + IdCCND-125x1 - IdCCND-125p = . 4

*** \\

IdCC-253p + 1xC-253

= 100.8

ldCCPC-25311

- IdCCPC-253p

= 10.2

ldCCNC-253n

- ldCCNC-253p

\\

10.8

* * * C o n s t 5 - 5b Case Preferences Doctor D I CMG 350 "*

PD-3 5 0 :
+ 1xD-350 + ldCD-350n - ldCD-350p = 2.4
sm-3 5 O :
+ ldCD-350p + IdCCPD-350n - IdCCPD-350p = . 6
SLD-350 :
+ ldCD-350n + IdCCND-35On - IdCCND-350p = . 4
\\

* * * Constraints Sc: Global Case Bounds

\\

\\
\\
\\

* * * Constrainr 6:
OR Availability
***
* * * Assigned cases must not use more OR tirne than * * *
* * * is available
*"

OR :
+ 74.0OOxA-185

+ 159.000xA-350
+ 216.000xC4253

+ 222.000xC-251
c = 50000

\\
\\
\\

* * * Conetraint

Bounds

+
+

Bed Availability
use more bed days

* * * Assigned cases must not

* * * than are available

Beds :
+ 29.000x.A-185
+ 23.100xC-251
c = 3900

END

7:

+ 255.000xB-251
+ 183.000xD-125

8.900xA-350
+ 16.100xC-253

+
+

17.500xB-251
11.000xD-125

+
+

\\
\\
\\

215.0OOx.B-253
368.000xD-350

***
***
***

\\
\\
\\

13.700xB-253
2.000xD-350

Appendix B: Example Cost Model

\\*****************************************************\\
\\
\\
\\
\\
\\
\\
\\
\\
\\
\\
\\

\\
\\
\\
\\
retained.
\\
\\
Benchmark data has been used in thie mode1 to set \ \
soft bounds in instances where the benchmark
\\
is less than the user specif ied soft bound ( O - 1 )* \ \
Preferred. Benchmark data may be found in
\\
Appendix D.
\\
\\*****************************************************\\
Cost Model Version 5 - 0 : This file contains an

LP listing of the example problem described in

Chapter 5. The mode1 is in standard CPLEX LP
format. Super & sub scripts have not been

MINIMIZE

\\

Pre-emptive goal #11 & 12

\\
\\

Pre-emptive goal #21 & 22 - determine changes to case

time

\\
\\

Pre-ernptive goal #31

direct case cost

&

32

determine changes to LOS

- determine changes to

\\

\\
\\

\\
\\

\\
\\
\\

* * * \\

* * * Constraint la

LOS ehould be set so that total bed days consumed

equals total available beds.

\\
\\

Beds :
+ 26.41B-251

+ 28.81C-251

+ 67.218-253
+ 100.81C-253

+ ldBedsn

ldBedsn = 3606.96

\\

* * * Constraint lb
ttt
Ensure Doctor B gets as many or more) bed days
as were assigned by the volume model

\\
\\
\\

* * * Constraint lb
*tt
Enure Doctor C gets as many or more) bed days
a s were assigned by the volume model

\\
\\

BC:
+

\\
\\

28.81C-251

+ 100.81C-253

* * * Constraints lc
Preferred LOS

&

>=

2242.17

Id

- Doctor B , CMG 251

\\
Soft lower Dound on LOS - Doctor B, CMG 251
L1B-251:
+ IdLB-2Sln t IdLLnB-2Sln - IdLLnB-2Slp = 1.75
Preferred LOS - Doctor B , CMG 253
PlB-253 :
+ 11B-253 + ldtB-253n - ldLB-253p = 13.7

\\

\\
Soft lower bound on LOS - Doctor B, CMG 253
L1B-253 :
+ ldLB-253n + IdLLnB-253n - ldLLnB_253p= 1.50

+ * * Constraints lc & Id
Preferred LOS - Doctor C,CMG 251
PIC-251 :
+ 11'2-251 + IdLC-2Sln - l&C-251p
= 23.1

\\
\\

\\
Soft lower bound on LOS - Doctor C, CMG 251
L1C-251:
+ ldLC-251n + IdLLnC-25ln
IdLLnC-25lp = 6.80

\\

Preferred LOS
Doctor C, CMG 253
PIC-253 :
+ llC-253 + ldLC-253n - ldLC-253p = 16.1

\\
Soft lower b o u n d o n LOS - Doctor C, CMG 253
L1C-253 :
+ ldLC-253n + ldLLnC-253n- ldLLnC_253p= 3.90

\\
\\
\\

* * * Constraint 2a
*"
Case time should be set so that total OR time
but no more is used up

ORS :
+
26.4008-251
+
2.40uD-125

\\
\\
\\

67.2008-253
2.40oD-350

+ l d O R n - ldORp = 22221.67

* * * Constraint 2b
Ensure Doctor B gets as much OR time as was
assigned by the volume model

OB :
+

\\
\\

+ 67.20B-253

26.4oB-251

>=

21167.00

Ensure Doctor D gets as much OR time as was

assigned by the volume model

OD :
+

\\
\\

2.40D-125

2.40D-350

>= 555.26

* * * Constraints 2c & 2d
Preferred time Doctor B I CMG 251
IdOB-2Slp = 255

\\
Soft lower bound on case tirne, Doctor B I CMG
LOB-2 5 1 :
+ ldOB_251n+ IdONB-2Sln - ldONB-251p = 39.616
\\

Preferred time Doctor B, CMG 253

PoB-2 53 :
+ loB-253 + IdOB-2S3n - ldOB-253p = 215

\\
Soft lower bound on case tirne, Doctor B I CMG
LOB-2 53 :
+ ldOB-253n + ldONB-253n - ldONB-253p = 21.5
\\

Preferred time Doctor D, CMG 125

PoD-12 5 :
+ loD-125 + IdOD-I2Sn - ldOD-125p = 183

\\
Soft lower bound on case tirne, Doctor D I CMG
LOD-125 :
+ ldOD-125n + ldOM3-125n - ldOND-125p = 18.3
\\

\\
\\
\\

Preferred time Doctor D, CMG 350

PoD-3 5 O :

\\
Soft lower bound on case tirne, Doctor D, CMG 350
\\LOD-350:
+ ldOD-350n + ldOND-35011 - idOND-350p
= 211.063

\\

\\
\\

* * * Constraint 3a
*** \\
Set direct case cost so that profit goals are met \ \
Direct :
- 2.4OkA-185
- 21.6OkA-350
- 26.40kB-251
- 67.20m-253
9051.4291B-251
- 17167.08318-253
- 28.80kC-251
- 100.80kC-253 - 6857.1431C-251
- 28173.9131C-253
- 2.40kD-125
- 2.40kD-350 + ldDirectn - ldDirectp = -3216966.38

\\
\\
\\

* * * Constraint 3bl
*** \\
Ensure Doctor A is assigned contribution less
\\
than or equal to that assigned by the volume rnodel\\

DirA:
+
2.40kA-185 + 21.6OkA-350 >= 149215.10

\\
\\

Ensure Doctor D is assigned contribution less

\\
than or equal to that assigned by the volume model\\

* * * Constraints 3c and 3d
Preferred direct cost Doctor A, CMG 185
PkA-185 :
+ lkA-185 + 1-185x1
ldKA-185p = 9411.00

\\
\\

*mt

\\
Soft lower bound, direct cost Doctor A, CMG 185
\LKA_lSS :
\ \ + ldKA-18511 + IdKNA-185n - IdKNA-185p = 941.1
\\

Preferred direct cost Doctor A, CMG 350

PkA-3 5 0 :
+ lkA-350 + ldKA-350n

- ldKA-350p

= 5862.44

\\
Soft lower bound, direct cost Doctor A, CMG 350
LKA-3 5 O :
+ ldKA-350n + ldKNA-350n - IdKNA_3SOp= 586.244
\\

Preferred direct cost Doctor D, CMG 125

PkD-125:
+ lkD-125 + ldKD-12511 - IdKD-12Sp = 3937.00

\\
Soft lower bound, direct cost Doctor D, CMG 125
LKD-125 :
+ l m - 1 2 5 n + IdKND-12Sn - IdKND-12Sp = 393.7

\\

Preferred direct cost Doctor D, CMG 350

PkD-350 :
+ 1k.D-350 + ldKD-350n
1dK.D-350p = 4785

\\
Soft lower bound, direct cost Doctor D I CMG 350
SKD-350 :
+ ldKND-350n - ldKND-350n + ldKD-350p = 478.5

\\
\\
\\
\\
\\
\\
\\

* f * Constraint 3b4
***
Constraint 3b4 is an equity constraint tying
direct cost in the cost model to direct cost in
the volume rnodel. It differs from the equity
constraints for Doctorts A and D because the
volume model suggested reductions to Doctor B and
C 1 s case load and bed allocation.

\\
\\
\\
\\
\\
\\
\\

Preferred direct cost Doctor B, CMG 251

\\
PkB-251:
+ IkB-251 + 342.85714291B-251 + ldKB-25111 - ldKB-251p = 18006.55

\\

\\

Soft lower bound, direct cost Doctor B, CMG 251

LKB-2 5 1 :
+ ldKB-25111 + IdKNB-25ln - l m - 2 5 1 p = 1200.655

\\

\\

Preferred direct cost Doctor B I CMG 253

\\
PkB-253:
+ IkB-253 + 255.474452618-253 + 1dK.B-253n - ldKB-253p = 11118.75

\\
Soft lower bound, direct cost Doctor B I CMG 253
LKB-2 53 :
+ ldKB-253n + ldKNB-253n - IdKNB-S53p = 761.875

\\

* * * Constraint 3c & 3d
*** \\
Preferred direct cost, Doctor C, CMG 251 after
\\
factoring in savings from changes in LOS
\\
PkC-251:
+ lkC-251 + 238.101C-251 + ldKC-251n - ldKC-251p = 16729.00

\\
\\
\\

\\
\\

Soft lower bound, direct cost Doctor C, CMG 251

after factoring savings from changes to LOS
SKC-251:
+ ldKC-251x1 + ldKNC-25111 - IdKNC-25lp = 1122.9

\\
\\

Preferred direct cost, Doctor C, CMG 251 after

\\
factoring in savings rom changea in LOS
\\
PkkC-253 :
+ lkC-253 + 279.501C-253 + ldKC-253n- ldKC_253p= 13814.19

\\
\\

\\
\\

Soft lower bound, direct cost Doctor C, CMG 251

after factoring savings from changes to LOS
LKC-2 53 :
+ ldKC-253n + IdKNC-2S3n - IdKNC-2S3p = 1872.125

\\
\\

\\
\
\\

* * * C o n s t r a i n t 4a
* * * S e t bounds to ensure
+ * * budget.

TotCost :
+ 2.40kA-185
+ 26.40kB-252
+ 28.80kC-251
+
2.40kD-125

Bounds

END

w e don't exceed global

***
***
***

\\
\\
\\

+ 21.60k.A-350

+ 67.20kB-253

+ 9051.42918-251

+ 100.80kC-253 + 6857.1431C-251

2.40kD-350

<= 3216966 -38

17167.8831B-253
28173.9131C-253

Appendix C: Input Database Table Lsyout

Table: BLAKE - Patient Level Case Cost Data

Field
encounter niimber
dert mtmber
P m quanti&
var
fixed
indir
total

Description
HospitaI assigned encounter nirmher
De~artmentname
Quamity of departmentai services deiivered
Variable direct cost of deparcmental services
Fixed direct cost of dmartmental services
Iodirect cost of departmemai services
Total departmental cost of services delivered

+
encounter number

ladmit date

bdrn cat

Text

Tes

1C

Long
Long
Lona

4
4
4
4
4

Long
Lona

T&&: HRData - Patient Level HeaWr Records Data

1
Field
1
Description
Institution num - indicates idoumatient
~ o s ~ i tassigned
al
encorner mimber
Phvsician rnimber
Admission date
case mix group 1 Day procedure group
1st speciai care unit
1st SCU length of stay
Total hospital length of stay
Operathg room rnimber
Operating room time
Clinical team rnimber
Admission categow (E/U/L)
Typical / atypicai iadicator
2nd speciaI case unit
2nd SCU length of stay
3rd special care unit
3rd SCU iength of stay
4th speciat are unit
4th SCU t e n d of sw
5th special care unit
5th SCU length of stay

Size

Type

S k

Type

Lon~

Long
Loue
Text
Text
Long
Long
Long
Long
Long
Long
Text
Text
Long
Long
Long
Text

Text
Text
Text
Text

Key
4

1
1

4
4
2551
255

4
4

4
4
4
4

255
255
4
4
4

2551
255
255
255
255

Appendix D: Benchmark Data

Benchmark data used in the resource ailocation mode1 was derived fkom a variety of sources.
Length of stay data was taken from the 1994/95 Canadian Institution for Heaith Information

(CIHI)length of stay database. Direct case cost information was obtained from records
maintained by the Ontario Case Costing Project (OCCP)and was derved from OCCP member
hospitals during the 1994/95 fiscal year. Case time information used as benchmark data was

denved f?om 1994/95 Mount Sinai Health Records data by averaging operating case tirne for each

CMGDPG.
Table DA Benchmark Data

CMG
010
040
050
052
O55
056
057
075
076
077
081
084
087
O88
O89
091
125
126
127
131
184
185
187
192
250
251
252
253
254
255
256

Description
NEOPLASM OF NERVOUS SYSTEM
TRACHEOSTOMY & GASTROSTOMY PRO
ORBITAL PROCEDURES
RETINAL PROCEDURES
CENS INSERTION WITH CC (MNRH)
LENS INSERTION WITHOUT CC (MNR
OTHER OPHTHALMIC PROCEDURES (M
RADICAL LARYNGECTOMY AND GLOSS
MAJOR HEAD AND NECK PROCEDURES
LESS EXTENSIVE HEAD AND NECK P
SALJVARY GLAND PROCEDURES
MISCELLANEOUS AR, NOSE AND TH
SINUS PROCEDURES
ETHMOIDECTOMY (MNRH)
DENTAL U(TRACTION/RESTQRATION
NASAL PROCEDURES (MNRH)
MAJOR CHEST PROCEDURES
OTHER RESPIRATORY PROCEDURESW
OTHER RESPIRATORY PROCEDURES W
RESPIRATORY NEOPLASMS
MAJOR RECONSTRUCTIVE VASCULAR
AMPUTATION EXCEPT UPPER LlMB A
OTHER VASCULAR PROCEDURES
VElN LlGATlON AND STRIPPING
EXTENSIVE GASTROINTESTINAL PRO
GASTROSTOMY 8 COLOSTOMY PROCED
MAJOR ESOPHAGEAL, STOMACH & DU
MAJOR INTESTINAL AND RECTAL PR
MAJOR INTESTINAL AND RECTAL PR
LESS EXTENSIVE ESOPHAGEAL, STO
,LESS EXTENSIVE ESOPHAGEAL.STOM

Case Time
LOS
231.O0
7.80
122.00
49.70
63.96
2.70
58.21
2.00
1.70
81.00 1
1.40
68.22
63.82
1.30
510.38
20.40
275.1O
10.40103.64
2.10
118.50
2.20
48.64
1.60
83.16
1-50
66.42
1.O0
117.07
1.50
62.39
1.00
197.78
9.30
75.56
2.50
121.O6
9.30
61.O5
8.00
259.67
8.90
69.43
12.20
109.11
2.90
99.85
3.40
416.83
19.60
215.38
16.30~
224.33
13.30
193.69
12.20
195.99
8.90'
134.63
13.40
108.81
8.10,

1
A

Direct Cost
3,845.27
38,497.78
1,796.97
1.378.23
1,228.11
787.n
820.38
14,706.52
9,023.31
2,088.26
1,648.27
1,067.18
1,320.78
1,070.11
1,093.46
875.15
6,525.76
1,555.35
14,166.07
3,344.71
8,931.28
10,045.04
3,848.20
980.n
17.951.O0
12.039.39
8,216.91
7,442.07
4,426.45
7,660.03
4,644.53

3 02

CMG
257
258
265
267
268
269
270
280
289
291
292
294
295
296
300
31 1
312
314
323

Description
LESS EXTENSIVE ESOPHAGEAL, STO
W A R O T O M Y WlTH CC
ABDOMINAL LAPAROSCOPY
ANUS AND STOMAL PROCEDURES, AG
ANUS AND STOMAL PROCEDURES, AG
BILATERAL HERNlA PROCEDURES, A
BILATERAL HERNlA PROCEDURES. A
DIGESTIVE SYSTEM MAUGNANCY WI
INFLAMMATORY BOWEL DISEASE
G.I. OBSTRUCTION AGE 1 8 4 9 WIT
G.I.OBSTRUCTION AGE 18-69 WIT
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITIS, GASTROENTERITIS &
OTHER G.I.DIAGNOSES AGE > 18
MAJOR PANCREATIC PROCEDURES
MAJOR HEPATOBILIARY PROCEDURES
OTHER HEPATOBlLiARYAND PANCRE
ClRRHOSlS AND ALCOHOLIC HEPATI

327
350
352
353
354
355
356
357
361
362
366
368
370
371
372
374
377
378
379
380
391
428
432
433
437
438
477
479
504

LlVER DISEASES EXCEPT ClRRHOSl

MULTIPLE OR BILATERAL JOINT RE
HlP REPLACEMENT WlTH CC
HIP REPLACEMENT WITHOUT CC
KNEE REPLACEMENT
,REATTACHMENTPROCEDURES OR LOW
FRACTURED FEMUR PROCEDURES WIT
FRACTURED FEMUR PROCEDURESWIT
MUSCULOSKELETAL BIOPSY FOR MAL
MUSCULOSKELETAL BIOPSY WITHOUT
BACK AND NECK PROCEDURESWlTHO
MAJOR HI? AND KNEE PROCEDURES
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR LOWER WTREMITY PROCEDUR
MAJOR UPPER EXTREMITY PROCEDUR
MINOR LOWER EXTREMIW PROCEDUF?
WOUND DEBRIDEMENT AND SKIN GRA
SOFT TISSUE PROCEDURES (MNRH)
OTHER MUSCULOSKELETALPROCEDUR
OTHER LOWER EXTREMITY PROCEDUR
SECONDARY NEOPLASMS AND PATHOL
BREAST PROCEDURES EXCEPT BIOPS
SUBTOTAL MASTECTOMY & OTHER BR
SUBTOTAL MASTECTOMY 8 OTHER BR
OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES
PARATHYROID PROCEDURES
THYROID PROCEDURES
MAJOR URINARY TRACT PROCEDURES

Case Time

LOS

6.1O
8.70
1.90
2.90
2.20
3.70
2.20
7.50
6.20
3.70
5.10
2.30
3.40
5.70
5.10
18.20
12.30
8.50
7.50

,.

6.50
14.30
10.30
8.50
9.10
16.40
16.50
9.70
8.00
9.50
3.90
4.40
5.60
3.30
3.90
2.20
4.20
2.80
2.10
1.90
10.60

137.06
94.55
92.09
52.10
53.51
127.76
94.76 83.33
37.84
35.00
38.33
49.14
43.76
38.33
37.59
298.38
377.83
162.19
87.30

Direct Cost
3.010.56
5,239.23
2.064.06
1.404.22
1,140.16
2,109.23
1.457.26
3,530.31
2.526.03
1.158.19
2,019.54
941.44
1,559.54
2,541.66
2.869.59
17,071.40
7,166.32
4,270.04
3,484.02

30.57
156.94
163.52
150.92
132.10
238.59
131-54
105.71
63.23
61-34
116.09
115.04
135.00
107.07
212.67
97.70
164.84
102.00
87.41
59.70
249.50

3,619-30
8.076.43
7,663.37
6.424.94
6,400.39
8,414.65
7,281.O1
4,363.69
4,951.25
3,600.31

78.18
85.37
60.93
72.15
115.77
113.45
162.38 ,

1,494.30
1,599.91
1,295.12
2,948.70
2,305.84
1.793.56
5.423.68,

2,399.85

3,646.64
3,392.57
1.953.52
3,198.10
1,569.90
2,474.19
2,681.96
1,590.55
1,205.41
4,024.15

p
P
P

2.30
2.70
1.90
3.90
3.00
2.50
7.70

,
,

CMG
508
510
513
554
578
n g
580
703
711
725
727
728
729
731
732
734
735
736
750
756
801
803
804
805
821
900
904
DO4
005
008
028
035
D60
D61
D63

Description
MlNOR UPPER URINARY TRACT PROC
TRANSURETHRAL PROSTATECTOMYWI
OTHER TRANSURETHRALOR BIOPSY
MISCELLANEOUS MALE R E f RODUCTIV
MAJOR GYNECOLOGICAL PROCEDURES
MAJOR GYNECOLOGICAL PROCEDURES
MAJOR GYNECOLOGICAL PROCEDURES
OTHER O.R. PROCEDURES OF THE B
RETICULOENDOTHELW AND IMMUN1
MAJOR LEUKEMlA AND LYMPHOMA PR
ACUTE LEUKEMlA WITHOUT MAJOR P
~LYMPHOMAAND CHRONIC LEUKEMIA
LYMPHOMA AND CHRONIC LEUKEMIA
LYMPHOMA AND CHRONIC LEUKEMIA
LYMPHOMA AND CHRONIC LEUKEMW
ILL-DEFINED NEOPLASM WlTH OTHE
RADIATIONTHERAPY
CHEMOTHERAPY
,MULTISYSTEMIC OR UNSPECIFIED S
POST-OPERATIVEAND POST-TRAUMA
WOUND DE8RIDEMENT FOR DlAGNOSl
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DIAGNOSIS
OTHER PROCEDURES FOR DiAGNOSIS
COMPLICATIONS OF TREATMENT AGE
]EXTENSIVE UNRELATED O.R. PROCE
NON-EXTENSIVE UNRELATED O.R. P
ORBITAL AND OTHER N E PROCEDUR
LENS PROCEDURES
EXTERNAL EYE PROCEDURES
GI PROCEDURES (ENDOSCOPY)
BLADDER & URETHML PROCEDURES
DENTAL SURGERY
BIOPSY
TRANSFUSIONS

1
1
1

LOS
5.80
5.00
2.30
1.40
4.90 1
4.70
3.70
3.60
6.20
7.00
18.00
4.30 1
11-90 7.90
10.20
4 .O0
5.20
2.60
10.80
5.1 0
5.30
3.30
6.20
10.00
4.90
20.00 1
9.10

CaseTime
113.55
57.70
45.81
99.21
105.71 1
i17-63
110.38
95.74 1
40.00 1
133.79
999,999.00
62.04 1
87.58
44.75
999,999.00
85.78
999.999.00
999,999.00
70.36
37.33
81.63
74.97
153.32
153.80
56.75
197.20 1
73.50
55.96
60.85
43.49 ,
36.87
22.09 70.81 ,.
27.8 1
292.95

Direct Cost
4,456.47
2.771 -11
1,167.31
970.70
2.945.77
2.433.93
1,853.40
3,064.22
3.191-98
7,347.41
10.374.58,
2.409.66
7,385.20
4,521.92
5,460.94
3,436.58
2,354.05
1,391-24
13,815.88
2,410.91
6,790.98
2,626.51
5,229.55
7,501-26
2,953.08
11.723.26
9,110.45
215.19
252.59
287.51
128.87
197.27
251.31
318.05
302.79

Appendix E: Default Mode1 Weights

Table E.1 Penalty Weights Employed in Production Runs

Weight

1
1

w;

Description

Negative deviabons from achieving profit goais.

-

--

.. . --

w;

( Positive deviahons frorn achieving profit goals.

w;

1 Negaiha deviaons frorn remainingM i n the global budget

w*,

Positive deviaions from remaining W i n the global budget

w,

Positive and negave deviaons from physician billing targets.

Value

1 100.0000.0 1

0.0

100,000.0
1

w,

w,

1 Positive and negative deviations from case targets within soft bounds.

10,000.c)

Negative and positive deviations from case targets between soft and hard bounds.
Positive and negative deviaons from length of stay target wrthin soft bounds.

wu

1 Posive and negative deviation from length of stay targets between soft and hard bounds. 1

w,

Positive and negative deviations from case time targets within soft bounds.

w,

Positive and negative deviations from case time targets behiveen soft and hard bounds.

w,

Positive and negative deviations frorn direct cost targets m i n soft bounds.

w,

1 Positive and negave deviationsfrom direct cost targets between soft and hard bounds. 1

0.1

10,000.0
000.1

10,000.0
0.001
10,000.0
0.001
10.000.0

AppendU F: Background Information and Scenario Presented to Team Leaders

November 11,1996

A Goal Programming Approach to

Resource Aiiocation in
Acute Care Hospitals (Overview)

John Blake

Introduction: Heaith care in Canada, tike most other developed nations, is undergohg enormous
change as a result of economic restraint. While health care spending in Canada increased throughout
the 1970's and 1980's at a rate exceeding inflation, recent policies of fiscal restraint adopted by both
the federai and provincial governrnents have resulted in a net decrease to health expenditures
throughout the 1990's. On a per capita basis, heaith care fnding has decreased by 0.2% per annum
since 1992. Hospitals, the single largest component of the Canadian health care system have been
particulady hard hit by these policies. In Ontario, hospital fnding d l be reduced by 18% over the
next three years. As a result of funding changes, hospitals now face a number of difficult strategic
and operational issues. To help hospitals deai with difncult resource allocation decisions, we have
proposed the implementation of a number of techniques fiom industrial settings.
Background: Canada, Iike ail western nations, has struggled to improve health care efficiency
through market mechanisms and structural reform. In Ontario the provincial govemment has, over
the past 10 years, changed the manner in which hospitals are funded. Hktoncally, hospitals were
given a global budget. Every year the province granted hospitals funds to cover the cost of its
operations for the upcorning year. For a number of reasons, this mechanism was found to be
ineffective. Accordingly, hospitals will, in future, be funded on a rate-based or prospective payrnent
system.
The decrease in total funding and the change in funding mechanisrn has significant implications for
Ontario hospitals; the risk of institutionai loss under a rate-based system is measurably greater than
under a global budget. Econornic pressures have lead to an increased focus on efficiency and
effectiveness of care. However, in a system that has been consistently squeezed over the past five
or six years, there is little operational "fat" that can be trirnmed to meet reduced levels of funding;
changes in patient volume, rnix andlor the costs of seMce are indicated. Nonetheless, traditional
"across-the-board" reductions in volume, mix or cost may be ineffective (or counter productive!)
under a rate-based ttnding systern. Clearly, new ways of looking at the problem are necessary.
Problem Synopsis: As part of my doctoral work in the Department of Mechanicd and Industrial
Engineering at the University of Toronto, 1 have investigated techniques for identif$ng econornically
feasible mixes of cases in acute care hospitals. Under the assumption that case mix decisions must
sati* the goals and objectives of a hospital's physicians as well as those of its administrators, 1 have
adopted an algorithm (recipe), developed originally for industrial settings, to select the number and
m k of seMces for physicians/clinical tearns within any given institution. The algorithm produces a
case rnix for each physician a d o r team which, when viewed at an aggregate level, guarantees:
1)

The hospital is able to generate sufficient prospective revenue to offset the fixed
and variable costs of operation and return a given arnount of profit (if any).

2)

The total cost of running the institution are less than or equal to the upper lirnit
imposed by the provincial government.

3)

Physicians enjoy a stable income from year to year.

4)

Physicians are able to pefom, as much as possible, a "preferred" volume and mix
in this context, means that physicians are able to
of cases. (A preferred case
do the same volume and rnix of cases this year as they did last year.)

By recognizing that hospitals consist of both a corporate body and an associated medical staff, we
have created a model that identifies an economicdly feasible case m k for a hospital whiie presewing
physician income and rninimizing disturbance to practice. It is our belief that profit rnaximization
does not adequately describe the behaviour of either physicians or hospitals. Both are profit
satisificers; d e r achieving an acceptable level of income, hospitals and their medical staff tend to
pursue goals related to education, research and societal well-being rather than pure profit. By
minimizing the disturbance to physician practice, while achieving econornic break even, it is our belief
that we can produce an allocation of resources that is acceptable to al1 decision makers.
Preliminary Resulb: Preliminary results suggest that small, but targeted. changes to case mix have
a significant impact on a hospital's bottom line. These results lead us to believe it is possible for
hospitals to respond to econornic pressures by expanding efficient services rather than making broad,
across the board cuts.

Extensions: Initial response from physicians indicates that professional autonomy is a highly valued
commodity; physicians are reluctant to accept the notion of a dictated case mix. They may. however,
be willing to accept an allocation of resources (operating room time, inpatient beds, fixed costs of
operation) extrapolated from such a mix.
Through an extension to Our case m k selection modei, we have developed an algorithm that identifies
the minimum changes (variable cost, patient length of stay, operating room tirne) to a physician's
practice necessary to achieve the econornic goals suggested by the case mix model, while still
allowing him or her to deliver a preferred rnix and volume of services.
Reviewers' Role: Theory, of course, only goes so far. The acid test for any decision making
aigorithm is whether or not it actuaily makes an intelligent, rational decision. However, it is difficult
in practice to differentiate a "good decision from a "bad" decision. This is where we would like to
ask your assistance. We would like your general comments on the model and the decisions it makes.
We are looking for two kinds of feedback: general comments on the model's structure and specific
comments on the decisions it produces.
1)

Generd Comments / Face Validity: If the assumptions underlying a model are

not correct. it cannot produce decent, intelligent answers. Without worrying

about the mechanics of the model. how reasonable does this approach seem to
you? Does it capture the issues you think are important to the problem of
resource allocation? Are any of the factors the model takes into consideration

unimportant? Would you be willing to use an approach like this in a real

situation?
2)

Specaic Comments: Even ifthe assurnptions that go into a model are reasonable,
bad resuhs cm sometimes be generated. Bad results rnay arise because of errors
in the model database or an incomplete data set. Programmuig bugs and errors
in output processing can aiso produce erroneous results. More comrnonly, the
model may make a recommendation that would be dinicult to implement in
practice because of clhicai, organizatonal, or political reasons not included in the
model. If there are any results that appear to be inconsistent with your
understanding of how case mix should be set andor resources should be
allocated, please point them out and make suggestions as to what more
appropriate values might be.

What Happens From Hem: Over the next several weeks, we will be visiting CO-teamleaders from

the Surgical and Perioperative Planning Council to get their feedback on the model. General
shortcomings in the model will be corrected or explained. In instances where the model makes
rational, but unimplementable. decisions additional factors will be added to the model or the range
of numbers over which the model can make decisions will be restncted. When the model has been
validated by CO-tearnleaders, we will test its results against the actual allocation of resources iom
1WSI96.
While this research is still theoretical, there are obvious practical applications for it. As rate based
funding becomes a reality in Ontario, rnodels such as this one, will becorne increasingly important
planning tools. Thank you for your time.

November 11,1996

A Goal Programming Approach to

Resource Aiiocation in
Acute Care Hospitais (Scenano)

John Blake

Part A: Case Mzx

Imagine for a moment, that it is Apnl 1, 1995. The Surgical Planning Council at Mount Sinai
Hospital has been asked to reduce its total costs by $ 3 .O million from $39.8 million to $36.8 million
(a reduction of about 7.2%). The hospital has detemiined that f 1.0 million can be cut by reducing
operating room lime by 50,000 minutes and eliminating 2,000 bed days. Nevertheless, an additional
$2.0 million must be eliminated through changes to case mix. Rather than make across the board
cuts, the hospitai has decided it wdi look at alternative ways of redistributing resources. Using a new
cornputer program, the hospital identifies a mix of cases for each tearn. Because the mix afEects the
practice of the cliraical teams, each council must approve the model's recommendations before they
can be implemented. Imagine that you have been presented with the list of recornmended cases given
in Table 1. 1s this mix acceptable? Are there any mode1 recommendations that cannot be
implemented? If changes to the rnix are required, would you accept the rnk on the understanding that
these changes be made?
Part B.-Practzce Changes
M e r some debate, your team decides that it cannot change its case mix. Tables 2a - 2c give a list
of changes to practice, in terms of length of stay, operating room time, and direct cost that are
necessary for your team to accornplish the sarne results listed in table 1. Are the practice changes
listed in Table 2a -2c reasonable? Are there any recommendations that cannot be implemented? If
you feel your tearn would accept changes to practice rather than changes to mix, why do you think
this is the case?

Table F.1 Validation Case Mix Recommendations Presented to Team Leaders

Teun
GI Tcam

Dcscriptiou

CMC

Targtt
374.00
86.00
154.00
56.00
10.00
62.00
14.00

ModdA
354.00
86.00

ModdB

ModdC
349.80
80.44

253
254
255
256
957

GTHER DEFINED BY MODEL

GASROSTOMY& COLOSTOMY PROCED
MAJOR IIWSTDJAL AND RECTAL PR
MAJOR DJTESTNALAND RECTAL PR
LESS EXTENSCVE ESOPHAGEAL STO
LESS EXTENSIVE ESOPHAGEALSTOM
LESS EXTENSIVE ESOPHAGEAL, STO

29 1
292
294
295
296
,300

b.1.OBSTRUCTION AGE 1849 UAT

'G.I. OBSTRUCTION AGE 1849 WIT
ESOPHAGiTIS. GASTROENTERlTIS &
ESOPHAGiS. GASTROENTERITIS &
ESOPHAGITIS. GASTROENTERITIS &
OTHER G.I. DIAGNOSES AGE > 18

44.00
28.00
56.00
48.00
30.00
44.00

52.00
34.00
66.00
56.00
36.00
52.00

54.00
30.00
76.00 f
64.00
20.00
54.00

41.15
26.19
52-33
44.89
28.06
41.15

312
314
323
325
327
438

MAJOR HEPATOBILIARY PROCEDURES

OTHER HEPATOBILIARY AND PANCRE
~IRRHOSISAND ALCOHOLIC HEPATI
PANCREAS DISEASES LW EE*T MALIG
LIVER DISEASES LXCEPT CIRRHOSI
OTHER DERMATOLOGIC PROCEDURES

4.00
10.00
34.00
32.00
20.00
2.00

6.00
8.00
32.00
38.00
24.00
4.00

2.00
24.00
16.00
48.00
18-00

3.74
9.35
31.80
29.93
18.71

2.00 1

1.87

756
821
900

POST-OPERATIVE AND POST-TRAUMA

COMPLICATiONSOF TREAT'MEN AGE
L\TENSIVE UNRELATED 0 . R PROCE

6.00
6.00
6.00

8.00
8.00
4.00

12.00
8-00
2.00

5.61
5.61
5.61

D6 1

BIOPSY
TRANSFUSIONS
OTHER DEFINED BY MODEL
BACK AND NECK PROCEDURES WHO
r
MMOR IJPPER URINARY TRACT PROC
TRANSURETHRAL PROSTATECTOMY W I
OTHER TRANSURETHRAL OR BIOPSY
MlSCELLANEOUS h W E REPRODUCTlV
OTHER 0.R PROCEDURES OF THE B
I L L D E M D NEOPLASM WITH OTHE
POST-OPERATIVE AND POST-TRAUMA
mOTHERPROCEDURES FOR DiAGNOSIS
COMPLICATiONSOF TREATMENT AGE
O W R DEFINED BY MODEL
BLADDER & WETHRAL PROCEDURES
.BIOPSY

000
25 1

D63
& Sp<icidtySurgery 000

366
508
5 10

,5 I3
554
703
734
756
805
821
W O
D35
D6 1

--

- -

- -- -

--

- --

--

154.00
56.00
8-00
52.00
10.00

400.00
82.00
164.00
50.00
4.00
76.00
16.00

144.04
52.38
9.35
57.99
13.09

126.00
134.00
12.00
10.00
251.03
240.00
26.00 - - - -26.00
16.00
16.00
18.00
18.00
22.00
22.00
70.00
70.00
8.00
10.00
2.00
4.00
6.00
8.00
4.00
6.00
8.00
10.00
218.00
184.00
948.00
996.00
40.00
38.00

1 17.85

104.00
8.00
290.00
34.00

20.00
12.00
26.00
70.00
8.00
4.00
16.00
4.00
4.00
228.00
1,228.00
40.00

1 1.22
224.47
24.32
14.96
16.84
20.58
65.47
7.48
1.87
5.61
3.74
7.48
203.90
886.67
37.41
-

CMG

000
185

Ducription

OTHER DEFINED BY MODEL

&dPUATION EXCEPT UPPER LiMB A

HIP REPLACEMENT WTTH CC

HIP REPLACEMENT WITHOUT CC
354
756

~ E REPLACEMENT
E
~ C T U R E FEMUR
D
PROCEDURES WIT

368
370
3 71
372
374
577
378

W O R HIP AND KNEE PROCEDURES

MAJOR LOWER EXTREMITY PROCEDUR
MAlOR LOWER EXI'REMTY PROCEDUR
AlATOR UPPER EXTREMITY PROCEDiJR
MINOR LOWER EXlREMITY PROCEDUR
WOUND DEBRIDEMENT AM) SKIN GRA

380
437
750
756
801
803
804
805
821

OTHER LOWER EXTREMITY PROCEDUR

OTKER DERMATOLOGIC PROCEDUES
MULTISYSTEMIC O R NSPECiFiED S
POST-OPERATWE AND P O S T - T R A M
NOUND DEBRIDEMENT FOR DiAGNOSI 1
OTHER PROCEDURES FOR DIAGNOSIS
D T H E R PROCEDURES FOR DIAGNOSIS
OTHER PROCEDURES FOR DIAGNOSIS
COMPLICAnONS OF TREATMENT
AGE
-

IWO

O00
125
126
127
184
187
192

293
703
DO0

M8
O00
050
r
075
076
I

077
081
84
088
125
r
1126
I

SOFT TISSUE PROCEDURES(MNRH)

Targct
188.00
6.00

Modcl A
198.00
4.00

Modd B
160.00
8.00

Modd C
175.84

130.00
88.00
124.00

143.24
88.00
124.00

48.00

42.00

120.00
74.00
146.00
44.00

121.59
82.31
1 15.98
44.89

14-00
38.00
70.00
2-00

16.00
40.00
58.00
(0.00)
38.00
4.00
12.00

32.00

2.00
10.00

OTHER RESPIRATORY PROCEDURES W

MAJOR RECONSTRUCTIVE VASCULAR
OTHER VASCULAR PROCEDURES
VEM LIGATION AND STRPPING
ESOPHAGiS. GASTROENTERITIS&

56.00
2.00
4.00
2.00
6.00
20.00
12.00
8.00
2.00
232.00 (
52.00
34.00
8.00
8.00
40.00
16.00
1 12.00
12.00

OTHER 0.R PROCEDURES OF THE 8

OTHER DEFINED BY MODEL
GI PROCEDURES (ENDOSCOPY)
OTHER DEFtNED BY MODEL

--

~ T H DERNED
E ~
BY MODEL
OTHER DEFiNED BY MODEL
W

R CHEST PROCEDURES

~ T H ERESPIIWTORY
R
PROCEDURES W

ORBlTAL PROCEDURES

RADICAL LARYNGECTOMY AND GLOSS

MAJOR HEAD AND NECK PROCEDRES
LESS EXTENSIVE HEAD AND NECK P
SALNARY GLAND PROCEDURES
MISCELLANEOUS EAR, NOSE AND TH
SINUS PROCEDURES
E T H M O D E C T O M Y (MNRH)
M N O R CHEST PROCEDURES

~THER
RESPIRATORY PROCEDURES W

66.00
4.00
6.00
(0.00)
5.67
18.00
10.00
6.00
4.00
244.00
60.00
45.48
6.00
6.00

5.6 1

12.00
12.00
50.00
2.00
20.00
4.00 1
2.00

13.09
35.54
65.47
1.87
29.93
1.87
9.35

60.00
2.00
8.00
2.00
4.00
14.00
12.00
6.00
4.00
258.00
44.00
10.00
6.00
2.00

--

52.38
1-87
3.74
1.al

5.6 1
18.71
1 1.22
7.48
1.87
216.99
48.64
31.80
7.48
7.48

12.00
166.00
98.00
438.00

40.00
14.00
94.00
14.00
10.00
192.00
82.00
372.00

26.00
14.00
76.00
14-00
6.00
170.00
112.00
374.00

37.41
14.96
104.75
1 1.22
1 1.22
155.26
9 1.66
409.66

4.00
6.00
34.00

2.00
8.00
28.00

4.00
10.00
34.00

3.74
5.61
3 1.80

18.00
36.00

14.00
30.00

10.00
34.00

16.84
33.67

12.00
6.00
2.00
100.00
34.00 1

14.00
4.00

6.00
10.00
2.00
104.00
18.00 1

11.22
5.61
1.87
93.53
31.80

84.00
36.00 1

CMG

127
131
25 1
252
253
254
I

258
I

267
280
295
.
00
3 12
3 14
I

324
3 54
355
356
357
36 1
362
368
L

370
37 1
372
374
377
379
380
383
39 1

428
432
433
437
438
477
479
504
508
510
5 13
578
579

5 80
703
925
727

Dtscription
OTHER RESPRATORY PROCEDORES W
RESPIRATORY NEOPLASMS
~ASI'ROSTOMY & COLOSTOMY PROCED
W U O R ESOPHAGEAL. STOMACH & DU
MAJOR INTESTINAL AND RECTAL PR
MAJOR VESNAL AND RECTAL PR
W A R O T O M Y WiTH CC
ANUS AND STOMAL PROCEDURES. AG
DIGESTIVE SYSTEM MALIGNANCY W i
ESOPHAGITIS. GASTROEN'TERlTlS

~ T H E G.I.
R DLAGNOSES AGE > 18
MAJOR PAVCREATK PROCEDURES
MAJOR HEPATOBILIARY PROCEDURES
OTHER HEPATOBLIARY AND PANCRE
PANCREAllC CANCER OR OTHER MAL
KNEE REPLACEMENT
REAITACHMENT PROCEDURES OR LOW
FRACTURED FEMUR PROCEDWRES W i T
FRACTURED FEMUR PROCEDURES WIT
M U S C U B S E L E T A L BIOPSY FOR MAL
I~WSCULOSKELETAL BIOPSY WITHOUT 1
MAJOR HiP AND KNEE PROCEDURES
hlAlOR LOWER EXTREMITY PROCEDUR

MAJOR LOWER EXTREMITY PROCEDUR

MAJOR WPER EX'TREMITY PROCEDUR
MINOR LOWER EXiXEMITY PROCEDUR
WOUND DEBRIDEMENT .WD S W GRA

JOTHER

MUSCULOSKELETAL PROCEDUR
~ITHERLOWER E,YTREMITY PROCEDUR
JOINT REPLACEMENT FOR MALIGNAN
SECONDARY NEOPLASMS AND PATHOL
BREAST PROCEDURES LYCEPT BIOPS
SUBTOTAL MASTETOMY & OTHER BR
SUBTOTAL MASTECTOMY & OTHER BR
OTHER DERMATOLOGIC PROCEDURES 1
3THER DERMATOUXiIC PROCEDURES
PARATHYROID PROCEDURES
THYROiD PROCEDURES
MAJOR tlRMARY TRACT PROCEDURES
MINOR UPPER URINARY TRACT PROC
TRANPROSTATECTOMY WI
DTHER TRANSURETHRAL OR BIOPSY
MAJOR GYNECOUXiICAL PROCEDURES
MAJOR GYNECOLOGICAL PROCEDURES
MAlOR GYNECOLOGICAL PROCEDURES
OTHER 0 . R PROCEDURES OF THE B
MAJOR LEUKEMM AM) LYMPHOMA PR
ACUTE LEUKEMfA WITHOUT MAIOR P

MOM B

~ q e t ~ o d Ad

1 MOMc

14.00

10.00

34.00

13.09

68.00

80.00

50.00

63.60

24.00

20.00

28.00

22.45

6.00

8.00

2.00

5.61

58.00

63.44

72.00

54.25

44.00

44.00

46.00

41-15

6.00

8.00

2.00

5.61

4.00

6.00

6.00

3.74

24.00

20.00

48.00

2.00

4.00

2.00

--

10.00

12.00 [

12.00

22.45

1.87
9.35

14.00

18.00

6.00

13.09

18-00

14.00

24.00

16.84

12.00

10.00

6.00

11.22

34.00

40.00

26.00

3 1.80

2.00

4.00

4.00

1.87

14-00

10.00

20.00

13.09

4.00

2.00

12.00

3.74

8.00

6.00

4.00

7.48

46.00

46.00
48.00

48.00

34.00

58.00

43.02

44.89

10.00

8.00

16.00

9.35

6.00

4.00

2.00

5.6 1

8.00

6.00

8.00

10.00

8.00

16.00

14.00

10.00

20.00

13.09

14.00

10.00

18.00

13.09

48.00

40.00

44.00

44.83

2.00

4-00

4-00

1.87

36.00

42.00

26.00

33.67

7.48
9.35

20.00

22.00

18.00

18.71

66.00

76.00

162.00

6 1.73

86.00

80.00

120.00

80.44

58.00

51.63

68.00

26.00

28.00

36.00

24.00

28.00

34.00

8.00

6.00

28.00

7.48

108.00

90.00

110.00

101.01

20.00

16.00

18.00

18.7 1

54.25

1
1

24.32
22.45

2.00

4.00

2.00

1.87

8.00

10.00

10.00

7.48

52.00

60.00

30.00

48.64

22.00

18.00

26.00

20.58

28.00

22.00

26.00

26.19

48.00

40.00

64.00

44.89

6.00

4.00

2-00

5.61

22.00

20.00

34.00

20.58

6.00

6.00

8.00

5.61

Modd C

Validation LOS Recommendations

Tenm

CMC

Targtt
3.82
24.49

Dcse~iption

GI Team
GI Team

~ O O
25 1

PTHER DEFINED BY MODEL

GI Team

253

MAJOR INTESTINAL AND RECTAL PR

15.39

GI Team

P54

_MAJOR INTESTINAL AND RECTAL PR

11.33

GI Team

p55

-S

EXTENSlVE ESOPHAGEAL. STO

GI Team
GI Team
GI Team

GASROSTOMY & COLOSTOMY PROCED

b56

LESS EXTENSIVE ESOPHAGEALSTOM

257
258

LAPAROTOMY WITH CC

GI Team

265

GI Tcam

,267

GI Team
GI Team

b68
b69

LESS EXTENSIVE ESOPHAGEAL STO

ModclB

MddA

3.82

4.55

24.49

20.24

ModdC
3.79
24.32
15-28
1 1.26
16.68
8.97

15.39

15.85

11.33

10.72

16.80

16.80

10.00

9.03

9.03

8.39

8.57

8.57

8.13

8.5 1

17.64

17.64

9.00

17.51

ABDOMINAL W ~ O S C O P Y

2.27

2.27

4.58

2.26

ANUS AND S T O W PROCEDURES. AG

3.79

3.79

2.81

ANS AND STOMAL PROCEDURES. AG

BUTERAL HERNiA PROCEDURES. A

239
6.58

2.39
6.58

7.00

3.77
2.38
6.54

2.37

GI Tcun

270

BUTERAL HERMA PROCEDURES. A

2.57

2.57

3.42

2.55

GI Tram

289

iNFWCiATORY BOWEL DISEASE

8.40

8.40

9.59

8.34

GI T m
GI T-

,29 1
b92

G.I. OBSTRUCTION AGE 18-69 WIT

4.14

4.14

3.44

4.11

G.I. O B S T R U ~ O NAGE 1849WTT

4.93

4.93

8.00

4.89

GI Team

29 5

ESOPHAGITIS, GASTROENTERITIS&

5.46

5.46

6.28

5.42

GI Team

296

ESOPHAGITIS, GASTROENTERiTiS 8r

10.53

10.53

9.O0
12.67
44.00 1
2.00
11.75
6.75

10.46

2 1.45
10.40
8.64

$31 Team

300

OTHER G.I. DIAGNOSES AGE > 18

p l1

GWOR

PANCREATiC PROCEDURES

7.32

36.00

7.32

36.00

7.27
35.75

GI Team

312

MAJOR HEPATOBUARY PROCEDURES

5.00

5.00

GI Team

314

OTHER IiEPATOBiLIARY AND PANCRE

21.60

2 1.60

GI Team

b23

~CIRRHOSISAND ALCOHOLIC HEPATI

10.47

10.47

Gl T m

b27

LlVER DISEASES LYCEPT CiRRHOSI

8.70

8.70

8.56

CI Tcam
GI Team

438

OTHER DERMATOLOGIC PROCEDURES

2.00

2.00

4.00

1.99

750

bIULnSYSTEMIC OR NSPECiFTED S

16.00

16.00

3.33

15.89

GI Team

82 1

COMPLICATIONS OF TREATMENT AGE

2.33

2.33

13.00

2.3 2

GI Team

900

EXTENSIVE UNREUTED 0.R PROCE

15.33

15.33

28.00

15.23

GI T<rm

904

NON-EXTENSIVE UNRELATED 0 . R P

27.40

27.40

16.50

27.2 1

Iwo

IOTIIER DERNED BY MODEL

028

GI PROCEDURES (ENDOSCOPn
BIOPSY
TRANSFUSIONS
3.42
8.56

3.42

3.99

3.39

8.56

4.88

8.50

16.43

16.43

5.60

16.32

6.50

6.50

7.50

6.46

2.50

2.50

2.50

2.48

GI ~ c
GI Team

GI Team

W1

GI Team

D63

Gen & Spcciaity Surgery O00

OTHER DEFINED BY MODEL

Gen & Specialty Surgery 366

BACK AND NECK PROCEDURES WlTHO

Ga Br Specialty Surgay 508

Gen & Specialty Surgery 5 10

MINOR UPPER URENARY TRACT PROC

TRANSURETHRAL PROSTATECTOMY WI

Gai & Spccialy Surgery 821

M en & ~peciaity~ u g e r b
y35

~OMPLICATIONSOF TREATMENT AGE

~THER
DEFINED BY MODEL
'BLADDER & URETHRAL PROCEDURES

G m & Speciaity Surgcry D6 1

BIOPSY

Gen & SDecialtv SumeryDOO

4.97

Tuun
Musniloskeletal
Musculoskelaal
Musculoskeletal
Musnildelaal
Musculoskclaal
Musculoskeletal
Musculoskeletal
klusculoskeletaI
Musculoskeletal
hlusculosktletal
Musculoskeletal
Musculdeled
Musculaskeletal
~LIusculdeletal

CIMG

000
lia
350
352
353
354
356
b57
b62
366
368
370
371
372

2374

hIusculoskrletal
hIusniloskeletal

377
b78

Xfusculoskel~
380
Musculoskeletal
437
750
Musculoskelcral
M ~ ~ ~ ~ l o s ) c c l c t a l 756
bfusculmkelctal
1801
bf usculoskeletal
803
Musculoskelctal
804
hlusculoskeletai
805
Musculoskeletal
821
~lusculoskeletai
DO0
000
%o&c Sufgery
125
Thoracic Surgeq
126
nioncic Surgay
Thoracic Surgery
127
184
Thoncic Surgcry
187
Thoracic S u r g t m
1192
Thoracic Surgery
Thoracic Surgcry
$95
L

Thoracic S u r a m

LTho&c Surgay
r

nomJ.+ surgay
Oncology
v
Oncology
Oncolow
Oncology
Oncology

Oncoloev
Oncology
.Oncolopv
Oncology
Oncology
Oncoloav
~coloRy

703
DO0
~ 2 8
000
3 10
050
075
076
077
081
084
087
088
125
1126

acription

5-33

43.67

HIP REPLACEMENT WITHOUT CC

KNEE REPLACEMENT
F R A a D FEMUR PROCEDURES WTT
FRACTZlRED FEMUR PROCEDURES WIT
IZUSCULOSKELEXAL BIOPSY WiiOUT

BACK AND NECK PROCEDURES W i i O

W O R HIP AND KNEE PROCEDURES
M N O R LOWER EXTREMITY PROCEDUR
MAJOR LOWER EXTREMfn PROCEDUR
MAJOR UPPER EXTREMZTY PROCEDUR
~
O LOWER
R LYTREMITY PROCEDUR

WOUND DEBRIDEMENT ANI3 SKiN GRA

SOFT TISSUE PROCEDURES (MNRH)
37'HER LOWER E X R E M I Y PROCEDR
OTHER DERMATOLOGIC PROCEDURES
MULTISYSTEMICOR UNSPECIFED S
POST-OPERATIVE AND POST-TRAUMA
WOUND DEBRIDEMENT FOR DIAGNOSI
OTHER PROCEDURES FOR DiAGNOSIS
OTHER PROCEDURES FOR DIAGNOSIS
~THER
PROCEDURES FOR DIAGNOS IS
COMPLICAnONS OF TREATMENT AGE
OTHER DEFiNED BY MODEL
OTHER DEFINED BY MODEL
W O R CHEST PROCEDURES
3THER RESPIRATORY PROCEDLJRES W
OTHER RESPIRATORY PROCEDLJRES W
MAJOR RECONSTRUCTTVE VASCULAR
OTHER VMCLAR PROCEDURES
VEIN LIGATION AND STREPPING
ESOPHAGITIS, GASTROENTERmS &
OTHER 0.R PROCEDURES OF THE B
OTHER DEFMED BY MODEL

Mode1 B

ModdA
533
43.67

Targct

OTHER DEFINED BY MODEL

AMPUTATION EXCEPT UPPER LlMB A
bIULTIPLE OR BaATERAL JOINT RE
HIP REPLACEMENT WTTH CC

8.00
8.63
8.24
7.50
13.00
10.32
5.00
6.14
6.00
6.63
4.89
13.00
3.94
2.00

8.00
8.63
8.24
7.50
13.00
1032
5.00

5.29
43.37
7.94
83 7
8.18
7.45
12.91
10.25

2.67
2.00

1.77
2.98

4.25
17.00
51.00
2.43

6.46
21.85
22.5 1
5.66
13.24
10.43
2.98

4.60

6.14
6.00
6.63
4.89
13.O0
3.94
2.00
4.60

1.79
3.00
6.50
22.00
22.67
5.70
13.33
10.50
3.00

1.79
3.00
6.50
22.00
22.67
5.70
13.33
10.50
3.00

20.00

3.69
6.25
1.75
4.75
14.36
6.50
1.66
1.33

3.69
6.25
1.75
4.75
14.36
6.50
1.66
1.33
2.50

3.45
21.80
2-00
1 1.00
13.15
7.86
2.1 1
4.57
2.67

6.20

5 -72
16.00
3.00

2.50

Mode1 C ,

5.76
14.00
9.20
9.38
7.97
7.01
11.59
1 1.29
1.O0
4.56
7.67
9.00
4.24
2-00
2.701
1.50
6.00

11.67
9.67

4.97

6.10
5.96
6.59
4.85
12.91
3.91
1.99
4.57

3.67
6.2f
1.74
4.72
14.26
6.46
1.65
1.32
2.48

br PROCEDURES (ENWSCOPY)
OTHER DEFTNED BY MODEL
N E O P M M OF NERVOUS SYSTEM
ORBlTAL PROCEDURES
'RADICAL LARYNGECTOMY AND GLOSS
MAJOR HEAD AND NECK PROCEDURES
LESS E X E N S m HEAD AND NECK P
SALIVARY GLAND PROCEDURES

6.20
13.00
1 .O0
39.67
12.59
3.33
2.50

MISCELLANEOUS EAR. NOSE AND TH

SMUS PROCEDURES
ETHMOIDECTOMY (MNRH)
MAJOR CHEST PROCEDURES
~THER
RESPCRATORY PROCEDURES W

1.17
2.67
1.O0
11.08
2.82

13.00
1.00
39.67
12-59
3.33
2.50
1.17
2.67
1-00
1 1.08
2.82 1

32.80
12.00
8.20
2.53
1.67
4.60
1.O0
10.13
3.44 1

6.15
12.91
0.99
39.39
12.50
3.31
2.48
1.16
2.65
0.99
1 1 .O0
2.80

1253
b54
h58

MAJOR tWESNAL AND RECTAL PR

MAJOR INTESTINALAND RECTAL PR
LAPAROTOMY WITH CC

267
280

ANUS ;WD STOMAL PROCEDURES. AG

DIGESTNE SYSTEM MALIGNANCY WI

'MAJOR PANCREATIC PROCEDURES

Oncology

pl1
512
314
324
354
355
356
357
36 1
062
368
370
$7 1
372
374

Oncology
Onwlogy
Oncolow
Oncology
Oncology
t

Oncology
Oncology
Oncology
Oncology
Oncology
Oncotogy
Oncolog).
~colow
Oncology
Oncology
---

Oncology
Oncology
Oncology
Oncology
!

Oncology

377

Oncalogy

378

Oncology

379
380

Oncology
Oncolow
3ncology
Oncotogy

3 83
39 1
428

Onmiogy
1

OncoIogy
Oncology
Oncology
Oncology

ioncolom
Oncology
Oncology
Oncology
0nc010gy
1

OncoIogy
Onmlogy
Oncology
Oncology
Oncology
Onology

+miow

b32
1433
437
438
477
,479
504
508
510
513
578
579
580
703
71 1
725
727

h W O R HEPATOBILIARY PROCEDURES
OTHER HEPATOBILIARY AND PANCRE
PANCREATIC CANCER OR OTHER MAL
KNEE REPUCEMENT
REATACHMENT PROCEDURES OR LOW
F R A W D FEMUR PROCEDURES WIT
_FRACTLIRED FEMUR PROCEDURES WIT
MUSCULOSKELETAL BIOPSY FOR MAL
MUSCULOSKELETAL BIOPSY WITHOUT
-M;UOR HP AND KNEE PROCEDURES
MAJOR LOWER L-MITY
PROCEDLTR
MAJOR LOWER LXTREMITY PROCEDUR
MAJOR UPPER E?CREMITY PROCEDUR
MMOR LOWER LYTREMITY PROCEDUR
WOUND DEBIUDEMENT AND SKIN GRA
SOFT TISSUE PROCEDURES (MNRH)
OTHER MUSCULOSKELETAL PROCEDUR
OTHER LOWER EXREMITY PROCEDUR
JOiNT REPLACEMENT FOR MALIGNAN
'SECONDARY NEOPLASMS AND PATHOL
BREST P R O C E D W S EXCEPT BlOPS
SUBTOTAL MASTECTOMY & OTHER BR
SUBTOTAL hMTECTOMY & OTHER BR
O m E R DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES
PARATHYROID PROCEDUES

THYROi PROCEDURES
[MAJOR U R ~ A R Y
TRACT PROCEDURES
UPPER LRiNARY TRACT PROC

OR

TRANSURETHRAL PROSTATECTOMY WI
OTHER TRANSURETHRALOR BIOPSY
?L4JOR GYNECOLOGICAL PROCEDURES
h n t v 0 ~GYNECOLOGICAL PROCEDURES
MAIOR GYNECOLOGICAL PROCEDURES
OTHER O.R.PROCEDURES OF THE B
,RETICvLoENwTHELIAL AND IMMUNI

AIO OR LEUKEMIA AND LYMPHOMA PR

~ACUTELEUKEMIA W~THOUTMAJOR P

'

11.38
9.95
13.67
1.50
20.50

1138
9.95
13.67
1 -50
20.50

15.06
10.00
9.00
3.67
8.79

11.30
9.88
13.57
1.49
20.36

32.43
11.33
8.17
13.41
4.00
15.57
8.50
13.50
6.17
4.50
5.60
21.00
10.00
1 I .60
4.57
16.57
6.83
7.46
2.00
13.56
19.40
1 .O3
1.65
2.28
2.54
2.42
3.00
3.06
6.20
16.00
3.25
1.65
4.64
4.79
2.58
3.00
10.95
16.36
33.00

32.43
1 1.33
8.17
13.41
4.00
15.57
8.50
13.50
6.17
4.50
5.60
21.00
10.00
11.60
4.57
16.57
6.83
7.46
2.00
13.56
19.40
1.03
1.65
2.28
2.54
2.42
3.00
3.06
6.20
16.00
3.25
1.65
4.64
4.79
2.58
3.00
10.95
16.36
33.00

84.00
1 1.92
18.33
13.23
10.00
15.90
16.50
7.50
1 1.23
4.2 1
14.13
15.00
5.25
11.00
10.00
9.89 1
6.54
4.86
3.50
17.08
11.67
1 .O6

32.20
11.26
8.11
13.32
3.97
15.46
8.44
13.41
6.12
4.47
5.56
20.85
3.93
11.52
4.54
16.46
6.79
7.41
1.99

1.97
2.06
3.33
4.06
3.14
2.69
7.89
80.00
2.60
i .n
5.15
3.23
3.16
7.00
8.16
7.24
24.75

1.64
2.26
2.52
2.40
2.98
3 .O3
6.16
15.89
3.23
1.64
4.60
4.75
2.57
2.98
10.88
16.25

--

- -

- -

- -

13.46

19.27
1 .O2

Team
Oncology
Oncohgy
Oncology
Oacology
I

Oncology
Oncology
Oncology

,DentlEyJENT
Denb'EyeENT
Dent/Eve/ENT
DentiEyeiENT
DenUEyJENT
DentfEveENT
DentlEyeJENT
Dent/EyeJENT
DentlEyeENT
DentEvelENT
Dent/Eyr/ENT
DentEyfiNT
Dent/EyeJENT
DentEvdENT
DentXyciENT
DenUEyeENT
DentlEveENT'

CMC

728
729

73 1
732

D28
D35
D63

DcntEycENT
DenEy&NT
DentXyelENT

Modd E
3.00
19.80
16.86
8.75

1.18

088
089
09 1

ETHMOIDECOMY (MNRH)
DENTAL EXIUCTION/RESTORATiON
NASAL PROCEDURES
iOHER RESPMTORY PROCEDURES W
OTHER RESPIRATORY PROCEDURES W
ESOPHAGIS, GASTROENTERTllS &
rOTHER DERMATOLOGIC PROCEDURES

1.29
1.55
1 .O7
1.33

1.33

2.00
1 .O0
1.25

2.00
1 .O0
1.25

2.92
2.42
2.00
1.00
3.80

2.92
2.42
2.00
1.00
3.80
1 .O0

2.78
2.50
2.00
1.14
1 .O0
5.50

437

3.97
21.05
1 1-59
20.63

GI PROCEDURES (ENDOSCOPY)
BLADDER & LlRETHRAL PROCEDURES
TRANSFSIONS

S.DIUS PROCEDURES

126
127
294

ModdC

1084

077
08 1

477

DcnVEveENT

4.00
21.20
11.67
20.78

b87

057
076

-9
h03
803
804
805
W O

DenVEydENT

1 1.67
20.78

ModdA

1.23
1 .O2
1.06
1 .O0
1.50
1.83
1.30
1.58
1.13
1.15
1.1 1
1.33
5.00
5.00
1 .O0

DenUEydENT
ht/EyeENT

Targc
4.00
21.20

ORBiAL PROCEDURES
RETINAL PROCEDURES
LENS iNSERTiON WITH CC MNRHI
LENS INSERTION WITHOUT CC (MNR
~THER
OPHTHALMTCPROCEDURES (M
P OR HEAD AND NECK PROCEDURES
LESS EXTENSIVE HEAD AND NECK P
SALWARY G U N D PROCEDURES
MISCELLANEOUS EAR. NOSE AND TH

050
052
055
1056

DenUEycfENT

Dent/Evc/ENT

Dcseiiption
LYMPHOMA AND CHRONIC LEUKEMIA
LYMPHOMA AND CHRONiC LEUKEMIA
LYMPHOMA AND CHRONIC LEUKEMIA
LYMPHOMA AND CHRONiC LEUKEMIA

W4

PARATHYROID PROCEDURES
THYROiD PROCEDURES
OTHER 0 . R PROCEDURES OF THE B
OTHER PROCEDURES FOR DiAGNOSIS
OTHER PROCEDURES FOR DIAGNOSIS
OTKER PROCEDURES FOR DiAGNOSIS
~TWER
DEFINED BY MODEL
ORBITAL AND OTHER EYE PROCEDUR

DOS

&.ENS PROCEDURES

2.17
1.26

2.17
1.26

1.19
1 .O0
1.10
4.50
3.25
1 .78

1.19
1.00
1.10
4.50
3.25
1.78

1.38

1.38
t.18
1.29
1.55
1.07

1 .O0

1.46

t .50

2.16
1.25
1.18
0.99

1 .O9
4.47
3.23
1.77
1.37
1.17
I

1.28
1.53
1.06
1.32
1.99
0.99
1.24

2.90
2.40
1.99
0.99
3.77
0.99

Validation Case Time Recommendations

CMG

Team
GI Team
GI Tcam

2 51

Dtsrriptron
OTHER DEFINED BY MODEL
GASTROSrOMY & COLOSTOMY PROCED

GI Tcam
GI Team

253
254

MTES?WAL AND RECTAL PR

MNOR NlXSTiNAL AND RECTAL PR

000

GI Tcam
GI Team
GI Team
GI Team
-

QS5
256

GI Tcam

IGI T-

---

OR

268

ANUS AND STOhCV. PROCEDURES. AG

BILATERAL HERNIA PROCEDURES, A

--

GI Team
GI Team
GI Team
GI Tcarn

289

GI T m

b91

GI T m

292

3.1. OBSTRUCTION AGE 18-69 WIT

G.I. OBSTRUCON AGE 18-69 WIT

294
hg5

ESOPHAGITIS. GASTROENTERITiS
&
-~SOPHAGITIS.G A S T R O E N T E ~ Sa

hg6
300

269
$70

Gf Team

109.71

PROCEDURES.AG 1

113.61
6931
58.46

155.29
123.66

137.64
102.61

113.61
62.38 1

ModdC

57.70
210.24
192.69
210.03

101.41
--

106.26

6 1.75

76.59
119.76

64821

29.39

53.5 1
152.58
107.77
29.39

107.74
26.76

103.97
27.49

4.37

(0.00)
4.37

1.86
6.62

4.09

152.58
111.16

B U T E R A L KERNIA PROCEDUES, A
MFLAMMATORY BOWEL DISEASE

147.16
109.71

193.69
202.10

224.56
167.76

A B D O W A L LAPAROSCOPY
~ANUSAND STOMAL

156.9 1
108.31

206.02

115.81
147.16

265
h67

128.03

58.61
215.38

tESS EXENSIVE ESOPHAGEALSTO

LESS E?XENSIVE ESOPHAGEAL, STO
WAROTOMY WITH CC

108.8 1

61.69
224.78

-LE!~s EXTE5NSIVE E S O P ~ ~ A O E ~ S T O M

257
258

167.76

ModdB
57.19
230.96
214.99
202.88
103.16
.-

ModdA

Tqd

54.68
I

142.71

17.01
17.01
2 C 7 4 T 21.74 1

19.38
21311

15.91
t0.33

ESOPHAGITIS. GASTROENTERmS &

OTHER G.I.DIAGNOSES AGE > 18

27.82
21.89

27.82

22.46

26.02

21.89

24.4 1

20.47

311

MAJOR PANCREATIC PROCEDURES

109.16

109.16

154.16

102.10

GI T m
GI Team
GI Team
GI Team
GI T ~ u n
GI Team
GI Tearn
G I Team

3 14
J23

OTHER HEPATOBILIARY AND PAPU'CRE

CIRRHOSIS AND ALCOHOLIC HEPAX
PANCREAS DISESES LYCEPT MALIG
LiVER DISEASES LXCEPT CTRRHOSI

139.76
52.71

139.76

56.16

130.72

52.71

16.06

49.30

1.95
11.85

1.95
11.85

12.62
22.40

1.82
1 1.08

OTHER DERUATOLOGIC PROCEDURES

MULTiSYSTEbllC OR UNSPECIFIED S

46.16
90.12

46.16
90.12

39.16
27.1 1

43.17
84.29

POST-OPERATIVE AND POST-TRAUMA

~OMPLICATIONSOF TREATMENT AGE

63.44

0.00
63.44

8.03
9.29

1 Tcam

900

75.16

GI Team

904
WO

272.16
47.13

k;r T-

PITem
GI Team
GI Tcam

--

GI T c ~ n

325
327

438
750
756
82 1

--

LX'ENSIVE UNREWTED 0 . R PROCE

NON-EXENSIVE UNRELATED 0.R P
OTHER DEFiNED BY MODEL

272.16
47.13
30.45

MISCELLANEOUS MALE RE%DUCTIV

OTHER 0.R PROCEDURES OF TKE B
ILLDEFMED NEOPLASM W i H OTHE
POST-OPERATIVE AND POST-TRAMA

28.92

72.58
35.19

107.73

107.73 -

121.02

56.9 1

56.9 1

60.91

53.23

73.16

73.16
0.00

42.66

68.43

43.08

43.08

76.58

40.29

42.24

0.00
40.13

46.20

39.5 1

21.40

21.40

22.07

20.02

36.22

36.22

34.92

33.88

--

Gen & Specialty Surgery 703

Gai & Speciahy Surgery 734
Gen & Specialty Surgcry 756

Gen & Specialty S u r g q 805

OTHER PROCEDURES FOR DIAGNOSIS

Gai&SpccialtySwgny 821
Gai & SvccialS Sumcry DO0

COMPLICATIONS OF TREATMEh AG&

OTKER DEFINED BY MODEL

Gai & Sptcialty Surgery D35

Gen & Speciaity Sumery D6 1

BLADDER & URETHRAL PROCEDURES

BIOPSY

59.34
254.55
43.08

28.48

100.76

Tcnm

CMG

b00
Musculwkeld

1185

Muscul6Jcelctd

b50

Musculaskelaal
Musculoskeld

1354
b56

~MusculoskeIetal

b ~ 7

hiusculoskeld

b62

Musculoskcled

b66

.-

LWOR

b68
Musculwkeletal

370

Musculoskeletd

173.96

178.96

Modtl: B
59.73
87.9 1
149.06

169.92

169.92

160.94

158.93

150.13

150.13

145.13

140.41

117.0 1

117.0 1

120.07

109.44

111.89

111.89

102.69

104.65

87.54

87.54

81.88

213.16

2 13.16

62.16
100.38

H P AND KNEE P R O C E D L T R E S ~ 110.16

Targct
72.13

IIUP REPLACEMENT WITHOLJTCC

~
E REPLACEMENT
E
FRACTURED FEMUR PROCEDURES WiT
FRACTLlRED FEMUR PROCEDURES WiT
MUSCULOSKELETAL BIOPSY WiTHOUT
WITHO
BACK AND NECK PROCEDURES
. -

3 53

MuscuIoskeld

80.16

ModdA
69.43
80.16

Description
OTHER DEFiNED BY MODEL
AMPUTATION EXCEFT UPPER LMB A
MULTIPLE OR BILATEML JOINT RE

- -

Model C

67.46
74.97
167.38

199.37
. . .

- -

B W O R LOWER EXTRELLITY PROCEDUR

MAJOR LOWER EXIREMIY PROCEDUR
UPPER LYTREMITY PROCEDUR
MTNOR LOWER EXTREBUTY PROCEDUR
WOUND DEBRIDEMENT AND SKIN GRA

119.74

110.16
119.74

171.33T
128.66

103.03

120.56

120.56

106.04

112.76

195. 16

195.16

89.16

80.41

80.4 1

80.06

75.2 1

115.16

115.16

81.16

107.71
97.42

111.99
J

Musculoskeletal

37 1
572

Musculoskeletal

374

hlusculoskeletal

377

Musculoskeletal

378
379

1sor;r~

S S PROCEDURES
~ E
(MNRH)
(OTHER MUSCUUSKELETAL PROCEDUR

104.16

104.16

80.16

86.8 1

86.8 1

97.83

81.19

~OTHERLOWER EXTREMITY PROCEDR

~ T H E RDERMATOUXIC PROCEDURES
MULSYSTEMIC OR UNSPECIFIED S
POST-OPERATIVE AND POST-TIMUMA
~WOUNDDEBRIDEhENT FOR DWGNOSI 1
OTHER PROCEDURES FOR DIAGNOSIS
3THER PROCEDURES FOR DIAGNOSIS
~THER
PROCEDURES FOR DIAGNOSIS 1
OF TREAThiENT AGE
~OTHERDEFINED BY MODEL

70.98

70.98

77.96

66.39

Musculoskeld
Musculoskeflrtal
Musculoskclccal

380
437

Musculoskcld

750

Musculoskeletal

756

kIuscuIoskrled

1801

hIusculoskeletal

803

Musniloskeletal
[Mwculoakeictal

804
1805

Musculoskeled

82 1

Musculoskeletal

DO0

Thoracic Surgay
Thoracic Surgery

~ O O
125

Thoncic Surgcry

126

Thoracic Surgery
oncic Surgery

127
1184

%oracic Surgcry

187

Thoracic Sur~ery

192

oncic s u r g ~ r y

h9s
703

Thoracic Surnerv

DO0

horacic Surgety

D28

Oncoloay
bnco~ogy

000
010

Oncology

050

Oncology

075

Oncology

376

OR

...

OTHER VASCULAR PROCEDURES

L E M LIGAnON AND STRIPPiNG
~ O P H A G I T I S .G A S ~ O E N ~ E Ra~ S

73.16

73.16

80.16

68.43

65.66

60.66

61.41

0.00

49.16

OTHER 0.R PROCEDURES OF THE B

OTHER DEFINED BY MODEL
Gr PROCEDLRES (ENDOSCOPY)
OTHER DEFINED BY MODEL
'NEOPLASM OF NERVOUS SYSTEM
ORBITAL PROCEDURES
RADICAL LARYNGECTOMY AM) GLOSS
MAlOR HEAD AND NECK PROCEDURES

108.16

159.66 1

101.16

84.36

84.36

89.16

78.90

178.99

178.99
2 12.66

2 12.66

117.99
254.16

167.4 1

198.90

24.04

29.33

23.67

22.20

21.09

30.97

20.76

1 18.66

1 18.66

142.36

110.98

89.9 1

89.9 1

98.83

84.09

131.16

131.16
259.74

93.16

122.67

285.95

118.29
-

60.80

56.69

106.46
60.80

55.95

53.02

110.19

110.63

39.66

120.16

120.16

160.49

9.34

8.87

12.17

112.39
8.74

37.89

36.87

35.70

35.44

89.72

80.75

96.64

83.9 1

6.23

65.88

5.83

113.99

95.16

118.47

6.23
126.66

.
1

2 6 2 . 8 5 r 267.451

491.83

491.83

545.76

460.01

367.22

330.50

25 1.57

343.46
115.82

077

LESS EXTENSIVE HEAD AND NECK P

123.83

111.44

SALIVARY GLAND PROCEDURES

152.66

137.39

123.36
127.81

Oncalogy

1084

66.33

59.69

56.16

62.04

Oncology

087

MISCELLANEOUS EAR NOSE AND TH

SINUS PROCEDLRE!S

113.49

102.14

189.56

106.15

Oncology

088

Oncalogy

125

Oncology

126

ETHMOlDECTOMY (MNRH)
MAJOR CHEST PROCEDURES
OTHER RESPIRATORY PROCEDURES W

--

jO8 1

0.00

25.31

Onailogy

p""bgy

108.16 (

56.69

182.53

65.66

ICOMPLICA~ONS

~OTHERDEFZNEDBY MODEL
L W O R CHEST PROCEDURES
OTHER RESPIRATORY PROCEDURES W
OTHER RESPIRATORY PROCEDURES W
l h i i U 0 ~RECONSIRUCWE VASCU[

142.78

100.16

90.14

37.16

93.68

222.06

199.85

2 12.24

207.69

94.16

84.74

102.94

88.07

Tc~m
Onculom

CMG
127

oncoh3Y

13 1

Oocology

,250
b 51

Oncologv
Oowlogy

252
253
254

c010gy

bss

Onmlogy

267

Oncoloev

280

O~COIOW

295

Oncology

300

Oncoloav

3 11

(Oncolow

b12

Oncology
Oncologv

3 14
324

cology
3ncolog):

b54

Oncology

3 56

Oncrilogy
Oncolom

362

Oncology
Oncology

379

Oncolom

380

Oncology
Oncolow

428

Oncology

432

Oncology

433

lOncoIow
cology

437

355

36 1

378

39 1

- -

1438

Oncology
Oncology

479

Oncology

504

477

Oncology

508

ioncolom
Oncology

5 10

Oncology
Oncoloav

5 13
578
579

Oncology
Oncology

703

Oncologv

71 1

Oncology

725

Oncolonv

727

580

Dcscdptron
Taqet
OTHER RESPIRATORY PROCEDURES W
85.85
RESPIRATORY NEOPLASMS
14.13
EXTENSIVE GASTROINTESTLNAL PRO
345. 16
bASl'ROST0MY& COLOSTOMY PROCED
254.16
MVOR ESOPHAGEAL. STOMACH & DU
234.16
206.59
MAJOR INTESTINAL AND RECTAL PR
W O R INTESTINALAND RECTAL PR
213.96
~ A R O T O M Ywrm CC
35-77
5 1.66
ANUS AND STOMAL PROCEDURES, AG
114.18
DIGESTIVE SYSTEM MALIGNANCY WI
GASTROENTERITIS gt
~THER
G.I. DIAGNOSES AGE > 18
61.56
MAJOR PANCREATIC PROCEDURES
320.02
(MAJOR HEPATOBILLARYPROCEDURES 1
442.49 1
OTHER HEPATOBILIARY AND PANCRE
229.6 1
18.70
PANCREATIC CANCER OR OTHER MAL
~KNEEREPLACEMENT
1 106.161
(REAITACHMENTPROCEDURES OR LOW
255.16
~ c T U R E DFEMUR PROCEDURES WIT
590.16

35.77

110.16

33.46

51.66

113.16

48.32

102.76

29.09

106.79

IESOPHAGITIS.

MUSCULOSKELETAL BIOPSY FOR MAL

MUSCULOSKELETAL BIOPSY WlTHOUi

SOFT ?ISSUE PROCEDURES (MNRH)

OTHER MUSCUUISKELETAL PROCEDR
OTHER LOWER LXTREMITY PROCEDUR
SECONDARY NEOPLASMS AND PATHOL
BREAST PROCEDURES EXCEPT BIOPS
MASTECTOMY gt OTHER BR
MASTECTOMY & OTHER BR
~THER
DERMATOLOGICPROCEDURES 1
DEWTOL~GIC
PARATHYROID PROCEDURES
THYROID P R O C E D W
MAJOR URiNARY TRACT PROCEDURES
MINOR UPPER URiNARY TRACT PROC
TRANSPROSTATECTOMY Wi
OTHER TRANSURETHRAL OR BIOPSY
W O R GYNECOLOGICAL PROCEDURES
M ' O R GYNECOUXiICAL PROCEDURES
GYNECOLXXilCAt PROCEDURES
OTHER 0.R PROCEDURES OFTHE B
~TICULOENDOTHELIALAND IMMUNI
MAJOR LEUKEMIA AND LYMPHOMA PR
ACUTE LEUKEMIA WlWOUT MAJOR P

ISUBTOTAL

FUBTOTAL
b~

OR

PROCEDVRES 1

55.40

57.49

57.58

298.38

472.83

299.3 1

41 1.74 1

413.86 1

398.24

206.65

29.05

214.75

16.83

11.95

17.49

106.161

156.661

229.64

252.76

238.65

99.29

53 1.14

225.16

551.98

70.22

63.20

76.6 1

65.68

77.72

69.94

67.15

72.69

139.41

125.47

110.29

130.39

116.37

104.73

1033 7

108.84

54.16

54.16

93.16

50.66

6.82

6.13

5 1.24

6.38

59.52

53.57

56.06

55.67
90.22

96.46

86.82

91.11

106.82

96.13

95.16

99.90

84.62

76.16

75.27

79.15

82.91

- --

74-62]

107.1O

140.91

126.82

131.87

77.55

136.33

122.69

137.89

127.5 1

18 1.66

163.49

175.27

169.91

69.9 1

62.92

66.96

65.39

54.16

48.74

6 I .69

50.66

128.89

116.00

114.08

120.55

139.80

125.82

125.24

130.76

126.79

1 14.11

124.60

118.58

118.16

106.34

236.16

110.52

2.15

1.94

3.43

2.0 1

152.98

137.68

146.34

143.08

131.79

(0.00)

Team

CMC

Oncoloav

728

Oocology

729

0ac010gy

n i

Onmlonv

732

cology

p34

Oncology

735

LYMPHOMA AND CHRONIC LEUKEMIA

LYMPHOMA AND CHRONIC LEUKEMiA
~DEFINED
NEOPLASM WlTH OTHE
W I A T I O N THERAPY

Oncoloav

36

~CKEMOTHERAPY

Oncology
Oncology

756

~ST-OPERATIVE AND POST-TRAUMA

OTHER PROCEDURES FOR DIAGNOSIS
NON-LX4TNSWE UNREWTED 0.R P

Oncolow

804
904

DenEyeENT

~ O O

DenUEydENT
DrntlEvelENT

b50
b52

Dent/Eye/ENT
056
Wt/Eye/ENT
057
b76
DenUEvdENT
D~~~/EYC/ENT o n
DentEyc/ENT
08 1
DenVEvdENT
1084
r

DcntEydENT
Dent/EydENT
Dent/Eve/ENT
Dcnt/EyuENT
DenVEyeiENT
Den'EycfENT
Denl/EveENT
Dent/Eye/ENT
Dent'EyJENT
DenEye/ENT
Dent/Evc/ENT

1087

b88
b89
09 1
126
127
294
437
438

,477
kt9

91.22

(0.00)
82.10 1

DcnUEye/ENT
Dent~EvfiNT

DO0
DO4

OTHER 0.R PROCEDURES OF THE B

OTHER PROCEDURES FOR DlAGNOSIS
OTHER PROCEDURES FOR DIAGNOSIS
OTHER PROCEDURES FOR DIAGNOSIS
OTHER DEFINED BY MODEL
iORBiTAL AND OTHER EYE PROCEDUR

DentEyc/ENT
DentEvcfENT

DO5
W8

85.32

305.53

40.25

55.81

39.63

61.35
58.2 1

68.97
61.46

59.26

68.22

73.07

66.36

76.10
171.76

68.49
171.76

61.97
29.63

71.18
160.65

124.51

124.51
1 12.04

74.43
98.43

116.45
104.79

70.95

PARATHYROID P R O C E D W S
THYROiD PROCEDURES

80.19
12.82

47.98

LENS INSER'IION WITHOUT CC (MNR

BINUS PROCEDURES
EZHMOIDECTOMY
DENTAL LXRACTiON/RESTORAnON
NASAL PROCEDURES (MNRH)
OTHER RESPEWTORY PROCEDRES W
~ T H E RRESPIRATORY PROCEDURESW
IESOPHAGITIS. GMTROENTERITIS &
OTHER DERMATOLOGIC PROCEDURES
OTHER D E W T O U X I C PROCEDUES

65.60

122.66 1

42.37
61.35
63.36

MISCELLW'EOUS EAR. NOSE AND TH

30.08
171.83

293.99

326.66

~THER
DEFINED BY MODEL
ORBITAL PROCEDWRES
REii3iA.L PROCEDURES

OTHER OPHTHALMIC PROCEDURES (M

MNORHEADANDNECKPROCEDURES
LESS EXTENS~VEHEAD AND NECK P
SALNARY GLAND PROCEDURES

90.22 1

Modd C

(0.00)

703
804
805

Modd B
67.04
108.16
8.3 1

Modd A
63.13
77.16
1233

5 130

Den'Eye/ENT
DentlEveIENT
DenuEydENT
Dent/Eye/ENT

803

Tqct
70.14
85.74
13-70

Dacriptlon

LYMPHOMA AND CHRONIC LEUICEMIA

LYWHOMA AND CHROMC LEUKEMIA

1 12.04

57.38

48.62
76.37

50.25
56.96

75.97
103.94

60.90
75.97
103.94

66.47
130.87

67.15
56.59

62.39
56.59

66.16
39.93

71.05
97.22
62.80

46.26

64.26

43.27
30.17

39.26
33.26

58.93
72.57

120.76

107.10
101.36

53.72

53.72
60.90

46.26
32.26
63.0 1
f1.59
1 14.5 1
10837

32.26
63.01
77.59
1 14.51

108.37

111.24

121.76

121.76

88.76

47.84

47.84

123.86
1 19.26

123.86

59.97
86.26

52.93

113.88
44.74

50.78
56.0 1

119.26
48.24
55.96

72.26
50.56

115.85
1 1 1.54
47.49

56.87

52.38

LENS PROCEDURES

62.57

56.55

5 8.26

58.52

EXiERNAL EYE PROCEDURES

39.78

39.78

3 8.49

37.2 1

Vaiidation Direct Cost Recommendations

Tenm

CMC

Description

TW?d

GI Teyn

000

GI T e m

2 5t

S I Ttam
GI Team

253
254

GI T a n
S I Tcam

255

Gl Tcam
GI Tearn

257
b58

GI Team
GI Tcam

265
267

GI Team
GI Team

1268
b69

GI Team
Gl Tkun

270
289

GI Team
GI Team

5 91

ABWMMAL LAPAROSCOPY
+4US AND STOMAL PROCEDURES, AG
A W S AND STOMAL PROCEDURES. AG
BUTERAL HERNiA PROCEDURES. A
BILATERAL HERNlA PROCEDURES, A
ENFLAMMATORY BOWEL DISEASE
G.I. OBSTRUCTION AGE 18-69 WIT

292

G.I. OBSTRUCnON AGE 18-69 WTT

1.272. 14

CI Team
GI Team

294
b95

ESOPHAGIIS, GASTROENTERITIS &

ESOPHAGITIS-GASTROENTERiTIS&

1,015.07
1.234.83

GI Tram

'296

ESOPHAGITIS. GASTROENTEIUTIS &

2,472.53

Gl Tram

300

3THER G.I.DIAGNOSES AGE > 18

Gf Team
GI Tcam

31 1
312

MAJOR PANCREATIC PROCEDURES

MNOR HEPATOBILIARY PROCEDURES

256

OTHER D E M D BY MODEL
GASTROSTOMY & COLOSTOMY PROCED
MAlOR WESTlNAL. AND RECTAL PR
MNOR LNTEST[NAL AND RECTAt PR
LESS EXTENSIVE ESOPHAGEAL STO
LES3 EXTENSIVE ESOPHAGEALSTOM
LESS EXTENSrVE ESOPHAGEAL. STO
LAPAROTOMY WITH CC

1.176.72
9.432.18
5.6 18.96
4.686.43
7.0 15.20
3.140.90
3318.00
5,401.09
1,525.73
1,345.41
892.33
2.402.50
1.289.86
1-819.82
891.18

2.0 14.27
20.665.00
2.80 1.50
5,147.60

3 14

OTHER HEPATOBILIARY AND PANCRE

323

2.485.06
2.237.69
2.260.20

GI Tearn

325

GI Team

327

CiRRHOSIS AND ALCOHOLIC HEPATI

PANCREAS DISEASES EXCEPT MALIG
LIVER DISEASES EYCEPT CIRRHOSI

GI Team
GI Team

438
750

OTHER DERMATOLOGIC PROCEDURES

bCLTISYSTEMIC OR mSPECIFED S

83 1.00
5.724.00

GI Team
GI Tram

756

POST-OPEIWTIVE AND POST-TRAUMA

1,035.67

82 1

~OMPLICATIONS
OF TREATMENT AGE

79 1.33

GI Team
Gl Tcam

DO0
D28

OTHER DEFINED Bk' MODEL

GI PROCEDLlRES (ENDOSCOPY)

118.67
121.79

GI Tcarn
GI Team

D6 1
D63

BIOPSY
TRANSFUSIONS

341.95
285.50

GI Team
GI Team

Gen & Speciahy Surgery 000

Gen & Specialty Surgery 366

Gai & Specidty Surgery 508

Gen & Specialtysurgery 510

OTHER DEFINED BY MODEL

BACK AND NECK PROCEDURES WITHO
MNOR UPPER URINARY TRACT PROC
TRANSURETHRAL PROSTATECTOMY Wi

1.258.3 1
1.703.33
5,058.43
1.7 13.00

Gen & Spccialty Surgcry -5 13

OTHER RANSURETHRAL OR BIOPSY

Gai & Spefiafty Surgery 554

Gen & Specialty Surgay 703

Gen & SpeCialty Surgery 805

MSCELLANEOUS MALE REPRODUCTIV

OTHER 0 . R PROCEDURES OF THE B
ILLDEFPED NEOPLASM WH OTHE
WST-OPERATIVE AND POST-TRAUMA
OTHER PROCEDURES FOR DIAGNOSIS

777.00
759.00
671.00
2.455.00
3.092.00

Gai gt Specidty Surgery 82 1

COMPLICATIONS OF TREATMENT AGE

609.50

Gai & SPCC-

OTHER DERNED BY MODEL

BLADDER & URETHRAL PROCEDURES
BIOPSY

188.01
185.81
279-35

Gen & Specialty Surgery 734

GUI & SptCialty Surgery 756

SWW Do0

Gai Bc Spcciahy Surgay D35

Gai & Speciahy Surgay D6 1

820.67

Mwlaskelelal

185

Musniloskeldal
Musculoskeletal

350
352

Musculoskeletal

353

Musniloskcletai
Musculaskeletal
Musculoskctaal

b54
356
357

MuscuIosktletal

366

Musculoskeletal

368
370

Musculoskeletal
klusculoskeletal

3 71
372
374

bf usculwkeletal
Musni1oskclctal
Musculoskeld

i377

Musculoskeletal

378

Musniloskeleial

379

Musculoskele!al
Musculoskeletal

437
750

Musculoskcld
blusculoskelaal

80 1

blusculoskeletal
hIusnilaskeld

803
804

b~~loskelctal
Musniloskeldal

805
82 1

756

MusculoskeId
Thofacic Surgcry

DO0
O00

Thor;;rcic Surgery

125

Thoracic Surgay
Thoracic Surgery
Thoracic S u r n i

126
127

184

1187

- -

--

Thotacic Sugcry
boracic Surgay
1

192

Thoracic S u r e m
nioracic Surgery

295
703

Thoracic Surgery
Thoracic Surecrv

LM0
M8

Oncology

~ O O

Oncoloev

b10

Oncolo@

b50

Oncology
Oncology

b75
1076

Oncolom

bn

087

Oncology
Oncologv

08 1
084

Oncology
Oncology
Oncoloev

388
125
-

Oncology

126

Oncolow

127

AMPUTATION EXCEPT UPPER LiMB A

MUtTIPLE OR B U T E R A L JOINT RE
%
REPLACEMENT
I
I
'
WITH CC
HIP REPLACEMENT WITHOUT CC
K N E E REPLACEMENT
FRACTURED FEMUR PROCEDURES WT
JRACTURED FEMUR PROCEDURES WlT

8.13233
4.495.20
4.472.53

4.25 1.00

4.472.53

4.143.35

3,714.27

4.14335

3.105.37
3.933.27

3323.4
3240.69

2.740.88

2 6 10.95

1,045.33

1393.00

1
1

2.610.95
1393.00
1.862.71

1393.00
1.862.71

2.329.00

1.862.71

1.815.21
1.529.74

1.815.21
1.529.74

2781.00
1.198.48

1,815.21
1,529.74

4,045.00
1,109.25

4,045.00
1.109.25

4,045.00

1.743.00
1.177.40

1.743.00
1.177.40

520.00
952.60
816.50
1.414.00

1.109.25
1.743.00
1.177.40

843.76

998.94

843.76

1.088.00

1.047.00
1.766.50

1
843.76 1

'

1.047.00
1.766.50

1,047.00
1.766.50

3,622.00

3,622.00

3.525.00

3525.00
1.658.20
4.174.33

1,658.20
4,174.33
4.090.50
490.00

- -

3.525.00

1.042.86
3.074.00

1,658.20
4.174.33
4.090.50
490.00

184.78
151.95
.- - 1.028.50
1.21 1.96

3.402.58
1.041.50

8.581.60
1.122.67

3.402.58
1,041.50

2.392.00
7.146.57
7.146.57
2,011.50~ 2,011.50
936.88
936.88

8.184.00
6.68 1.69

2392.00
7,146.57

1.041.50
2.392.00
--

839.33

3,385.14
953.95

1.532.17

1.195.00
1.761.33

32.64
73.88

32.64
73.88

41.41
88.14

1,969.80
3.896.00

1.969.80

1.798.04

3.896.00

2975.71

1.969.80
3.896.00

935.00
24.230.33

1.178.00
14,355.80

935.00
24.230.33

4,458.41
1.326.67

5,574.47

4,458.41

2.555.20

1326.67
1.406.89

24.230.33
4.458.41
1.326.67
. 1.406.89

2.011.5
936.88

839.33
1.532.17

935.00

3.622.00

9.736.50

184.78
1.21 1.96

IOTHER

1290.50
3.577.00

3.872.00
4.489.50

184.78

4.090.50
490.00

1.21 1.96
3,402.58

1
MAJOR RECONSTRUCTIVE VASCULAR 1

~ T H E RVMCULAR PROCEDURES
VEM LIGATION AND STRIPPMG
ESOPHAGITIS. GASTROENTERITIS &
~ T H E O.R
R PROCEDURES OF THE 8
DEFINED BY MODEL
ki1 PROCEDURES IENDOSCOPYI
OTHER DEFINED BY MODEL
NEOPLASM OF NERVOUS SYSTEM
ORBlTAL PROCEDURES
RADICAL LARYNGECTOMY AND GLOSS
MAJOR HEAD AND NECK PROCEDURES
LESS EXTENSIVE HEAD AND NECK P
SALIVARY GLAND PROCEDURES
MISCELLANEOUS EAR, NOSE AND TH
SINUS PROCEDURES
ETHMOIDECTOMY (MNRH)
MAJOR CHEST PROCEDURES
OTHER RESPRATORY PROCEDURES W
-0THER RESPIRATORY PROCEDRES W

8,13233
4.495.20

3323.40
3,240.69
2.6 10.95

OTHER DERMATOUXilC PROCEDURES

MULTISYSTEhffC OR U N S P E C E D S
P O S T - O P E R A m AND POST-TRAUMA
WOLWD DEBRIDEMENT FOR DIAGNOSI

F~ER

4,873.00
4.658.60

4,14335
3323.40
3.240.69

BACK AND NECK PROCEDURES WlTHO

MAJOR HIP AND KNEE PROCEDURES
MAJOR LOWER EXREMlW PROCEDUR
MAJOR LOWER EXWhdiTY PROCEDUR
MAJOR UPPER EXTREMITY PROCEDUR
MINOR U)WER EXTREMW PROCEDUR
WOUND DEBRIDEMENT AND S U N GRA
~ O F TISSUE
T
PROCEDURES (MNRH)
~ H E MUSCUU~SKELETAL
R
PROCEDUR

OTHER PROCEDURES FOR DIAGNOSIS

FOR DIAGNOSIS
OTHER PROCEDURES
PROCEDURES FOR DIAGNOSIS
~OMPLICATIONSOF TREATMENT AGE
(OTHER
DEFINED BY MODEL
-OTHER DEFINED BY MODEL
MAJOR CHEST PROCEDURES
OTHER RESPIRATORY PROCEDURES W
IOTHER RESPiR4TORY PROCEDURES W

8.13233
4.495.20
4.472.53

839.33
1.532.17
32.64
73.88

685.50

1.406.89
685.50

1,212.76
738.67

1.187.67
843.00

1.18'7.67
843.00

2311.80
45 1.O0

1.187.67
843.00

5.140.50
1,408.41

5.140.50
1,408:41

4.512.75
1,58433 -

5.140.50
1,408.41 -

8,157.14

8,157.14

3.152.29

8,157.14

- -

685.50

Oncology

131

Oncology
Oncoloav

250

Oncology
cology

251
252
253

colow

254

Oncology

267

OncoIow

280

COIO~Y

Oncology

~ S S
356

RESPiR4TORY NEOPLASMS
m N S I V E GASTROINTESTiNAL PRO
~ASIROSTOMY & COLOSTOMY PROCED
MAlOR ESOPHAGEAL, STOMACH & DU
MAJOR NEsNAL AND RECTAL PR
b W O R INTESTMAL AND RECTAL PR
ANUS AND STOMAL PROCEDURES. AG
DIGESTiVE SYSTEM MALIGNANCY W1

IREAT~ACHMENT

- -

PROCEDURES OR MW
FRACTURED FEMUR PROCEDURES WTT

2.519.12

2519.12

2.499.32

2,519.12

8.43 1.00
9.797.08

8.43 1.00
9.797.08

6.077.00
4,764.19

6,077.00
4,764.19

16.602.33
5272.57
5.230.00
5.056.94

9.797.08
6.077.00
4.764.19

4233.10

4233.10

4.100.17

4.233.10

745.50

745.50

1,648.00

745.50

4.78 1.33

4.781.33

1359.00

4.78 1.33

6.192.14
2.664.50

Oncolow

357

FRACTURED FEMUR PROCEDURES WIT

Epi7-

(361
362

~ S C U L O S K E L E T A BIOPSY
L
FOR MAL

Oncology
Oncology

37 1
372

Oncology
Oncology

374

3.780.25

1.948.00

2.664.50

(Oncoloev
cology

17384.50
5,124.67
2,974.00
3.780.25

8.43 1.O0

1948.00

3343.41
1.433.83

6.192.14
2,664.50
3.780.25

1.948.00 1

MUSCULOSKELETAL BIOPSY WiTHOT

MAJOR HIP AND KNEE PROCEDRES

1,294.50
1.682.20

1.294.50
1.682.20

3,219.13

1.294.50
1,68220

3,600.25
3.685.80

3,600.25

1.646.75

3.600.25

3.685.80

1.727.29
4.483.29

4,200.38
3.032.70

3,685.80
1.727.29

1.727.29
4.483.29

3.104.89

Oncology

379
380

MAJOR LOWER EYTREMITY PROCEDUR

MAJOR WPER L-MITY
PROCEDUR
MINOR LOWER EXTREMITY PROCEDUR
~WOWD DEBRIDEMENT AND SIUN G R 4 1
S O ~ S~ S U PROCEDURES
E
(MNRH)
OTHER MUSCULOSKELETAL PROCEDUR
LOWER E X T R E W PROCEDUR 1

IOTHER

2.34442
2.276.04
73 1.00

2344.42
2,276.04
73 1.00 1

2,133.17
1.665.14
1.697.50

Oncology

383

SOINT REPLACEMENT FOR MALIGNAN

7.494.50

7.494.50

7,734.85

73 1.00
7.494.50

Oncology
Oncology

39 1
428

3,99 1.90
590.18

3.99 1.90
590.18

1.8 14.67
604.01

3.99 1-90
590.18

Oncology
Oncology

432
433

958.12
1,141.48

958.12
1.141.48

1.062.00
1.057.68

958.12
1.14 1.48

Oncologv

437

SECONDARY NEOPLASMS AND PATHOL

BREAST PROCEDURES EXCEPT BIOPS
SUBTOTAL MASTECTOMY & OTHER BR
SUBTOTAL MASTECTOMY & O T I E R BR
OTHER DERMATOLOGIC P R O C E D W S

1.025.54

1.025.54

1.230.28

1.02534

1,535.25
1.509.80

1,535.25

1,595.71

1.535.25

1.509.80

1,460.11

1.509.80

2.4 14.90
5.433.00

3.636.00
14.588.00

2414.90
5,433.00

1.215.75
707.58

1,215.75

1.004.60
802.27

1.215.75
707.58

1.678.9 1

1,678.9 1
1,564.50

1,560.38

1.678.9 1

1389.92

1,564.50

1308.54

1.240.34
3,579.00

1308.54
1,894.67

2.188.21
3.905.06

3,555.10
7.500.00
10,485.00

368

cologv

Oncology
Oncology

377
378

Oncoloav

725

Oncology

727

Oncology

728

PARATHYROID PROCEDURES
THYROID PROCEDURES
MAJOR iJMNARY TRACT PROCEDRES
MMOR UPPER W A R Y TRACT PROC
TRANPROSTATECTOMY MA
OTHER TRANSURETHRALOR BIOPSY
MAJOR GYNECOLOGICAL PROCEDURES
MAlOR GYNECOLOGICAL PROCEDURES
M4JOR GYNECOLOGICAL PROCEDUES
OTHER 0 . R PROCEDURES OF THE B
RETICLOENDOTHELIAL AND IMMUN1
MAJOR LEUKEMIA AND LYMPHOMA PR
ACUTE LEUKEMIA WITHOUT W O R P
LYh.lPH0M.A AND CHRONlC LEUKEMlA

729

L W H O M A AND CHRONIC LEUKEMIA

477
479
504
508
5 10

513

k
cology

578
579

Oncology

580
703

Oncology

71 1

Oncolow

2.414.90
5.433.00

1.564.50

707.58

4,483.29
2,34442
2.276.04

1308.54
1.894.67

1,894.67

3.555.10
7.500.00

3.555.10
7.500.00

10,485.00

10.485.00

8,393.50

1.580.88

1.580.88

1,763.20

1,580.88

5,63 1.60

5.63 1.60

6,382.80

5,63 1.60

~YMPHOMAAND CKRONIC LEUKEMIA

LYMPHOMA
AND CHRONIC LEUKEMIA

3.042.44

3,042.44

4383.86

5.8 11.67

5.8 I 1-67

2170.00

5.8 I 1-67

ILLDEFiNED NEOPLASM WITH OTHE

RADIATION THERAPY
CHEMOTHERAPY

2264.09

2,26409

3.883.33

2.26409

2,252.03

2.252.03

2.3 10.82

2252.03

1246.54

1 J46.54

1.027.5 1

1346.54

Oacology

73 1

0nca10a~
Oncology

nz
p34

3awlogy
Oncolonv

b 5
tn6

Oncology

756

POST-OPERATIVE AND POST-TRAUMA

Oncoloav

804

OTHER PROCEDURES
FOR DWGNOSIS
- -UNRELATED 0.R P

3.042.44

449.00

449.00

2,150.50

449.00

3344.00
6.530.00

3344.00
6.530.00

4.910.33

3.344.00

ncotogy

b04

Oncalogy
Oncology

DO0

ON-EXTENSIVE
IOTHER DEEWED BY MODEL

ORBITAL AND OTHER EYE PROCEDR

191.85
69333

191.85
69333

284.94

DO4

156.83

191.85
693.33
377.56

--

9,837.00

D8

EXERNAL EYE P R O C E D W

377.56

341.47

300.57

D28

GI PROCEDLRES (ENDOSCOPY)

149.41

134.47

154.98

149.41

O
, n~logy
Oncoloav

p35
ID63

215.79

197.27

215.79

DentXyJENT
Dent/Eyc/ENT
Denb'Eve/ENT
DentfEyetENT
DenWEy&NT
htfEvr/ENT
DaitlEyelENT
DenVEydENT
&t/EvdENT
I~entfEvdE~~

~ O O

b50

BLADDER & URETHRAL PROCEDURES

TRANSFUSIONS
3THER DEFiNED BY MODEL
ORBITAL PROCEDLRES

b52

RETINAL PROCEDWRES

055

LENS INSERTION WlTH CC 0

056
057

LENS RJSERTION -OUT

CC (MNR
~THER
OPHTHALMIC PROCEDURES CM

.Oncologv
Oncology
I

77

DniuEydENT
Dmt'EyeENT
DcnUEve/ENT
DentlEye/ENT
DentfEye/ENT
P U E Y & T
DenUEyeRNT
DenWEydENT
Dcnt/EydENT

IO8 i
1084
O87
388
089
09 1
126

1127
294

~7

43 8

ESOPHAGITlS, GASTROENTERITIS &

OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES

ID08

EX'TERNAL EYE PROCEDURES

437

DenVEyeiENT

MAJOR HEAD AND NECK PROCEDURES

LESS EXENSIVE HEAD AND NECK P
SALNARY GLAND PROCEDURES
~ISCELLANEOUS
EAR NOSE AND TH
S i N S PROCEDURES
ETHMOIDECTOMY ( h m
DENTAL LXTIUCTiONRESTORATiON
'NASALPROCEDURES
RESPUTORY
PROCEDURES W
OTHER
-- THER RESPLRATORY PROCEDURES

313.32

302.79

252.11
278.05

713.53

713.53

721.16

697.83

697.83

712.02

697.83

565.90

565.90

601.03

565.90

705.00

705.00

746.05

705.00

3 1332

713.53

793.29

793.29

753.33

793.29

950.09

950.09

846.38

950.09

679.37

679.37

729.00

67937

807.67

807.67

996.33

644.00

644.00

1.086.00

807.67
L

644.00

365.00

365.00
723.00

n8.00
539.00

365.00

723.00
1314.33

1.314.33

2.120.33

1314.33

298.12

27.51 1

237.20

723.00

Appendix G: Model Recornmendations - 5% Budget Reduction

Table G.1 lists a surnmary of volume and cost model recommendations for each clinical
team; Table G.2 provides a detailed listing of volume and cost model recommendations. The results
appearing in the two tables assume sofi bound on case deviations of *5% and hard bounds of 130%.

Soft bounds of *5% and hard bounds of 120%of current physician practice were also assumed for
the runs used to generate the results appearing in G. 1 and G.2.

Table G.1 Model Recommendation Summary - 5% Budget Reduction

.-

GI
Gen & Spec
,Msk
Thoracic

Oncology
PenEvelENT

Cases
4,077
1,823
1,194
517
3,535
3.114

Preferred
Bed Days OR Time Direct Cost
222,870 $ 4,477,864
12,684
67,142 $ 934,059
1,949
120,939 $ 2,516,790
6,786
32,800 $ 598,558
1,297
325,785 $ 6,820,616
17,684
. 2.563. 206.167s 1.562.517.

Volume ~ o d ~ecommendations
e ~
Cases
Bed Days OR Time Direct Cost
4,266
12,213 224,753 $ 4,314,788
1,754 67,095 $ 868,564
1,867
6,511 121,268 $ 2,443,628
1,226
1,167 32,294 $ 541,514
569
16,488 326,682 $ 6,335,392
3,902
2,453 , 203.609 S 1.51 1.900
3,128 .

Cost Mode1 Recommendations

Cases
Bed Days OR Time Direct Cost
12,213 222,870 $ 4,314,788
4,077
1,823
1,825
67,142 $ 868,564
1,194
6,603 120,939 $ 2,443,628
517
1,211
32,800 $ 544,856
3,535
16,488 325,785 $ 6,335,392
3.1 14
2.474 - 206.167 .$ 1,508,558
.

b ' z ~1'

1Z'PL

0'1
9'8
9'9

0'0
0'8 1
O'ZP

O'SZP' l
6'b2PdZ
1Xl.6'1

Z'PL

(O' 1

0'61
O'OP

.1

O'L 1

rn

O't'

l0't'

10'19L'Z

'O 1
6'6
P'S

O'PE
O'LZ
10'6~

6'LZP'Z
9'PP'Z
IP'LZC'1
0'101 1

O'P9
0'901

P'06L
L'Pt0 Z
OZP9

1SOHtltll3 l d 3 3 X 3 S3SWSl U3An

9llVW ld33X3 S3SV3SIO SMkJ3NVd
IlVd3H 3llOH037V CINQ SlS0HtlM3l
3t13NVd 0NV AUVIllElOlVd3H 83HlO)
1

LZE
Z6
EZE
PL6

O'E

O'Z

0'6P
0'90 1
0'1
O'EE

O'L 1
0'89
10'99
O'CZ
0'61
O31
0'69
O'L
1

O'E
S.

O'OL
O' LZ

P'981' 1
6'L6' 1

,O.SS

~O'PS1
Z'6PZ
L'LZZ
t1.65

0'1
0'98
O'L8E

LIM6 9 8 1 3 0 V N0113fl&JlS80'1'0 162

3 S f l S l 1 3 M 0 8 AUOLVWWVldNI 682
IMA3NVNE)lVW W31SAS 3AIlS391 082
V 'S3tln330tld VlNt13H l V t j 3 l t n I B OLZ
V ' ~ 3 U n 0 3 3 0 t f dVINtl3H lVtl31Wl18 692
9~ ' ~ 3 t l n 0 3 3 0 IVWOLS
~d
0 ~ tSl~ N W8 9 ~
9~ ' ~ ~ t l m 3 3 0 t~lVdW O L S
QNV S ~ N VL ~ Z
Ad03SOUVdW IVNIWOaW 9Z
33 H U M AWOlOWdVl 0Z
OIS ' 1 ~ 0 V H d 0 ~
3AlSN31X3
3
SS31 LZ
W 0 l ~ ' l t l 3 9 t l H d 0 S 3 3AISN31X3 SS37 992
O I S 'lV39VHd0S3 3AISN31X3 SS31 Z
Nd l V 1 3 3 t l CiNV 7VNIlS31NI HOWYV PSZ
a d l V 1 3 3 U NV 7 t l N I 1 ~ 3 l N%0l'Vl4
l
na '8 H W W O l S 'lV3EltlHdOS3 8OrtlW' ZSZ
0 3 3 0 t l d AWOlS0103 '8 AWOlSOtllSVEI 1SZ
1300W A 8 CUNI330 WHLO 000

CMG
Description

Cases

Preferred
LOS
Case

Direct

Cases

Recommended
1 Case

LOS

Direct

438 OTHER DERMATOLOGIC PROCEDURES

756
801
803
804
805
821
900
904
O00
028
, D61
063
Gen
000
& Spec
040
127
294
295
,

366
437

438
504
508
510
513
554
703
711
734

POST-OPERATIVEAND POST-TRAUMA
WOUND DEBRIDEMENT FOR DlAGNOSl
OTHER PROCEDURES FOR DIAGNOSIS
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
COMPLICATIONS OF TREATMENT AG
EXTENSIVE UNRELATED O.R. PROCE
NON-EXTENSIVE UNRELATED O.R. P
OTHER DEFINED BY MODEL
Gl PROCEDURES (ENDOSCOPY)
BIOPSY
TRANSFUSIONS
OTHER DEFINED BY MODEL
TRACHEOSTOMY & GASTROSTOMY PRO
OTHER RESPIRATORY PROCEDURES W
ESOPHAGITIS, GASTROENTERITIS 8
ESOPHAGlTlS, GASTROENTERITIS &
BACK AND NECK PROCEDURES WlTHO
OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES
MAJOR URINARY TRACT PROCEDURES
MlNOR UPPER URINARY TRACT PROC
TRANSURETHRAL PROSTATECTOMY W l
OTHER TRANSURETHRAL OR BIOPSY
MlSCELiANEOUS MALE REPRODUCTIV
PTHER O.R. PROCEDURES OF THE 6
RETCULOENOOTHELIALAND IMMUN1
ILL-DEFINED NEOPLASM WlTH OTHE

9.0
1.O
1.0

5.9
2.0
10.0

5.4
37.2
187.2

3.0

26.3
3.O

283.2

1.O
7.0
4.0

7.0
135.0
2,286.0
115.0
10.0
265.0
1.0
1.O
1.O
1.O
26.0
4.0
3.0
5.0
17.0
18.0
22.0
70.0
8.O
1.O

3.0

8.4
18.5
24.3
0.0
0.0
0.0
0.0
3.7
87.0
1.O
2.0
3.0
6.2
1.O
4.3
5.6,.

10.1
6.8
2.3
1.1
1.O
1.O
1.3-

10.0
1.O
0.0

5.9
2.0
10.0

2.0

26.3

65.2
32.5
222.9

1,547.4
860.0
5,379.0
9,097.3
829.0
2,826.0
6,376.8

2.0
4.0
2.0

3&0
8.4
18.5

54.4
31.9
35.4
25.0
155.5
52.3
166.2
58.2
O. O
61.2
110.5
60.9
75.2
173.4
83.3
75.2
68.5
114.4
58.9
0.0
52.8

5,947.6
128.1
126.7
311.2
41 1.6
1,230.8
39,238.0
679.0
374.0
1,634.0
1,438.3
544.8
1,462.7
2,443.2
3,112.0
1,735.9
881.6
815.8
874.0
377.O
613.7

4.0
176.0
2,401.O
121.0
9.0
260.7
0.0
2.0
2.0
0.0
26.0
4.0
0.0
5.0
17.0
18.0
22.0
70.0
9.0
2.0
4.0

24.3
0.0
0.0
0.0
0.0
3.5
79.9
1.O
1.9
2.9
5.7
1.O
4.0
53
9.1
6.3
2.2.
1.O
1.O
1.O
1.3

5.4
37.2
187.2
283.2
65.2,
32.5
222.9
54.4
31.9
35.4
25.0

155.5
52.3
166.2
58.2
0.0
61.2
110.5
60.9
75.2
173.4
83.3
75.2
68.5
114.4
58.0
0.0
52.8

2,826.
6,376.
5,947.
121.
120.
295.
411.
7,142.1
37,276.1
645.1
355.3
1,552.3
1,366.4
517.9
1,388.
2,321 .C
2,956.4
1,649.1
837.6
775.1
830.
358.2
583.C

CMG
Description

1.

756
801
803
804
805
821
DO0
D35

Msk

D61
000
185
350
352
353
, 354
356
357
362
366
368
370
371
372
374
377
378
379
380
391
437
1 438

POST-OPERATIVE AND POST-TRAUMA

WOUND DEBRIDEMENT FOR DIAGNOSI
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DIAGNOSIS
COMPLICATIONS OF TREATMENT AGE
OTHER DEFINED 8 Y MODEL
BIADDER & URETHRAL PROCEDURES
BlOPSY
OTHER DEFINED BY MODEL
AMPUTATION EXCEPT UPPER LlMB A
MULTIPLE OR BILATERAL JOINT RE
HIP REPLACEMENTWITH CC
HIP REPLACEMENT WITHOUT CC
KNEE REPLACEMENT
FRACTURED FEMUR PROCEDURES WIT
FRACTURED FEMUR PROCEDURES WIT
MUSCULOSKELETAL BIOPSY WITHOUT
BACK AND NECK PROCEDURES WITHO
MAJOR HlP AND KNEE PROCEDURES
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR UPPER EXTREMITY PROCEDUR
MINOR LOWER EXTREMITY PROCEDUR
WOUND DEBRIDEMENTAND SKlN GRA
SOFT TISSUE PROCEDURES (MNRH)
OTHER MUSCULOSKELETAL PROCEDUR
~OTHERLOWER EXTREMITY PROCEOUR
SECONDARY NEOPLASMS AND PATHOL
OTHER DERMATOLOGIC PROCEDURES
IOTHER
DERMATOLOGIC PROCEDURES

Cases

11.0
1 .O
1 .O

2.0
4.0
6.0
223.0
1,088.0
40.0
174.0
7.0
15.0
130.0
88.0
135.0
48.0
39.0

1
1

11.0
16.0
13.0
25.0
60.0
2.0
26.0
3.0
6.0
35.0
58.01
1 .O
2.0
1 .0l

Recommended
1 Direct Cases LOS 1 Case 1
Cost
Time
0.0
1,338.9
12.0
0.0
5.0
50.2
1 0,272.0
0.0
17.1
50.2
0.0
192.2
4,345.0
192.2
9.9
180.2
3,873.0
1 .O
8.7
180.2
59.8
5,874.5
24.5
59.8
2.0
8.0,
721.5
0.0
0.0
2.4
219.1
222.5
0.0
44.3
44.3
0.0
21.8
194.1- 1,143.0
21.8
35.6
38.0
35.6
342.5
O.0
1,413.9 177.6
5.3
66.4
66.4
6,269.9
6.0
84.6
26.7
84.6
10.0
159.0
4,604.1
159.0
8.8
4,370.3 130.0
177.9
8.5,
177.9
88.0
166.1
3,962.9
1 66.1
7.8
3,205.5 135.0
147.4
6.9
147.4
39.0
3,558.1
12.4
11 8.4
118.4
32.0
2,667.6
2 07.9
107.9
10.7
2,498.9
85.2
7.0
4.6
85.2
1,197.4
5.3
17.0
149.7
149.7
12.0
6.8
138.4
2,077.9
138.4
121.9
2,047.0
27.0
7.2
121.9
114.5
114.5
1,391.7- 63.0
4.6
3.0
7.5
2,282.5
142.2
142.2
3.5
80.3
34.0
80.3
1,049.0
92.5
1.7
92.5
1,125.3
4.0,.
8.0
1,216.8
4.8
100.2
100.2
923.6
2.6
46.0
92.5
92.5
74.61
894.71
76-01
2.21
74.61
666.0
2.0
0.0
3.0
0.0
3.0
1,067.5
76.7
2.5
76.755.21
1,389.01
2.01
6.01
55.21

Preferred
LOS
Case
Time

5.3,
19.0
11.0
9.5
27.0
2.5
0.0
0.0
0.0
5.5,

26.7
8.8
9.0
8.1
7.2
12.4
1 0,7
4.6

5.3
6.8
7.2
4.6
7.5
3.5
1.7
4.8
2.6
2.21
3.0
2.5
6.01

Direct
Cost

1,272.C
9,244.
3,910.
3,679.4
5,580.8
685.4
211.4
174.7
318.1
1,343.2
6,269.9
4,604.1
4,151.8

3,558.1
2,667.6
2,498.8

1,197.4
2,077.5
2,047.0
1,391.7
2,282.5
1,049.C

894.71
666.
1,067.
1 ,389.4

CMG
Description

LOS

~LYMPHOMAAND CHRONIC LEUKEMIA

1
1.O
r 7 3 4 1 1 ~ ~ - D E F I N ENEOPLASM
D
W ~ T H E
1.O
750 MULTISYSTEMIC OR UNSPECIFIED S
6.0
2.0
756 POST-OPERATIVEAND POST-TRAUMA
, 801
WOUND DEBRIDEMENT FORKAGNOSI 5.0
803 OTHER PROCEDURES FOR DIAGNOSIS
17.0
804 OTHER PROCEDURES FOR DlAGNOSlS
12.0
805 OTHER PROCEDURES FOR DIAGNOSIS
7.0
821 ICOMPLICATIONS
OF TREATMENT AGE
l
3.0

1 728

000 OTHER DEFINED BY MODEL

125 MAJOR CHEST PROCEDURES
126 OTHER RESPIRATORY PROCEDURES W
127 ,OTHER RESPIRATORY PROCEDURES W
184 MAJOR RECONSTRUCTIVEVASCULAR
1 187 ~OTHERVASCULAR PROCEDURES
192 VEIN LlGATlON AND STRIPPING
250 EXTENSIVE GASTROINTESTINAL PRO
251 GASTROSTOMY & COLOSTOMY PROCED
2 5 2 'MAJORESOPHAGEAL,STOMACH
8 DU
292 G.I. OBSTRUCTION AGE 18-69 WIT
294 ESOPHAGITIS, GASTROENTERITIS &
295 ESOPHAGITIS, GASTROENTERITIS &
296 ESOPHAGITIS, GASTROENTERITIS&
703 OTHER O.R. PROCEDURES OF THE B
904 NON-EXTENSIVE UNRELATED O.R. P
DO0 ~OTHERDEFINED BY MODEL
D28
PROCEDURES (ENDOSCOPY)
Oncology 000 OTHER DEINED BY MODEL
010 NEOPiASM OF NERVOUS SYSTEM

Thoracic

1
1

IGI

48.0
17.0
7.0
5.0
27.0
15.0
94.0
2.0
1.O
1.a
1.a
2.0
13.0
1.C
9.C
1.C
168.0
105.C
406.C
12.C

Case
Tlme

Direct
Cost

Cases

LOS

Case
Tirne

Dlrect
Cost

CMG
#

CMG
Description

ORBITAL PROCEDURES

050

1 075 IRADICAL
LARYNGECTOMY AND GLOSS
1 077
081
084
087
088
089
125
126
127
131
187
250
251
252

253
254
255
256
257
258
267
268
280
295
300
311
312
314

~LESSEXTENSIVE HEAD AND NECK P

SALIVARY GLAND PROCEDURES
MISCELLANEOUS EAR, NOSE AND TH
SINUS PROCEDURES
ETHMOIDECTOMY (MNRH)
DENTAL EXTRACTlONlRESTORATlON
MAJOR CHEST PROCEDURES
OTHER RESPIRATORY PROCEDURES W
OTHER RESPIRATORY PROCEDURES W
RESPIRATORY NEOPLASMS
OTHER VASCULAR PROCEDURES
EXTENSIVE GASTROINTESTINAL PR
GASTROSTOMY & COLOSTOMY PROCED
MAJOR ESOPHAGEAL, STOMACH & DU
MAJOR INTESTINAL AND RECTAL PR
MAJOR INTESTINAL AND RECTAL PR
LESS EXTENSIVE ESOPHAGEAL. STO
LESS EXTENSIVE ESOPHAGEAL,STOM
LESS EXTENSIVE ESOPHAGEAL, STO
LAPAROTOMY WlTH CC
,ANUSAND STOMAL PROCEDURES, AG
ANUS AND STOMAL PROCEDURES, AG
DIGESTIVE SYSTEM MALIGNANCY Wl
ESOPHAGITIS, GASTROENTERITIS &
OTHER G.I. DIAGNOSES AGE 18
MAJOR PANCREATtC PROCEDURES
MAJOR HEPATOBlLlARY PROCEDURES
~OTHERHEPATOBILIARYAND PANCRE

Cases
-

Preferred
Case
LOS
Tlme

4.0
8.01

2.0
35.4 1

14.01
35.0
9.0
8.0
2.0
1.O
102.0
26.0
24.0
59.0
3.0
4.0

5.11
2.5
1.3
3.9
1.O

26.0
4.0

57.0
43.0
1.0
1.0
1.O
4.0

5.0

3.0
36.0
2.0
11.0
10.0
21.O
9.01

1.0
10.6
3.0
10.8
10.9
4.7
24.0
19.1
12.5
13.7
10.0
20.0
22.0

Direct

110.9
525.51

Cost
1,056.5
18,058.81

140.6
62.9
161.O
68.7
69.2
217.1
97.2
144.8
25.8
91.1
402.7
235.0
241.4
199.9
206.6
153.2
223.2

1,765.41
1,312.6
703.2
1,890.3
647.0
737.0
4,820.5
1,469.3
4,612.0
2,51 0.7
1,621.7,
14,559.5
7,360.8
5,865.3
4,949.1
4,162.0
11,O1 5.0
6,422.0

5.01

210.21

4,641.O1

12.5,
2.8
3.3
12.71.5
5.5
47.9
11.7

54.4
88.6
88.5
57.5
0.0
59.3
365.9
424.9
162.81

3,051.5
1,287.0
1,472.3
2,899.8
320.5
1,268.0
31,796.6
5,883.2
3.620.11

11.61

Cases

6.0

5.01
15.01
46.0
12.0
5.0
3.0

2.0
71.O
34.0
16.0
77.0
4.0
2.0
18.0
6.0
57.0
43.0
0.0
0.0

0.0(
6.0

7.0,
4.0
38.0
3.0
15.0
7.0
14.0
8.01

Recommended
LOS
Case
Time
1.9
35.41

110.9
525.51

4.71
2.3
1.3
3.5
1.O
1.O
9.5

123.71
140.6
62.9
161.0
68.7
69.2
217.1
97.2
144.8
25.8
91.i
402.7
235.0
241.4
199.9
206.6,
153.2'
223.2

2.8
10.8
9.9
4.7
21.9
17.5
11.9,
12.5
9.2
18.0

19.8
5 .O 1
11.3
2.7
3.0
11.5
1.4
5.1
47.9
11.1
10.51

210.2(
54.4
88.6
88.5

57.5
0.0
59.3
365.9
424.9
162.81

Direct
Cost
1,003.7
16.252.q
1,677.1
1,247~
668.1
1,701.2,
614-7

6,892.8

9,91 3.5
5,779.4

4,176.E(
2,888,G
1,222.7
1,325.1
2,754.8
304.
1,204.
28,616.
5,589.
3.439.

fl

CMG
#

CMG
Description

324
350
352
353
354
355

PANCRUTIC CANCER OR OTHER MAL

MULTIPLE OR BliATERAL JOINT RE
HIP REPLACEMENT WlTH CC
HI? REPLACEMENT WITHOUT CC
-KNEE REPLACEMENT
~REATTACHMENTPROCEDURES OR LOW

357
361
362
366
368
370
371
372

FRACTURED FEMUR PROCEDURES WlT

MUSCULOSKELETAL BlOPSY FOR MAL
MUSCULOSKELETAL BIOPSY WlTHOUT
BACK AND NECK PROCEDURES WlTHO
MAJOR HIP AND KNEE PROCEDURES
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR UPPER UTREMITY PROCEDUR
MINOR LOWER EXTREMlN PROCEDUR
DEBRIDEMENT AND SKIN GRA
SOFT TISSUE PROCEDURES (MNRH)
OTHER MUSCULOSKELETAL PROCEDUR
OTHER LOWER EXTREMlTY PROCEDUR
JOINT REPLACEMENT FOR MALIGNAN
SECONDARY NEOPLASMSAND PATHOL
BREAST PROCEDURES EXCEPT BlOPS

374
377
378
379
380
383
391
428

NOUND

433
437
438
477
479
504
508

SUBTOTAL MASTECTOMY 8, OTHER BR

OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES
~PARATHYROIDPROCEDURES
THYROID PROCEDURES
MAJOR URlNARY TRACT PROCEDURES
MlNOR UPPER URINARY TRACT PROC

Preferred
LOS
Case
Time
30.0
13.3
15.8
1.O
2.0
368.2
2.0
29.5
227.7
2.0
5.0
163.2
3.0
8.0
139.8
17.01
253.7 1
15.81

Cases

6.0
46.0
47.0
1.O
13.0
4.0
8.0

13.0..
17.0
16.0
36.0
46.0
3.0
31.O
19.0

63.0
31.O
29.0
18.01
109.0
19.0

2.0

11.5
8.6
4.3
10.0
10.8
19.5
7.6
11.2
7.8
12.8
6.6
6.2
3.0
15.0
15.7

205.3
73.3
71.2
375.2
122.0
326.9
180.2
241.O

2.2
3.0
3.4
3.1
2.9
7.0
48.0

Direct
Cases
Cost
2,706.9
33.9
3,702.0
2.0
5,870.0
1.O
3,614.5
3.0
4,215.0
2.0
6,893.51
11.0l
3,511.5
2,615.4
1,380.5
5,282.0
2,628.0
6,327.5

Recommended
LOS
Case
Time
12.0
15.8
1.9
368.2
27.1
227.7
4.8
163.2
7.6
139.8
15-81
253.71

Direct
Cost
2,571.6
3,516.
5,576.
3,433.
4,004.
6.548.4

196.4
120.0
110.2
80.2
202.5
27.9

4,002.5
2,495.2
3,707.9
2,203.6
1,983.9
1,375.3
7,595.3
2,960.6

4.0
46.0
47.0
0.0
12.0
2.0
5.0
9.0
11.0
11.0
38.0
43.0
4.0
21.O
20.0

10.5
8.0
4.1
9.0
9.9
17.6
6.9
10.1
7.8
11.5
6.0
5.6
2.7
14.7
14.3

205.3
73.3
71.2
375.2
122.0
326.Q
180.2
241.O
162.7
196.4
120.0
110.2
80.2
202.5
27.9

3,335.q
2,484.6
1,311.4
4,753.8
2,496.
5,694,
2,361,
3,602.4
2,245.7
3,337.1
2,093.4,
1,785.
1,237.
7,215.4
2,812.6

100.5
79.2
97.1
133.91
137.1
178.6
0.0

1,096.3
1,144.4
1,352.1
1,582.31
1,484.7
2,993.3
10.010.5

82.0
41.O
38.024.01
142.0
18.0
1.O

2.1
2.7
3.2
3.11
2.6
6.7
43.2

100.5
79.2
97.1
133.91
137.1
178.6
0.0

1,041.4

162.7

2,623.5

2,843.7
9.009.3

CMG
Description
510 TRANSURETHRAL PROSTATECTOMY Wl
513 OTHER TRANSURETHRAL OR BIOPSY
578 MAJOR GYNECOLOGICAL PROCEDURES
580
703
711
725
727
728
729
731
732
734
735
736
750

756
801
803
804
805
821
900
904
DO0
DO4
DO8
D28
035
D61

Cases

Preferred
1 Case

LOS

9.0
41 .O
24.0 1

2.9
1.7
4.91

MAJOR GYNECOLOGICAL PROCEDURES

56.0
,OTHR O.R. PROCEDURES OF THE B
4.0
IRETICULOENDOTHELIAL
AND IMMUN1
1
40.0 1
MAJOR LEUKEMIA AND LYMPHOMA PR
1
28.01
ACUTE LEUKEMtA WJTHOUT MAJOR P
5.0
LYMPHOMA AND CHRONlC LEUKEMIA
26.0
20.0
LYMPHOMA AND CHRONlC LEUKEMIA
23.0
LYMPHOMA AND CHRONIC LEUKEMIA
UMPHOMA AND CHRONIC LEUKEMIA
13.0
ILL-DEFINED NEOPLASM WlTH OTHE
23.0
65.0
RADIATIONTHERAPY
CHEMOTHERAPY
198.0
MULTISYSTEMIC OR UNSPEClFtED S
2.0
POST-OPERATIVE AND POST-TRAUMA
3.0
WOUND DEBRIDEMENT FOR DIAGNOSI
1.O
OTHER PROCEDURES FOR DlAGNOSlS
1.O
OTHER PROCEDURES FOR DlAGNOSlS
5.0
,OTHER PROCEDURES FOR DlAGNOSlS
1.O,
1
1.O[
[COMPLICATIONS OF TREATMENT AGE
EXTENSIVE UNRELATEO O.R. PROCE
4.0
NON-EXTENSIVE UNREIATED O.R. P
7.0
OTHER DEFINED BY MODEL
322.0
ORBITAL AND OTHER EYE PROCEDUR
9.0
EXTERNAL EYE PROCEDURES
48.0
428.0
GI PROCEDURES (ENDOSCOPY)
88.0
BLADOER & URETHRAL PROCEDURES
BlOPSY
1.O

2.9
4.0
9.6 1
10.81

26.4
3.6
20.9
14.8
17.1
8.7
3.4
3.2
44.0

7.3
3.0
7.0
12.8
10.0
22.01
14.0

22.3
0.0
0.00.0
0.0
0.0
0.0

Recommended

68.3
56.9

120.91
125.5
147.7
2.81
148.91
0.0
68.8
91.3
10.4
0.0
90.7
O.0
0.0
165.7
20.1
43.2
105.2
233.8
190.2
60.21
42.5
71.7
36.4
38.9
53.8
44.5
25.4
40.0

Dlrect

Cases

LOS

Case

1,098.4
742.2
1,614.71

12.0
54.0
32.01

2.7
1.6
4.71

68.3
56.9
120.9

1,269.6
2,315.8
2,90581
5,317.41
8,811.8
1,651
5,819.4
3,859.0
4,691.2
3,108.9
2,282.8
1,163.6
26,111.0
1.583.3
829.0
2,026.0
4,283.8
2,708.0
5,108.0l
4,265.5
7,474.9
225.7
335.7
343.9
152.3
236.1

73.0
2.0
42.01
19.01
4.0

2.8
3.6
8.71
9.8 1
26.4
3.4
20.9
13.3
15.4
7.9
3.4
2.9
39.6
6.6
2.9

125.5
147.7
2.8
148.9
0.0
68.9
91.3

.c

129.0

26.0
10.0
21.O
12.0
21.O
85.0
258.0
1.O

4.0
2.0
0.0
3.0

2.0
0.01
6.0
4.0
419.0
12.0
51.0
450.0
115.0
2.0

10.4
--

11.5
9.5
19.81
13.3
20.1
0.0
0.0
0.0
0.0
0.0

0.0
90.7
O. 0
O. 0
165.7
20.1
43.2
105.2
233.8
190.2
60.2
42.5
71.7
36.4
38.9
53.8
44.5
25.4

0.0

40.0

6.4,.

Direct

CMG
Description

Dental

Eye

,
,

Cases

Preferred
LOS

Case
Time

Direct
Cost

Cases

Recommended
LOS
Case
Time
0.0
148.0
49.9
1.Q
66.2
1.7
62.5
1.3
85.3
1.2
72.5
1.O
68.1
1.O
3.3
124.4
103.0
2.5
106.6,
1.8
50.9
1.370.7
1.4
71.2
1.2
120.3
1.3
66.6
1.O
48.3
1.3
58.3
4.0
94.3
4.016.1
3.0
10.0
34.6
58.3
1.2
55.4
6.0
118.3
2.8
109.9
2.4
2.0
110.8
3.0
114.3
56.3
1.1
2.6
107.1.
95.8
3.3
50.3
1.O
0.0
50.7

Dlrect
Cost

063 TRANSFUSIONS
148.0
266.0
0.0
296.4
217.1
169.0
000 OTHER DEFINED BY MODEL
2.0
717.8
330.0
681.9
347.0
49.9
050 ORBITAL PROCEDURES
706.9
706,9
60.0
1.7
64.0
66.2
052 RETINAL PROCEDURES
582.3
150.0
62.5
1.4
553.2)
158.0
055 LENS INSERTION WlTH CC (MNRH)
85.3
624.4
84.0
1.2
624.4
79.0
056 LENS INSERTION WITHOUT CC (MNR
513.0
292.0
72.5
278.0
1.O
487.3
057 OTHER OPHTHALMIC PROCEDURES (M
1.1
642.5
182.0
192.0
68.1
124.4
3.3
076 MAJOR HEAD AND NECK PROCEDURES
1,604.0
6.0
7.6
906.8
103.0
077 LESS EXTENSIVE HEAD AND NECK P
2.5
13.0
14.0
081 SALIVARY GLAND PROCEDURES
1.8
15.0
106.6
10.0
1,139.7
084 MISCELLANEOUS AR, NOSE AND TH
620.2
50.948.048.0
1.3087 SINUS PROCEDURES
731 .O
30.0
28.0
1.4
70.7
731.d
088 ETHMOIDECTOMY (MNRH)
773.3
71.2
1.2
48.0
45.0
773.3
089 DENTAL EXTRACTIONIRESTORATION
56.0
120.3
1.3
43.9
887.1
887.1
091 NASAL PROCEDURES (MNRH)
669.3
704.5
144.0
66.6
1.1
136.0
126 0
OTHER RESPIRATORY PROCEDURES W
902.0
902.
5.0
6.0
127 OTHER RESPIRATORY PROCEDURES W
58.3
938.7
938.7
3.0
4.0
4.0
257 LESS EXTENSIVE ESOPHAGEAL, STO
1,196.0
1,196.0
94.3
1.O
2.0
4.0
294 ESOPHAGITIS, GASTROENTERITIS 8
571
571.5
3.0
2.0
16.1
3.0
295 ESOPHAGITIS, GASTROENTERITIS &
1,640.0
34.6
1,640.C
10.0
2.0
1.O
437 OTHER DERMATOLOGIC PROCEDURES
686.2
58.3
5.0
686.2
6.0
1.2,.
438 OTHER DERMATOLOGIC PROCEDURES
6.0
1,717.3
55.4
1,717.3
4.0
6.0
477 PARATHYROID PROCEDURES
30.0
1,372.8
118.3
21.O
1,372.8
2.8
479 THYROID PROCEDURES
109.9
2.5
1,205.4
1,230.4
83.0,
119.0
703 OTHER O.R. PROCEDURES OF THE B
110.8
2.0
1,169.3
1,169.3
4.0
6.0
729 LYMPHOMA AND CHRONIC LEUKEMIA
114.3
3.0
1,984.C
1,984.0
2.0
1.O
803 OTHER PROCEDURES FOR DlAGNOSlS
468.5
56.3
468.5
13.0
1.1
10.0
804 OTHER PROCEDURES FOR DlAGNOSlS
12.0
2.6
1,155.1
9.0
107.1
1,155.1
805 OTHER PROCEDURES FOR DlAGNOSlS
1,051.3
95.8
1,051.3
6.0
3.3
4.0
904 NON-EXTENSIVE UNRELATED O.R. P
50.3
176.0
176.C
1.O
1.O
2.0
DO0 OTHER DEFINED BY MODEL
166.0
0.0
158.0
212.5
201 .q
50.7

.q

Appendix H: Mode1 Recommendations 1 1 % Budget Reduction

Table H. 1 lists a surnrnary of volume and cost model recommendations for each clinicai team after an 1 1% budget reduction; Table H.2 provides

a detailed listing of volume and cost model recommendations. The results appearing in the two tables assume sofi bound on case deviations of Il 1%
and hard bounds of 160%. Sofi bounds of *I 1% and hard bounds of &20%ofcurrent physician practice were also assumed for the nins used to
generate the results appearing in H. 1 and H.2.

Table H.1 Model Recommendation Sumrnary 11% Budget Reduction

1

GI
Gen & Spec

Msk
Thoracic

Oncology
DentlEveENT
Total

Cases
4,077
1,823
1,194
517
3,535
3.114
14.260

Preferred
Volume Model Recommendations
Bed Days OR Time Direct Cost
Cases
Bed Days OR Time Direct Cost
222,870 $ 4,477,864.
12,684
5,165
10,866 246,424 $ 4,026,175
67,142 $ 934,059~
1,949
836,126
1,968
1,559 68,225 $
6,786
120,939 $ 2,516,790
1,329
6,159 120,927 $ 2,366,311
32,800 $ 598,558
645
494,615
28,315 $
1,297
1,129
325,785 $ 6,820,616
17,684
4,351
15,455 318,405 $ 5,862,807
206,167 $ 1.562.517
2.563
3.1 19
2.129 193.407 $ 1.356.209
42.963
975,703&16,910,404
16.577
37,297 975,703 $ 14,942,243

Cost Mode1 Recornmendations

Cases
Bed Days OR Time Direct Cost
10,866 222,870 $ 4,026,175
4,077
1,704
1,823
67,142 $ 836,126
6,416 120,939 $ 2,366,311
1,194
32,799 $ 494,615
1,254
517
15,700 325,785 $ 5,862,807
3,535
2,296 206.1 67 $ 1,356.209
3,114
14.260
38.236 975,7023 14,942.243,

l'able H,i Detailed Mode1 Recommendations - 1 1 % Budget Reduction

Team

CMG
#

300
311
312
314
323
325
327
437
438

CMG
Description

OTHER G.I. DIAGNOSES AGE > 18

MAJOR PANCREATIC PROCEDURES
MAJOR HEPATOBILIARY PROCEDURES
OTHER HEPATOBILIARY AND PANCRE
ClRRHOSlS AND ALCOHOLIC HEPATI
PANCREAS DISEASES EXCEPT MALE
LlVER DISEASES EXCEPT ClRRHOSl
OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGlC PROCEDURES

Cases

Prefetred
1 Case

LOS

49.0
3.0
3.0
17.0
25.0

10.3
41.3

23.3
139.2

4.0

40.0

6.6
8.6

169.5
80.8
41.O
8.4
16.8

19.0
1.O
2.0

14.6
9.3

1.O
3.0

Direct

2,427.9
20,990.3
2,761.O
4,003.1
2,199.7
1,913.1
2,424.9

Cases

24.0
1.O

5.0
6.O
23.0
40.0
17.0

Recommended
LOS
Case

8.2
41.3

4.0

74.2

1,425.0

0.0

14.6
7.5.
5.5
8.6
1.O

42.7

884.0

4.0

3.0

23.3
139.2
169.5
80.8
41.O
8.4

16.82
74.2
42.7

Direct

2,160.d
20,990.3

2,761.C
3,562.
1,957.
1,702.7
2,158.4

1,425.
884.

CMG
Description

Gen
& Spec

750
756
801
803
804

MULTISYSTEMIC OR UNSPECIFIED S
POST-OPERATIVEAND POST-TRAUMA
WOUND DEBRIDEMENT FOR DlAGNOSl
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS

900
904
DO0
028
D61
D63
000
040
127
294
295
366

EXTENSIVE UNRELATE0 O.R. PROCE

NON-EXTENSIVE UNRELATED O.R. P
OTHER DEFINED BY MODEL
GI PROCEDURES (ENDOSCOPY)
BIOPSY
TRANSFUSIONS
OTHER DEFINEO BY MODEL
TRACHEOSTOMY & GASTROSTOMY PRO
OTHER RESPIRATORY PROCEDURES W
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITtS, GASTROENTERITIS &
BACK AND NECK PROCEDURES WlTHO

438 OTHER DERMATOLOGIC PROCEDURES

504 MAJOR URINARY TRACT PROCEDURES
508 MINOR UPPER URINARY TRACT PROC
510 TRANSURETHRAL PROSTATECTOMY W1
513 OTHER TRANSURETHRAL OR BlOPSY
554 MISCELLANEOUS MALE REPRODUCTJV
703 OTHER O.R. PROCEDURES OF THE 8
711 RETICULOENDOTHELIALAND IMMUN1
734 ILL-DEFINED NEOPLASM WlTH OTHE
750 MULTISYSTEMIC OR UNSPECIFIED S

Cases

Preferred
LOS
Case

7.0
9.0

10.6
5.9

3.0

26.3

7.0
135.0
2,286.0
115.0
10.0
265.0
1.0
1 .O
1.O
1.O
26.0

24.3
O. 0
0.0
O. 0
0.0
3.7
87.0
1.O

3.0

4.3

5.0

5.6

17.0
18.0
22.0
70.0
8.0
1.O
3.0
1.O

1O. 1
6.8
2.3
1.1
1.O
1.O
1.3
1.O

2.0
3.0
6.2

Direct

Cases

Recommended
LOS
Case

Direct

63.1
5.4

3,927.6
1,547.4

2.0
8.0

10.6.
5.9

63.1
5.4

3,927.
1,547.

31.9
35.4
25.0
155.5
52.3
166.2
58.2-.
0.0
61.2
110.5

128.1
126.7
311.2
411.6
1,230.8
39,238.0
679.0
374.0
1,634.0
1,438.3

216.0
3,418.5
184.0
4.0
207.0
0.0
2.0
2.0
0.0
26.0

0.0
0.0
0.0
0.0
3.3
81.6
1.O
2.0
3.0
5.1

31.9
35.4
25.0
155.5
52.3
166.2
58.2
0.0,
61.2
110.5

114.C
112.7
277.9
411.
1,095.
37,738,
679.
374.
1,634.C
1,280.1
1,301.8
2,174.4
2,769.7
1,545.
784.7'
726.1
777.9
377.C
546.
473.

1,462.7

O. 0

3.9

75.2

173.4~.

2,443.2

5.0

5.6

173.4

83.3
75.2
68.5
114.4
58.9
0.0
52.8
50.2

3,112.0
1,735.9
881.6
815.8
874.0
377.0
613.7.
473.0

17.0
18.0
22.0
70.0
13.0
2.0
5.0
2.0

8.0
5.6
2.1
0.9
0.9
1.O
1.3
1.O

83.3
75.2
68,s
114.458.9
0.0
52.8
50.2

75.2

Team

CMG

CMG
#

801
803
804
805
821
DO0

Msk
,

D35
D61
000
185
350
352
354
356
357
362
366
368
370

371
372
374
377
378
379
380
391
437
438
728

Description

WOUND DEBRIDEMENT FOR DlAGNOSl

OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
COMPLICATIONS OF TREATMENT AGE
OTHER DEFINEO BY MODEL
BLADDER & URETHRAL PROCEDURES
BJOPSY
OTHER DEFINED BY MODEL
AMPUTATlON EXCEPT UPPER LlMB A
MULTIPLE OR BILATERAL JOINT RE
HIP REPLACEMENT WlTH CC
KNEE REPLACEMENT
FRACTURED FEMUR PROCEDURES WlT
RACTURED FEMUR PROCEDURES WIT
MUSCULOSKELETAL BIOPSY WITHOUT
BACK AND NECK PROCEDURES WITHO
MAJOR HIP AND KNEE PROCEDURES
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR UPPER EXTREMITY PROCEDUR
MlNOR LOWER EXTREMITY PROCEDUR
WOUND DEBRIDEMENT AND SKlN GRA
SOFT TISSUE PROCEDURES (MNRH)
OTHER MUSCULOSKELETAL PROCEDUR
OTHER LOWER EXTREMITY PROCEDUR
SECONDARY NEOPLASMS AND PATHOL
OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES
LYMPHOMA AND CHRONIC LEUKEMIA

Cases

1.O

7 .O
2.0
4.0
6. O

--.

223.0
1,088.0
40.0
174.0
7.0
15.0
130.0
88.0

Preferred
LOS
Case
Time
50.2
19.0
192.2
11.0
9.5
180.2
27.0
59.8
0.0
2.5
O. 0
44.3
0.0
21.8
35.6
0.0
5.5
66.4
26.7
84.6
8.8
159.0
9.0
177.9

Direct
Cases
Cost
10,272.0
0.0
0.0
4,345.0.
3,873.0
0.0
5,874.5
1.O
10.0
721.5
222.5
248.0
194.1 1,251.4
45.0
342.5
1,413.9
154.0
6.0
6,269.9
4,604.1
6.0
130.0
4,370.3

Recornmended
Case
LOS
Tlme
19.0
11.0
7,7
21.6
2.5
0,O
0.0
0.0
4.9
26.7
8.8
8.0

Direct
Cost
I
10,272.0
4,345.4
3,447.C
5,228.3
642.1
198.C
172.8
304.4

50.2

192.2
180.2
59.8
0.0
44.3
21.8
35.6
66.4
84.6
159.0
177.9

135.01
48.0
39.0
11.0
16.0
13.0
25.0

60.0
2.0
26.0
3.0
6.0
35.0
58.0
1.O
2.0
1.O
1.O

12.4
10.7
4.6
5.3
6.8
7.2
4.6
7.5
3.5
1.7
4.8
2.6
2.2,
3.0

2.5
6.0
3.0

118.4.
107.9
85.2.
149.7
138.4
121.9
114.5
142.2
80.3
92.5
100.2
92.5
74.6
0.0
76.7
55.2
60.2

3,558.1
2,667.6
2,498.9
1,197.4
2,077.9
2,047.0
1,391.7
2,282.5
1,049.0
1,125.3
1,216.8
923.6
894.7
666.0
1,067.5
1,389.0
1.314.0

41 .O
34.0
4.0
14.0
5.0
20.1
53.0
4.0
42.0
5.0
10.0
56.0
93.0
2.0
4.0
2.0
2.0

12.4
10.7
4.6
5.3
6.8
7.2
4.6

7.5
3.5
1.7
4.8
2.6
2.2
3.0

- -

25
6.0
3.0

118.4
107.9
85.2
149.7
138.4
121.9
114.5
142.2
80.3
92.5
100.2
92.5
74.6
0.0
76,7
55.2
60.2

3,558.1
2,667.6
2,498.9
1,197.4
2,077.9
2,047.C
1,391.7
~

2,282.5
1,049.C
1,125.3
1,216.8
923.6
894.7
666.C
1,067,t
1,389.C
1.314.C,

---

CMG
Description

Thoracic

734
750
756
801
803
804
805
821
DO0
000
125
126
127
184
187
192
250
251
252
292
294
295
296
703
904
000
028

Cases

ILL-DEFINED NEOPLASM WlTH OTHE

MULTISYSTEMIC OR UNSPEClFlED S
POST-OPERATIVE AND POST-TRAUMA
~OUND
DEBRIDEMENT FOR DIAGNOS1
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
COMPLICATIONS OF TREATMENT AGE
1
~OTHERDEFINED SY MODEL
1
OTHER DEFlNED BY MODEL
MAJOR CHEST PROCEDURES
OTHER RESPIRATORY PROCEDURES W
OTHER RESPIRATORY PROCEDURES W
MAJOR RECONSTRUCTIVE VASCUUR
OTHER VASCULAR PROCEDURES
VElN LlGATlON AND STRIPPING
EXTENSIVE GASTROINTESTINAL PRO
~GASTROSTOMYa COLOSTOMY PROCED
1
[MAJOR ESOPHAGEAL, STOMACH & DU
1
~G.I.
OBSTRUCTION AGE 18-69 WIT
1
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITIS, GASTROENTERITIS B
ESOPHAGITIS, GASTROENTERITIS &
OTHER O.R. PROCEDURES OF THE B
NON-EXTENSIVEUNRELATED OFF
OTHER DEFINED BY MODEL
GI PROCEDURES ENDOSCOPY)
--

1.O
6.0
2.0

5.O
17.0
12.0
7.0
3.0
245.0
48.0
17.0
7.0
5.0
27.0
15.0
94.0
2.0
1.O
1O
.
1.O
2.0

13.0
1.C
9.a
1.c
168.0
105.C

Oncology
010 NEOPLASM OF NERVOUS SYSTEM
040 TRACHEOSTOMY & GASTROSTOMY PRO

12.C
1.C

Pieferred
LOS
Case

Direct

Cases

Recommended
LOS
Case 1

Direct

CMG
#

CMG
Description

Cases

Preferred
Case
LOS

Cases
Direct
Cost
18,058.8
3.0
13.0
5,016.4
1,765.4
12.0
56.0
1,312.6
15.0
703.2
3.0
1,890.3,
647.0
4.0
737.0
2.0
40.0
4,820.5

Tirne

075
076
077
081
084
087
088
O89
125
126
727
131
187
250
251
252
253
254
255

,256
257
258
267
268
280
295
300
311
312
314

RADICAL LARYNGECTOMY AND GLOSS

MAJOR HEAD AND NECK PROCEDURES
LESS EXTENSIVE HEAD AND NECK P
SALIVARY GLAND PROCEDURES
MISCELLANEOUS EAR, NOSE AND TH
SINUS PROCEDURES
ETHMOIDECTOMY (MNRH)
DENTAL EXTRACTIONIRESTORATION
MAJOR CHEST PROCEDURES
PTHER RESPIRATORY PROCEDURES W
OTHER RESPIRATORY PROCEDURES W
-RESPIRATORY NEOPLASMS
OTHER VASCUiAR PROCEDURES
EXTENSIVE GASTROINTESTINAL PRO
GASTROSTOMY & COLOSTOMY PROCED
MAJOR ESOPHAGEAL, STOMACH & DU
IMAJOR
INTESTINALAND RECTAL PR
MAJOR INTESTINALAND RECTAL PR
LESS EXTENS1VE ESOPHAGEAL, STO
LESS EXTENSIVE ESOPHAGEAL,STOM
LESS EXTENSIVE ESOPHAGEAL, STO
LAPAROTOMY WlTH CC
ANUS AND STOMAL PROCEDURES, AG
ANUS AND STOMAL PROCEDURES, AG
DIGESTIVE SYSTEM MALIGNANCY Wl
,ESOPHAGITIS, GASTROENTERITIS B,
OTHER G.I. DIAGNOSES AGE > 18
MAJOR PANCREATIC PROCEDURES
MAJOR HEPATOBILIARY PROCEDURES
OTWER HEPATOBILIARY AND PANCRE

8.0
34.0
14.0
35.0
9.0
8.0
2.0
1O
.
102.0

26.0

24.0
59.0
3.0
4.0
26.0
4-01
57.01
43.0
1.0
1.0
1.O
4.0
5.0
3.0
36.0
2.0
11.0
10.0
21.O
9.0

35.4
12.3
5.1
2.5
1.3
3.9
1.O
1.O
10.6

525.5
309.4
123.7
140.6
62.9
161.0
68.7
69.2
217.1

Recommended
Case
LOS
Time
28.3
525.5
309.4
10.4
123.7
4.2
2.2
140.6
1.3
62,9~
161.O
3.9
1.O
68.7
1.O
69.2
217.1
9.3

Direct
Cost
14,706.5
4,464.
1,571.
1,168.
625.
1,512.
5754

655.9
4,290.2

3.0

97.2

1,469.3

42.0,

2.5

97.2

1,307.7

10.8
10.9
4.7
24.0
19.1
12.5
13.71
10.0
20.0
22.0
5.0
12.5
2.8
3.3
12.7
1.5
5.5
47.9
11.7
11.6-

144.8
25.8
91.1
402.7
235.0
241.4
199.91
206.6
153.2
223.2
210.2
54.4
88.6
88.5

4,612.0
2,510.7
1,621.7
14,559.5
7,360.8
5,865.3
4,949.11
4,162.0
11,015.0
6,422.0
4,641.O
3,051.5
1,287.0
1,472.3
2,899.8
320.5
1,268.0
31,796,6
5,883.2
3,620.1

9.0
95.0
5.0
2.0
10.0
7.0
57.01
43.0
0.0
0.0
0.0
7.0
8.0
4.0

9,3
8.7
4.7
24.0
16.3
12.5
12.21
8.9
20.0
22.0
5.0
12.5
2.5
3.3
12.7
1.5
5.5
47.9,
10.4
11.6

144.8
25.8
91.1
402.7
235.0
241.4
199.91
206.6

4,104.7
2,234.q
1,443.g
12,958.C
6,551.1
5,220.1
4,404.q
3,704.
8,812.
5,137.
3,712.
2,715.
1,145.
1,177.
2,580.
285.
1,128.
25,437.
5,236.
3,221.

57.5
0.0
59.3
365.9
424.9
162.8

58,O
4.0
18.0
4.0
8.0

10.0

153.2

223.2
210.2
54.4
88.6
88.5

57.5
0.0
59.3
365.9
424.9
162.8

KGI

leam

Cases

LOS

350 MULTIPLE OR BILATERAL JOINT RE

352 H P REPLACEMENT WITH CC

1.O
2.0

2.0
29.5

353 HIP REPLACEMENT WITHOUT CC

2.0

5.0

3.0
17.0
8.0
6.0
46.0
47.0
1.O
13.0
4.0
8.0
13.0
17.0
16.0
36.01
46.0
3.0
31.O
19.0
114.0
103.0
63.0

8.0
15.8
14.5
11.5
8.6
4.3
10.0
10.8
19.5

29.0
18.0
209.0
19.0
2.0
9.0

3.4
3.1
2.9
7.0
48.0

354
355
356
357
361
362
366
368
370
371
372
374
,,377
378
379
380
383
391
428
432
433
437
438
477
479
504
508
510

Preferred

CMG
Description

KNEE REPLACEMENT
REATTACHMENT PROCEDURES OR LOW
FRACTURED f MUR PROCEOURES WlT
- FRACTURED FEMUR PROCEDURES WIT
MUSCULOSKELETAL BlOPSY FOR MAL
MUSCULOSKELETAL BIOPSY WITHOUT
BACK AND NECK PROCEDURES WlTHO
MAJOR Hi? AND KNEE PROCEDURES
MAJOR LOWEF? EXTREMITY PROCEDUR
MAJOR LOWER EXTREMllY PROCEDUR
MAJOR UPPER EXTREMITY PROCEDUR
MlNOR LOWER EXTREMITY PROCEDUR
W OUND DEBRIDEMENT AND SKlN GRA
~SOFTTISSUE PROCEDURES (MNRH)
OTHER MUSCULOSKELETAL PROCEDUR
OTHER LOWER EXTREMITY PROCEDUR
JOINT REPLACEMENT FOR MALIGNAN
SECONDARY NEOPLASMS AND PATHOL
BREAST PROCEDURES EXCEPT BIOPS
SUBTOTAL MASTECTOMY & OTHER BR
SUBTOTAL MASTECTOMY & OTHER BR
OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES
PARATHYROID PROCEDURES
THYROID PROCEDURES
MAJOR URINARY TRACT PROCEDURES
MlNOR UPPER URINARY TRACT PROC
.TRANSURETHRAL PROSTATECTOMY WI

7.6
11.2
7.8
12.8
6.6 1
6.2
3.0
15.0
15.7
1.1
1.8
2.2

2.9

Case
Tirne
368.2
227.7
163.2
139.8
253.7~
316.4
205.3
73.3
71-2
375.2
122.0
326.9
180.2
241.O
162.7
196.4,
120.01
110.2
80.2
202.5
27.9
57.1
93.3
100.5

97.1
133.9
137.1
178.6.
0.0
68.3

Cases
Direct
~ost
3,702.0
0.0.
0.0
5,870.0
1.O
3,614.5
1.O
4,215.0
6.O
6,893.5
4,509.6
3.0
2.0
3,511.5
46.0
2,615.4
47.0
1,380.5
0.0
5,282.0
2,628.0
5.0
6,327.5
1.O
2,623.5
3.0
4,002.5
5.0
6.0
2,495.2
6.0
3,707.92,203.61
40.01
40.0
1,983.9
1,375.3
5.0
7,595.3
12.0
2,960.6
31.0
600.0
183.0
1,O18.6
165.0
1,096.3
101.O
1,352.1
1,582.3
1,484.7
2,993.3
10,010.5
1,098.4,

47.0
29.0
175.0
7.0
0.0
15.0

Recommended
LOS
Case
1 ~ime
368.2
2.0
227.7
24.3
5.0
163.2
139.8
8.0
253.7
14.0
316.4
12.9
205.3
9.7
8.2
73.3
71.2
3.8
375.2
8.0
122.0
8.8
326.9
15.6
6.1
180.2
241.O
11.2
6.2
162.7
10.3
196.4
5.31
120.0 (
110.2
6.2
2.4
80.2
15.0
202.5
12.7
27.9
0.9
57.1
1.6
93.3
1.9
100.5

3.0
3.1
2.9
6.2
38.4
2.6.

97.1
133.0
137.1
178.6
0.0
68.3.

Direct
~ost

3,125.2
2,327.7

2,338.q

5,062.C
2,008.e
3,202.C
1,996.1
2,966.
1, ~ 6 l
1,500.#
1,205.4,
6,759.
2,634.
480.0
897.4
975.7

.q

1,203.4,
1,408.
1,187,
2,664.1
8,008.4
977.6)

CMG
Description

513
578
579
580
703
711
725
727

728
729
731
732
734
735
736
750
756
801
803
804
805
821
900
904
DO0
DO4
DO8
028
D35

OTHER TRANSURETHRAL OR BlOPSY

MAJOR GYNECOLOGICAL PROCEDURES
MAJOR GYNECOLOGICAL PROCEDURES
MAJOR GYNECOLOGICAL PROCEDURES
OTHER O.R. PROCEDURES OF THE 8
RETICULOENDOTHELIALAND IMMUN1
MAJOR LEUKEMIA AND LYMPHOMA PR
ACUTE LEUKEMIA WITHOUT MAJOR P
LYMPHOMA AND CHRONIC LEUKEMIA
LYMPHOMA AND CHRONIC LEUKEMIA
LYMPHOMA AND CHRONIC LEUKEMIA
LYMPHOMA AND CHRONIC LEUKEMIA
ILL-DEFINED NEOPLASM WlTH OTHE
IRADIAT~ON
THERAPY
CHEMOTHERAPY
MULTJSYSTEMIC OR UNSPECIFIED S
POST-OPERATIVE AND POST-TRAUMA
WOUND DEBRIDEMENT FOR DlAGNOSl
~OTHERPROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
COMPLICATIONS OF TREATMENT AGE
EXTENSIVE UNREiATED O.R. PROCE
NON-EXTENSIVE UNRELATED O.R. P
OTHER DEFINED BY MODEL
ORBITAL AND OTHER EYE PROCEDUR
EXTERNAL EYE PROCEDURES
GI PROCEDURES (ENDOSCOPY)
BLADDER & URETHRAL PROCEDURES

061 BIOPSY

Cases

41 .O
24.0
27.01
56.0
4.0
40.0
28.0
5.0
26.0
20.0
23.0
13.0,
23.0
65.01
198.0
2.0
3.0

1.O
1.O!
5.0
1.O
1.0
4.0
7.0
322.0
9.O
48.0
428.0
88.0
1.O

Preferred
LOS
Case

1.7
4.9
4.01
2.9
4.0
9.6
10.8
26.4
3.6
20.9
74.8
17.1
8.7
3.4)
3.2
44.0
7.3
3.0
7-01
12.8
10.0
22.0

56.9
120.9
132.81
125.5
147.7
2.8
148.9
0.O
68.9
91.3
10.4
0.0
90.7
0.01
0.0
165.7
20.1
43.2
105.2 (
233.8
190.2
60.2
42.5
71.7
36.4
38.9
53.8

14.0
22.3
0.0
0.0
0.0
0.0
0.0
0.0

44.5.
25.4
40.0

Direct
742.2
1,614.7
1,576.71
1,269.6
2,315.8
2,905.8
5,317.4
8,811.8
1,651.0
5,819.4
3,859,O
4,691.2
3,l 08,9
2,282.81
1,163.6
26,111 .O
1,583.3
829.0
2,026.01
4,283.8
2,708.0
5,108.0
4,265.5
7,474.9
225.7
335.7
343.9
152.3
236.1
129.0

Cases

66.0
39.0
44.0)
90.0
1.O
64.0
11.0
4.0
26.0
10.0
37.0
15.0
9.0
104.01
317.0
0.0
5.0

2.0
0.0 1
2.0
2.0

O. O
7.0
3.0
516.0
15.0
77.0
685.0
141.0
2.0

~ecommendedCase

LOS

1.5
4.9

3.61
2.9
4.0
7.7
10.8
26.4
3.2
16.7
11.9
13.7
7.0
3.4 1
2.6
44.0
5.9
2.7
5.71
12.8
10.0
22.0
14.0
17.0
0.0
O.O1
0.0
0.0
0.0
0.0

56.9
120.9
132.81
125.5
147.7
2.8
148.9
0.0
68.9
91.3
10.4.
0.0
90.7
0.01
0.O
165.7
20.1
43.2
105.21
233.8

190.2
60.2
42.5
71.7
36.4
38.9
53.8
44.5
25.4
40.0

Direct
660.6
1,437.1

4,732.4
7,842.3
1,469.4
5,179.9
3,434.5
4,175.1
2,766.9
2,031.71
930.
20,888.
1,409.
737.
1,803.1l
$81 2.E
2,41 O. 1
4,086.4
3,796.
6,652.
180.
268.
287.
121.
197.
114.

CMG
Description

Dental
E Y ~
NT

000
050
052
055
056
057
076
077
081
084
087

OTHER DEFINED BY MODEL

ORBITAL PROCEDURES
RETINAL PROCEDURES
LENS INSERTION WlTH CC (MNRH)
LENS INSERTION WITHOUT CC (MNR
OTHER OPHTHALMIC PROCEDURES (M
MAJOR HEAD AND NECK PROCEDURES
LESS EXTENSIVE HEAD AND NECK P
SALIVARY GLAND PROCEDURES
MISCELLANEOUS EAR, NOSE AND TH
SINUS PROCEDURES

1 126 ~OTHERRESPIRATORY PROCEDURES W

127
257
294
295
437
438
477
479
,703
729
803
804
805
904
DO0

Cases

OTHER RESPIRATORY PROCEDURES W

LESS EXTENSIVE ESOPHAGEAL, STO
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITIS, GASTROENTERITIS &
~OTHERDERMATOLOGIC PROCEDURES
~OTHERDERMATOLOGIC PROCEDURES
PARATHYROID PROCEDURES
THYROID PROCEDURES
OTHER O.R. PROCEDURES OF THE B
LYMPHOMA AND CHRONlC LEUKEMIA
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
QTHER PROCEDURES FOR DlAGNOSlS
NON-EXTENSIVE UNRELATED O.R. P
OTHER DEFINED BY MODEL

DO4 ORBITAL AND OTHER EYE PROCEDUR

Preferred
LOS

330.0
64.0
150.0
84.0
278.0..
192.0
6.0
14.0
15.0
48.0
30.0

2.0
1.7
1.4.
1.2
1.O
1.1
3.3
2.5
1.8
1.3,
1.4,

6.0)
3.0
1.O
2.0
2.0
5.0
6.0,
30.0
119.0
6.0
1.O
10.0
9.0
4.0
1.O
158.0
223.0

4.0
4.0
3.0
10.0
1.2
6.0
2.8
2.5
2.0
3.0
1.1
2.6
3.3
1.O
0.0
0.0

Case

Direct

49.9
66.2
62.5
85.3
72.5
68.1
124.4
103.0106.6
50.9
70.7

717.8
706.9
582.3
624.4
513.0
642.5
1,604.0
906.8
1,139.7
620.2
731.O

48.31
58.3
94.3
16.1
34.6
58.3
55.4
118.3
109.9
110.8
114.3
56.3
107.1
95.8
50.3
50.7

902.0)
938.7
1,196.0
571.5
1,640.0
686.2
1,717.3
1,372.8
1,230.4
1,169.3
1,984.0
468.5
1,155.1
1,051.3
176.0
212.5
210.3

56.4

Cases

528.0
25.0

133.0
33.0,
247.0
76.0
10.0
5.0
6.0
0.0
14.2

5.01
5.0
2.0
4.0
0.0
6.0
2.0
12.0
47.0
2.0
2.0
16.0
15.0
7.0
2.0
140.0
357.0

Recommended
LOS

Case

1.7
13
1.2
1.1
O.9
1.O
3.0
2.5
1.6
1.2
1.3

49.9
66.2
62.5
85.3
72.5
68.1
124.4
103.0
106.6
50.9
70.7

1.31

48.31
58.3
94.3
16.1
34.6
58.3
55.4
118.3
109.9.
110.8
114.3.
56.3
107.1
95.8
50.3
50.7
56.4

4.0
4.0
2.4
10.0
1.2
6.0
2.5
2.2
2.0
3.0
1.1
2.6
2.9
1.O
0.0
0.0

Direct
599.
629,
518.
555.7

1,014.4
552,C
650.6

802.4
835.4
1,064.4
508,
1,458.
610.7

1,095.1
1,040.7
1,765.8
417.C
1,028.q
935.
156.
189.1
187.2)

Appendix 1: Mode1 Recommendrtions

- 18% Budget Reduction

Table 1.1 lists a sumrnary of volume and cost model recommendations for each clinical team afier an 18% budget reductions; Table 1.2 provides
a detailed listing of volume and cost model recomrnendations. The results appearing in the two tables assume sofi bound on case deviations of &5%

and hard bounds of kl00%. Sofi bounds of *5% and hard bounds of k3O%of current physician practice were also assumed for the runs used to
generate the results appearing in 1.1 and 1.2.

Table 1.1 Model Recommendation Summarv - 18% Budget Reduction

Cases

GI
Gen & Spec
Msk
Thoracic

Oncology

4,077
1,823
1,194
517
3,535

Preferred
Bed Days OR Time

12,684
1,949
6,786
1,297
17,684

Direct Cost

222,870$ 4,477,864
67,142$ 934,059
120,939$ 2,516,790
32,800$ 598,558
325,785s6,820,616

Volume Model Recommendations

Cases
Bed Days OR Time Direct Cost

6,134
2,559
1,407
721
4,736

9,526
675
5,856
1,072
14,051

264,005$ 3,748,584
73,062$ 704,809
122,250$ 2,289,880
25,484$ 458,438
298,972$ 5,309,197

Cost Model Recommendations

Cases
Bed Days OR Time Direct Cost

4,077
7,823
1,194
517
3,535

9,934 222,870 3,748,584

1,621 67,142
704,808
5,856 120,939 2,289,880
458,438
1,090 32,800
14,584 325,785 5,309,197

rable 1.2 Detailed Model Recommendations 18% Budget Reduction

CMG
Descrlption

G1

Cases

OTHER DEFlNED BY MODEL

GASTROSTOMY & COLOSTOMY PROCED
MAJOR ESOPHAGEAL, STOMACH & DU
MAJOR INTESTINAL AND RECTAL PR
MAJOR INTESTINAL AND RECTAL PR
LESS EXTENSIVE ESOPHAGEAL, STO
256 LESS EXTENSIVE ESOPHAGEAL,STOM
257 . LESS EXTENSIVE ESOPHAGEAL. STO
258 LAPAROTOMY WITH CC
265 ABDOMINAL LAPAROSCOPY
267 ANUS AND STOMAL PROCEDURES, AG
268 ANUS AND STOMAL PROCEDURES, AG
269 BILATERAL HERNIA PROCEDURES, A
270 BILATERAL HERNIA PROCEDURES,A
280 DIGESTIVE SYSTEM MALIGNANCY WI
289 INFLAMMATORY BOWEL DISEASE
291 G.I. OBSTRUCTION AGE 18-69 WlT
292 G.I. OBSTRUCTION AGE 18-69 W l i
294 ESOPHAGITIS, GASTROENTERITIS &
295 ESOPHAGITIS, GASTROENTERITIS &
296 ESOPHAGITIS, GASTROENTERITIS &
300 OTHER G.I. DIAGNOSES AGE > 18
311 MAJOR PANCREATIC PROCEDURES
312 MNOR HEPATOBiLIARY PROCEDURES
314 OTHER HEPATOBlLlARY AND PANCRE
323 ClRRHOSlS AND ALCOHOLIC HEPATI
325 PANCREAS DISEASES EXCEPT MALlG

387.0
86.0
1.0
154.0
55.0
7.0
69.0
15.0
19.0
23.0
66.0
68.017.0
33.0
1.O
106.0
49.0
29.0.
66.0
56.0
25.0
49.0
3.0
3.0
17.0,
25.0
40.0

000
251
252
253
254
255

1 327 1LIVER DISEASES EXCEPT CIRRHOSI

19.01

Preferred
LOS
Case

4.2
22.5
35,O15.6
11.1
14.9
8.7
8.3
14.0
3.5
3.3
2.4
6.7
2.9
3.0
9.0
3.8
6.5
4.9
5.9
9.9
10.3
41,3
4.0
14.6

59.4
227.7
249.2
210.8
214.7
149.3
122.5
151.5
115.6
107.2
65.6
67.8
142.9
iog.9
0.0
28.0
1.0
5.5
18.4
21.8.25.7
23.3
139.2
169.5
80.8
41.0
8.4
16.81

9.3
6.6
8.61

Direct
1,225.9
8,505.2
19,927.0
5,801.4
4,442.3
5,886.4.
3,147.8
3,124.5
4,278.9
1,937.91
1,186.4
880.7
2,730.7
iI32a.o
642.0
2,034.7
790.4
1,473.9
1,101.8
1,327.4
2,454.6
2,427.9
20,990.3
2,761.0
4,003.1
2,199.7
1,913.1
2,424.91

Cases

0.0
86.0
0.0
154.0
55.0
3.0
7.0
0.0
10.0
3.0
3.0
136.0
2.0
0.0
2.0
106.0
98.0
58.0
132.0
50.7
22.0
20.0
3.0
6.0
0.0
23.0
40.0
17.01

Recommende
LOS
Case

3.4
16.3
35.0
12.2
8.9

14.9
7.1
8.3
9.8
2.6
2.7
2.0
6.7
2.2
3.06.3
3.1
5.1
3.4
4.2
6.9
7.2
41.3
4.0
10.8
7.5
5.4
6.5 1

59.4
227.7
249.2
210.8
214.7
149.3
122.5
151.5
115.6
107.2
65,6
67.8
142.9)
09.~1
0.01
28.0
1.O
5.5
18.4
21.8

25.7
23.3
139.2
169.5
80.8
41.0
8.4
16.81

Direct
7,005.3
6,974.
19,927.
4,757.
3,642.
5,886.4
2,581.2
3,124.9
3,549.7
1,589.1
972.5
722.2)
2,730.7'
1,089.~
642.C
1,668.4
648.2
1,208.e
903.5
1,088.9
2,012.8
1,990.E
20,990.1
2,761 .C
3,504.6
1,803.8
1,568.7
2,032.

Tea m

D61 BIOPSY
063 TRANSFUSIONS
Gen
8 Spec

000 OTHER DEFINED BY MODEL

TRACHEOSTOMY & GASTROSTOMY PRO

OTHER RESPIRATORY PROCEDURES W
ESOPHAGITIS, GASTROENTERITIS&
ESOPHAGITIS, GASTROENTERITIS&
BACK AND NECK PROCEDURES WlTHO
OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES
MAJOR URINARY TRACT PROCEDURES
MlNOR UPPER URINARY TRACT PROC
TRANSURETHRAL PROSTATECTOMY Wl
OTHER TRANSURETHRAL OR BIOPSY
MISCELLANEOUS MALE REPRODUCTIV
OTHER O.R. PROCEDURES OF THE 8
RETICULOENDOTHELIALAND IMMUN1
734 ILL-DEFINED NEOPLASM WlTH OTHE
750 ,MUtTlSYSTEMIC OR UNSPECIFIED S

040
127
294
295
366
437
438
504
508
510
513
554
703
71 1

115.0

0.0

10.0
265.0.
1.0
1.O

0.0
3.7
87.0

25.0
155.5
52.3
166.2

1.O

58.2

2.0
3.0
6.2
1 .O

0.0
61.2
110.5
60.9
75.2.
173.4
83.3
75.2
68.5
114.4

1.O
1.O

26.0
4.0
3.0
5.0
17.0
18.0

4.3
5.6
1O.1

70.0

6.8
2.3
1.1

8.0

1.O

22.0

1 .O

1 .O

3.0
1.O

1.3

1 .O

58.9
0.0
52.8
50.2

311.2
41 1.6
1,230.8
39,238.0
679.0
374.0
1,634.0
1,438.3
544.8
1,462.7
2,443.2
3,112.0
1,735.9
881.6
815.8
874.0
377.0
613.7
473.0

230.0
O.0
O.O
0.0
2.0
2.0
0,O

0.0
O.0
3.1
87.0
1 .O

26.0

4.5
1 .O
4.3
5.6
7.0
5.0
1.9
0.9

4.0
O.0
5.0
17.0
18.0
22.0

70.0
16.0
2.0
6.0
2.0

2.0
3.0

i .O

1 .O
1.3

1 .O

25.0
155.5
52.3
166.2
58.2
0.0

61.2
110.5
60,9

75.2
173.4
83.3
75.2
68.5
114.4
58.9
0.0
52.8
50.2

255.
411,
880.8

32,175.
556.
306,
1,339.
1,179.4
446.7
1,199.4
2,003.4
2,551.8
722,

669.
716.7
309.1
503.
387.

Tea m

Preferred

CMG
Description
756
801
803
804
805
821
DO0
D35
061

Cases

POST-OPERATIVEAND POST-TRAUMA
WOUND DEBRIDEMENT FOR DlAGNOSl
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
COMPLICATIONSOF TREATMENT AGE
OTHER DEFtNED BY MODEL
BLADDER & URETHRAL PROCEDURES
BlOPSY

LOS

11.0
1.O

5.3
19.0

1.O

11.0

2.0
4.0.
6.0
223.0
1,088.0

9.5
27.0-

7.0
15.0
130.0
88.0
135.0
48.01

26.7
8.8
9.0
8.1
7.2
12.41
10.7
4.6
5.3
6.8
7.2
4.6
7.5
3.5

2.5,
0.0
0.0

Case

Recommended

0.0
50.2
192.2
180.2
59.8O. 0
44.3
22.8

Direct

Cases

0.0
1,338.9
10,272.0
0.0
4,345.0
0.0
3,873.0
0.0
5,8745O. 0
721.5
12.0
2225
178.9
194.1 2,176.0

LOS

Case

Direct

5.3
19.0
11.0
9.5
27.0
2.5
0.0
0.0

O.O
50.2
192.2
180.2
59.8.
0.0
44.3
21.8

1,097.9

26.7
8.8

3.5
1.7
4.8
2.6
1.8
3.0

84.6
159.0
177.9
166.1
147.4
118.41
107.9
85.2
149.7
138.4'
121.9
114.5
142.2
80.3
92.5
100.2
92.5
74.6
0.0

6,269.9
4,604.1
3,868.4
3,5013
2,841.1
3,166.11
2,535.6
2,498.9
1,197.4
2,077.
2,047.
1,144.
2,282.
1,049.C
1,125.3
1,216.
923.
794.C
666.C

6.01

55.21

182.
135.9

Msk
MULTIPLE OR BILATERAL JOINT RE
HIP REPLACEMENT WITH CC
HIP REPLACEMENT WITHOUT CC
KNEE REPLACEMENT
~FRACTUREDFEMUR PROCEDURES WIT
FRACTURED FEMUR PROCEDURES WIT
MUSCULOSKELETAL BIOPSY WlTHOUT
BACK AND NECK PROCEDURESWlTHO
MAJOR HIP AND KNEE PROCEDURES
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR LOWER EXTREMliY PROCEDUR
MAJOR UPPER EXTREMITY PROCEDUR
'MINOR LOWER EXTREMITY PROCEDUR
WOUND DEBRIDEMENT AND SKiN GRA
SOFT TISSUE PROCEDURES (MNRH)
OTHER MUSCULOSKELETAL PROCEDUR
OTHER LOWER EXTREMITY PROCEDUR
SECONDARY NEOPLASMS AND PATHOL

1 438

~OTHERDERMATOLOGIC PROCEDURES

1
l

AMPUTATION EXCEPT UPPER LlMB A

185
350
,352
353
354
356
357
362
366
368
370
371
372
374
377
378
379
380
391

39.0

11.0
16.0
13.0
25.0
60.0
2.0
26.0
3.0
6.0
35.0
58.0
1.O

1.0l

1.7..
4.8
2.6
2.2
3.0
6.0 1

84.6
159.0

177.9
166.1
147.4
118.41
107.9
85.2
149.7
138.4
121.9
114.5
142.2
80.3
92.5
100.2
92.5
74.6
0.0

6,269.9
4,604.1
4,370.3
3,962.9
3,205.5
3,558.11
2,667.6
2,498.9
1,197.4
2,077.9
2,047.0
1,391.7
2,282.5
1,049.0
1,125.3
1,216.8
923.6
894.7
666.0

6.0
0.0
130.0
88.0
135.0
44.01
36.0
0.0
32.0
5.0
13.0
39.0
4.0
52.0
6.0
12.0
70.0
116.0
2.0
2.01

7.4
6.7
5.9
10.11
9.9
4.6
5.3
6.8
7.2
3.3

7.5

1.389.

CMG
Description

Team

728
734
750
756
801
803
804
805
821

DO0
000
125
126
127
184
187
192
250
251
252
292
294
295
296
703
904
DO0
028
Oncology 000
1 010
040

Thoracic

LYMPHOMA AND CHRONIC LEUKEMIA

ILL-DEFINED NEOPLASM WlTH OTHE
MULTISYSTEMIC OR UNSPECIFIEO S
POST-OPERATIVE AND POST-TRAUMA
WOUND DEBRIDEMENT FOR DlAGNOSl
OTHER PROCEDURES FOR DIAGNOSE
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
COMPLICATIONS OF TREATMENT AGE
OTHER DEFINED BY MODEL
OTHER DEFINED BY MODEL
MAJOR CHEST PROCEDURES
OTHER RESPIRATORY PROCEDURES W
OTHER RESPIRATORY PROCEDURES W
MAJOR RECONSTRUCTIVE VASCULAR
PTHER VASCULAR PROCEDURES
VElN LlGATlON AND STRIPPING
EXTENSIVE GASTROlNTESTlNAL PRO
GASTROSTOMY B COLOSTOMY PROCED
MAJOR ESOPHAGEAL, STOMACH & DU
G.I. OBSTRUCTION AGE 18-69 WIT
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITIS, GASTROENTERlTlS &
IESOPHAGITIS, GASTROENTERITIS 8
OTHER O.R. PROCEDURES OF THE B
NON-EXTENSIVE UNRELATED O.R. P
OTHER OEFINED BY MODEL
GI PROCEDURES (ENDOSCOPY)
OTHER DEFINED BY MODEL
NEOPLASM OF NERVOUS SYSTEM
TRACHEOSTOMY & GASTROSTOMY PRO

Preferred
LOS
Case
Time
60.2
1.O
3.0
77.2
1.O
1.O
62.3
6.0
5.0
24.6
19.5..
2.0
34.0
128.8
5.0
86.3
4.4
17.0
148.5
12.5
12.07.0
1O. 1
230.4
0.0
3.0.
14.3
245.0
0.0
27.4
26.2
48.0.
3.6
17.0
125.6
10.8
93.7
7.0
1.9
5.0
6.0
123.6
274.8
27.0
13.8
56.3
15.0
7.1
115.0
94.0
1.8
47.5
490.7
2.0
127.2
1.0
11.0
13.0
222.2
1.O
3.0
0.0
1.O
3.0
47.2
2.0
13.0
3.1
49.4
1-01
31.01
0.01
133.6
2.6
9.0
120.2
24.0
1.0
10.8
0.0
168.0
36.7
0.0
105.0
92.9
406.0
6.0
41.O
12.0
14.8
1.O
41.O
108.2

Cases

Dlrect
Cases
Cost
2.0
1,314.0
1,002.0
2.0
1,449.2
12.0
3,599.5
0.0
6,009.6
1.O
0.0
1,404.8
0.0
3,624.2
0.0
3,996.9
3,156.3
1.O
167.5490.0
96.0
1,127.9
4,925.8
17.0
1,076.3
O. 0
3,610.4
0.0
6,922.7
27.0,
2,652.5
2.0
943.8
2.0
24,788.5
0.0
3,778.0
2.0
5,059.0
0.0
732.0
2.0
553.0
4.0
1,030.821.3
12,000.01
0.01
1,608.6
0.0.
4,763.0
2.0
336.0
37.1
81.5
210.0
1,890.7
0.0
3,359.2
24.0
9.770.0
2.0

Recommended
Case
Time
60.2
3.0
77.2
1.O
62.3
5.0
19.5
24.6
34.0
128.8
86.3
4.4
12.5
148.5
230.4
1O. 1
0.0
14.3
0.0
27.4
26.2
2.9
8.9,
125.6,
1.9
93.7
6.0
123.6
9.6
274.8
56.3
7.1
1.5
115.0
47.5
490.7
11.0
127.2
13.0
222-2
0.0
3.0
3.0
47.2
3.1
49.4
0.01
133.6
2.6
120.2
24.0
10.8
0.0.
O. O
36.7
92.8
4.8
14.8
41 .O
41.O
108.2-

LOS

Direct
Cost
1,314.C
1,002.
1,449,
3,599,
6,009.
1,404,
3,624.
3,996.d
3,156.3
163.
789,
4,039.
882.
2,960.
5,407.3
2,175.1
660.6
17,951.C
3,098.C
4,148.4
600.
453.
845,
8,400.q
1,319.C
3,905.7
26.0
57.C
1,323.5
2,754.5
8.011.4i

Team

CMG
#

CMG
Description

050 ORBITAL PROCEDURES

075 RADICAL LARYNGECTOMY AND GLOSS
076 MAJOR HEAD AND NECK PROCEDURES
077 LESS EXTENSIVE HEAD AND NECK P
081 SALIVARY GLAND PROCEDURES
084 MISCELLANEOUS EAR, NOSE AND TH
087 SINUS PROCEDURES
088 ETHMOIDECTOMY (MNRH)
089 DENTAL EXTRACTION/RESTORATION
125 MAJOR CHEST PROCEDURES
126 OTHER RESPIRATORY PROCEDURES W
127 OTHER RESPIRATORY PROCEDURES W
131 RESPIRATORY NEOPLASMS
187 OTHER VASCULAR PROCEDURES

Cases

4.0
8.0
34.0
14.0
35.0
9.0
8.0
2.0
1.O
102.0
26.0
24.0
59.0

Preferred
1 Case
Tlme
2.0
110.9
525.5
35.4
309.4
12.3
123.7
5.1
2.5
140.6
1.3
62.9
3.9
161.0
1.O
68.7
1.O
69.2
10.6
217.1
3.0
97.2
10.8
10.9

LOS

Cases
Direct
Cost
8.0
1,056.5
0.0
18,058.8
0.0
5,016.4
1,765.4
3.0
1,312.6
70.0
703.2
18.01,890.3
0.0
647.0
4.0
737.0
2.0
4,820.5
0.0
1,469.3
52.0

2,510.7

Recomrnended
LOS

2.0
35.4
1O. 1
5.1
2.1
1.3
3.9
1.O
1.O
8.7
2.5

Direct
Cost
866.
14,706.
4,113.6
1,447.7
1,076.3
576.9
1,323.
530.
604.3
3,680.C
1,204.8

199.9
206.6
153.2
223.2
210.2
54.4 ,
88.6

4,058.
3,412.
7,71O.
4,644.
3,248.7
2,502. 2
1,055.4
1,140.2
2,377.9

Case
Tlme
110.9
525.5
309.4
123.7
140.6
62.9
161.O
68.7
69.2
217.1
97.2

118.0

251 GASTROSTOMY & COLOSTOMY PROCED

253
254
255
256
257
258
267

268
280
295
300
311
312
, 314

MAJOR INTESTINAL AND RECTAL PR

MAJOR INTESTINAL AND RECTAL PR
LESS EXTENSIVE ESOPHAGEAL, STO
LESS WTENSIVE ESOPHAGEAL,STOM
LESS EXTENSIVE ESOPHAGEAL, STO
LAPAROTOMY WlTH CC
ANUS AND STOMAL PROCEDURES, AG
ANUS AND STOMAL PROCEDURES, AG
DIGESTIVE SYSTEM MALIGNANCY Wl
ESOPHAGITIS, GASTROENTERITIS &
OTHER G.I. DIAGNOSES AGE > 18
MAJOR PANCREATIC PROCEDURES
MAJOR HEPATOBILIARY PROCEDURES
.OTHER HEPATOBILIARY AND PANCRE

57.0
43.0
1.0
1.0
1.O
4.0 .
5.0
3.0
36.0
2.O
11.0
10.0,
21.O

9.0.

13.7
10.0
20.0,
22.0
5.0.
12.5
2.8
3.3
12.7
1.5
5.5
47.9
11.7
11.6.

199.9
206.6
153.2
223.2
210.2
54.4
88.6
88.5
57.5
0.0
59.3
365.9
424.9
162.8.

4,949.1
4,162.0
11,O15.0
6,422.0
4,641.O
3,051.5
1,287.0
1,472.3
2,899.8
320.5
1,268.0
31,796.6.
5,883.2
3.620.1

57.0
43.0
0.0
0.0
0.0
8.0
10.0
0.0
72.0
4.0
22.0
0.0
0.0
16.0-

11.2
8.2
20.0
22.0
5.0
12.5
2.8
3.3
8.9
1.5
5.5
47.9
9.6
11.6-

88.5.
57.5
0.0
59.3
365.9
424.9
162.8

CMG
Description

Cases

Preferred
LOS 1 Case

lime
324
350
352
353
354
355

356
357
361
362
366
368
370
371
372
374
377
378
379
380
383
391
428
432
433
437
438
477
479
504
508

PANCREATIC CANCER OR OTHER MAL

MULTIPLE OR BlATERAL JOINT RE
HIP REPLACEMENT WlTH CC
HIP REPLACEMENT WITHOUT CC
KNEE REPLACEMENT
REATTACHMENT PROCEDURES OR LOW
FRACTURED FEMUR PROCEDURES WlT
FRACTURED FEMUR PROCEDURES WIT
MUSCULOSKELETAL BIOPSY FOR MAL
MUSCULOSKELETAL BIOPSY WITHOUT
BACK AN0 NECK PROCEDURES WlTHO
MAJOR HIP AND KNEE PROCEDURES
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR UPPER EXTREMITY PROCEOUR
MlNOR LOWER EXTREMITY PROCEDUR
WOUND DEBRIDEMENT AND SKlN GRA
SOFT TISSUE PROCEDURES (MNRH)
OTHER MUSCULOSKELETAL PROCEDUR
OTHER LOWER EXTREMITY PROCEDUR
JOINT REPLACEMENT FOR MALIGNAN
SECONDARY NEOPLASMS AND PATHOL
BREAST PROCEDURES EXCEPT BlOPS
SUBTOTAL MASTECTOMY & OTHER SR
SUBTOTAL MASTECTOMY & OTHER BR
OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES
PARATHYROID PROCEDURES
THYROiD PROCEDURES
MAJOR URINARY TRACT PROCEDURES
.MINOR UPPER URINARY TRACT PROC
,

30.0
1.O
2.0
2.0
3.0
17.0
8.0
6.0
46.0
47.0,
1.O
13.0
4.08.0
13.0
17.0
16.0
36.0
46.0
3.0
31.O
19.0114.0
103.0

63.0
31.O
29.0
18.0
109.0,
19.0
2.0

13.3
2.0
29.5
5.0
8.0
15.8
14.5
11.5
8.6
4.3
10.0
10.8
19.57.6
11.2
7.8
12.8
6.6
6.2.
3.0
15.0
15.7
1.1
1.8
2.2
3.0
3.4
3.1
2.9
7.0..
48.0

15.8
368.2
227.7
163.2
139.8
253.7
316.4
205.3
73.3
71.2
375.2
122.O
326.9
180.2
241.O
162.7
196.4
120.0
110.2
80.2
202.5
27.9
57.1
93.3

100.5
79.2
97.1
133.9
137.1
178.6
0.0.

Cases 1
Direct
Cost
60.0
2,706.9
3,702.0
1.O
0.0
5,870.0
3,614.5
0.0
0.0
4,215.0
2.0
6,893.5
0.0
4,509.6
3,511.5
0.0
2,615.4
46.0
1,380.5
47.0
5,282.0
0.0
2,628.0
O. 0
6,327.5
0.0
0.0
2,623.5
0.O
4,002.5
2,495.2
O. O
O. 0
3,707.9
2,203.6
0.0
1,983.9
0.0
1,375.3
6.0
7,595.3
11.O
38.0
2,060.6
600.0.
228.0
206.0
1,O18.6
1,096.3
126.0
62.0
1,144.4
58.0
1,352.1
36.0
1,582.3
218.0
1,484.7
0.0
2,993.3
0.0
10.010.5,

Recammended
LOS 1 Case 1

Tlme
9.3
2.0
29.5
5.08.0
12.9
14.5
11.5
7.0
3.5
10.0
10.8
19.5
7.6
11.2
5.4
10.1
4.6
4.4
3.0
12.3
11.0
0.9
1.5
1.8
2.1
2.8
2.6
2.4
5.7
48.0.

15.8
368.2
227.7
163.2
139.8
253.7
316.4
205.3
73.3
71.2,
375.2
122.0
326.9
180.2
241,O
162.7
196.4
120.0
110.2
80.1
202.5

27.9
57.1
93.3
100.5
79.2
97.1
133.9
137.1
178.6
0.0-

Direct
Cost
2,219.7
3,035.6
4,813.4
2,963.9
3,456.q
5,652.7
3,607.9
2,879.f
2,144.61
1,132.C
3,697.4
2,155.C
4,429.3
1,953.5
3,198.1
1,746.6

1,590.
1,127.
6,228.1
2,427.7
420.
713.
898.9
938.4
1,108.
1,297.
1,039.3
2,454.5
7,007.4

CMG
Description

510
513
578
579
580
703
71 1

TRANSURETHRAL PROSTATECTOMYWl
OTHER TRANSURETHRAL OR BIOPSY
MAJOR GYNECOLOGICAL PROCEDURES
MAJOR GYNECOLOGlCAL PROCEDURES
MAJOR GYNECOLOGJCAL PROCEDURES
OTHER O.R. PROCEDURES OF THE B

Cases

RETJCULOENDOTHELIAL
AND IMMN
7-

725 -MAJOR LEUKEMIA AND LYMPHOMA PR

DO8 ~EXTERNALEYE PROCEDURES

D28 GI PROCEDURES (ENDOSCOPY)
D35 BLAODER & URETHRAL PROCEDURES

D61 BlOPSY

9.0
41.0
24.0
27.01
56.0
4.0
40.0

2.9
1.7
4.9
4.01
2.9
4.0
9.6

28.0
5.01

10.8
26.41

N U T E LEUKEMIA W1THOUT MAJOR P

1
LYMPHOMA AND CHRONlC LEUKEMIA
26.0
LYMPHOMA AND CHRONlC LEUKEMlA
20.0
LYMPHOMA AND CHRONIC LEUKEMIA
23.0
LYMPHOMA AND CHRONIC LEUKEMIA
13.0
ILL-DEFINED NEOPLASM WlTH OTHE
23.0
735 RADIATION THERAPY
65.0
736 CHEMOTHERAPY
198.0
750 MULTlSYSTEMlC OR UNSPEClFlED S
2.0
756 POST-OPERATIVE AND POST-TRAUMA
3.0
1.O
801 WOUND DEBRIDEMENT FOR OlAGNOSl
1.O
803 OTHER PROCEDURES FOR DlAGNOSlS
5.0
804 .OTHER PROCEDURES FOR DlAGNOSlS
805 OTHER PROCEDURES FOR DlAGNOSlS
1.O
821 k0MPLlCATlONS OF TREATMENT AGE
1
1.01
900 IEXTENSIVE
UNRELATED O.R. PROCE
1
4.01
904 - ~ N - E X T ~ ~ S IUNRELATED
VE
O
F 7.0
DO0 OTHER DEFINED BY MODEL
322.0
727
728
729
731
732
734

48.01
428.0
88.0
1.O

Recommended

Preferred
LOS
Case
Ttme
68.3
56.9
120.9
132.8
125.5
147.7

2.8

3.6
20.9
14.8
17.1
8.7
3.4
3.2
44.0
7.3
3.0
7.0
12.8
10.0
22.01
14.01
22.3
0.0

148.9
0.0
68.9
91.3
10.4
0.0
90.7
0.0
0.0
165.7
20.1
43.2
105.2
233.8
190.2
60.2
42.5
71.7
36.4

0.01
0.0
0.0
0.0

53.8
44.5
25.4
40.0

CMG

Description

Dental

E Y ~

ENT

063
000
050
052
055
056
O57
076
077
081
084
087
088
089
091
126

127
257
294

29s
437
438
477
479
703
729
803
804
805
904
DO0

TRANSFUSIONS
OTHER DEFINED BY MODEL
PRBITAL PROCEDURES
RETINAL PROCEDURES
LENS INSERTION WlTH CC (MNRH)
LENS INSERTION
WITHOUT-- CC (MNR
- -- OTHER OPHTHALMIC PROCEDURES (M
MAJOR HEAD AND NECK PROCEDURES
~LESSEXTENSIVE HEAD AND NECK P
SALIVARY GLAND PROCEDURES
MISCELUNEOUS EAR, NOSE AND TH
SINUS PROCEDURES
ETHMOIDECTOMY (MNRH)
DENTAL EXTRACTIONIRESTORATION
NASAL PROCEDURES (MNRH)
OTHER RESPIRATORY PROCEDURES W
OTHER RESPIRATORY PROCEDURES W
LESS EXTENSIVE ESOPHAGEAL, STO
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITIS, GASTROENTERITIS &
,OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES
PARATHYROID PROCEDURES
THYROID PROCEDURES
OTHER O.R. PROCEDURESOF THE B
LYMPHOMA AND CHRONC LEUKEMIA
OTHER PROCEDURES FOR OlAGNOSlS
~OTHERPROCEDURES FOR DIAGNOSIS
OTHER PROCEDURES FOR DlAGNOSlS
NON-EXTENSIVE UNRELATED O.R. P
OTHER OEFINED BY MODEL
-

Cases

-- -

169.0
330.0
64.0
150.0
84.0
278.0
192.0
6.014.01
15.0
48.0
30.0

0.0
2.0
1.7
1.4
1.2
1.O
1.1
3.3
2.51
1.8
1.3
1.4

148.0
49.9
66.2
62.5
85.3
72.5
68.1
124.4
103.0 1
106.6
50.9
70.7

48.0

1.2

71.2

56.0
144.0
6.0
3.0
I,O
2.0
2.0
5.0
6.0
30.O
119.0
6.0
I.O

1,3
1.1
1.3
4.0
4.0
3.0
10.0
1.2
6.O
2.8
2.5
2.O
3.0
1.1

120.3
66.6
48.3
58.3
94.3
16.1
34.6
58.3
55.4
118.3
109.9
110.8.
114.3

10.0

Preferred
LOS 1 Case

9.01
4.0
1.O
158.0

56.3-

2.61
-

3.3
1.O
0.0

107.11
95.8
50.3
50.7

Cases

Direct
296.4
717.8
706.9
582.3
624.4
513.0
642.5
1,604.0
906.8l
1,139.7
620.2
731.O
..

773.3
887.1
704.5
902.0
938.7
4,196.0
571.5
1,640.0
686.2
1,717.3
1,372.8
1,230.4
1,169.3
1,984.0

468.5
1,15541
1,051.3
176.0
212.5

Rec immended
LOS

212.1
660.0
3.0
16.0
1.O
1.O
1.O
12.0
0.01
0.0
48.0.

0.0
1.6
1.4
1.1
1.O
0.8
0.9
3.3
2.5
1.0
1.1

3.0
0.0
0.0
2.0
10.6

1.2

6.0
2.0
4.0
0.0
10.0
0.0
0.0
O. O
0.0
2.0
20.0

mol
8.0

2.0
0.0

1.O
1.1
O. G
1.3
4.0
4.0
3.0
10.0
1.2
6.C
2.3
2.U
2.a
3.0
1.1
2.c
3.2
1.C
0.C

148.0

Direct

Team

CMG
Description

CMG
#

004
DO5
DO8
D60

ORBITAL AND OTHER EYE PROCEDUR

LENS PROCEDURES
EXTERNAL EYE PROCEDURES
DENTAL SURGERY

ne3

Cases

Preferred
Case
LOS

223.0
521.O
124.0
423.0

0.0
0.0
0.0
0.0

I.O

OO

Time
56.4
60.8
39.1
69.6

Direct
Cases
Cost
210.3
446.0
252.6 1,042.0
265.7
0.0
846.0
251.3
397 O
0.0

Recomrnended
LOS
Case
.

0.0
0.0
0.0
0.0
0.0

Tlme
56.4
60.8
39.1
69.6

Direct
Cost
447.
176.
217.9
175.9

3130O

739.4

Appendix J: Model Recommendations Mixed Strategy

Table J. 1 lists a summaiy of volume and cost model recommendations for each clinical team afier an 18% budget reduction, in which a mixed
solution strategy is implemented; Table J.2 provides a detailed listing of volume and cost model recommendations. The results appearing in the

two tables assume soft bound on case deviations of 18% and hard bounds of *6O%. Sofl bounds of 18% and hard bounds of &60%ofcurrent
physician practice were also assumed for the runs used to generate the results appearing in J. 1 and J.2.

Table J.1 Mixed Strategy Reeornrnendation Sumrnary

Cases
GI
Gen & Spec
Msk
Thoracic
Oncology
DenEveENT
Total

4,077
1,823
1,194
517
3,535
3.114
14,260

Required Price Changes

Bed Days OR Time Direct Cost

11,028,,
1,949
6,131..
1,297
15,911
2,564
38,880

222,871 $ 4,068,620
934,060
67,143 $
120,939$ 2,301,262
598,558
32,800 $
325,786 $ 5,821,402
206.169 $ 1.562.519
975.708 $15,286,421

Volume Model Recomrnendations

Cases Bed Days OR Time Direct Cost

4,593
1,996
1,331
617
4,321
3.157
16.015

9,895
1,523
5,513
1,118
13,713
2.113
33.875

231,750 $
69,203 $
121,608 $
30,619 $
321,360 $
201.163 $
975,703 $

3,728,051
831,528
2,150,060
502,108..
5,080,749
1,397,202
13,689,778

Cost Model Recommendations

Bed Days OR Time Direct Cost
10,203 222,871 $ 3,728,051
4,077

Cases

1,823
1,194
517
3,535
3.1 14
14.260

1,645 67,143 $ 831,528

5,978 120,939 $ 2,150,060
1,118 32,800 $ 502,108
13,952 325,786 $ 5,080,749
2.333 206.163 $ 1,397,282,
35,229 975.702 $ 13.689.778

Mixed Strategy Detailed Recommendations

Volume Model Results
LOS
Case

CMG

Description

Cases

1
Direct

1 Cases 1

Cost Model Results

Case

LOS

Direct

Cost

000
251
252
253
254
255
256
257
258

OTHER DEFINED BY MODEL

GASTROSTOMY d COLOSTOMY PROCEO
MAJOR ESOPHAGEAL, STOMACH 81DU
-MAJOR INTESTNA=
RECTAL PR
MAJOR INTESTINALAND RECTAL PR
LESS EXTENSIVE ESOPHAGEAL, STO
LESS EXTENSIVE ESOPHAGEAL,STOM
LESS EXTENSIVE ESOPHAGEAL, STO
LAPAROTOMY WlTH CC

267
268
269
270
280
289
291
292
294
295
296
300
311
312
314
323

NUS AND STOMAL PROCEDURES, AG

ANUS AND STOMAL PROCEDURES,AG

325

327
437

BIMTERAL HERNIA PROCEDURES, A

BILATERAL HERNIA PROCEDURES, A
DIGESTIVE SYSTEM MALlGNANCY WI
INFLAMMATORY BOWEL DISEASE
G.I.OBSTRUCTION AGE 18-69 WIT
G.I. OBSTRUCTION AG 18-69WIT
-ESOPHAGlTIS, GASTROENTERITIS &
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITIS, GASTROENTERITIS 8
OTHER G.I. DIAGNOSES AGE > 18
MAJOR PANCREATIC PROCEDURES
MAJOR HEPATOBILIARY PROCEDURES
OTHER HEPATOBILIARY AND PANCRE
ICIRRHOSIS
AND ALCOHOLIC HEPATI
PANCREAS DISEASES EXCEPT MALlG
LlVER DISEASES EXCEPT ClRRHOSl

OTHER DERMATOLOGIC PROCEDURES

472.2
86

2.83
18.77
35.00
-154
13.58
55
9.76
3
14.86
27
7.74
6
8.33
11
14.00

78
81
6
27

106
79
26
78
67
22
24
1
5
6
23 1
40
17

- -

2.96
2.13
6.71
2.88

3.00
7.33
3,76
6.52
3.85
4.69
9.92
10.27
41.33
4.00
14.65
9.28 1
6.58
8.63
1.O0
,,

59.37
227.7
249.2
210.8
214.7
149.3
122.5
151.5
115.6

$998.1
$7,586.7,
$19,927.0
$5,174.9
$3,962.61
$5,886.4
$2,807.91
$3,124.51
$4,279.01

387.0
86.0
1.O
154.0
55.0
7.0
69.0 15.0 1
19.0

2.8
16.4
35.0
13.6
9.8
14.9
7.7
7.8
9.6

59.4
227.7
249.2
210.8
214.7
149.3
122.5
151.5
115.6

$998.1
$6,974.3
$19,927.0
$4,757.2
$3,642.7
$5,886.4
$2,581.2
$3,005.6
$3,508.7

65.7
67.8
142.9
109.9

$1,058.3
$785.6
$2,730.7
$1,328,0
$642,0
$1,814.9
$705.1
$1,473.9
$982.8
$1,184.1
$2,454.6
$2,418.9
$20,990.3
$2,761.0
$4,003.1$2,199.71
$1,913.1
$2,425.0
$1.425.0

66.0
68.0
17.0
33.0
1.O
106.0
49.0
29.0
66.0
56.0
25.0
49.0
3.0
3.0
17.0
25.0 1
40.0
19.0
1.O

3.0

657
67.8
142.9
109.9

$972.9
$722.2
$2,109.2
$1,089.0
$642.0
$1,668.4
$648,2
$1,208.6
$903.5,
$1,088.5
$2,012.8
$1,990.9
$20,990.3
$2,761 .O,
$3,282.6
$1.803.8l
$1,568.7
$1,988.5
$1,425.0

1
1

28
1

5.5
18.4
21.8
25.7
23.3
139.2
169.5
80.8
41 1
8.4
16.9
74.2

2.1
4.3
2.9
3.0
6.3
3.3
6.5
3.3
4.0
6.6
6.7
41.3
4.0
14.7
9.3
6.6
8.6

1.O

,,

28.0
1.O

5.5

18.4
21.8
25.7
23.3
139.2
169.5
80.8
41.0 1
8.4
16.9
74.2

Tea m

CMG
Description

Volume r( ode1 Results

Case

lxGJ7G-

I
OTHER DERMATOLOGIC PROCEDURES
MULTlSYSTEMlC OR UNSPEClFlED S
POST-OPERATIVEAND POST-TRAUMA

1
1

803 ~OTHERPROCEDURES FOR DlAGNOSlS

804 OTHER PROCEDURES FOR DlAGNOSlS
805 OTHER PROCEDURES FOR DlAGNOSlS
821 COMPLICATIONS OF TREATMENT AGE
900 EXTENSIVE UNRELATED O.R. PROCE
904 NON-EXTENSIVE UNRELATED O.R. P
DO0 OTHER OEFINED BY MODEL

028 ~ GPROCEDURES
I
(ENDOSCOPY)

438
750
756

4
2
9

3.00
10.57
5.89

1
2
2

26.33
3.00
8.43
18.50

2
216
2,698.0 1
--

TRANSFUSIONS
Gen
OTHER DEFINED BY MODEL
& Spec
TRACHEOSTOMY & GASTROSTOMY PRO
OTHER RESPIRATORY PROCEDURES W
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITIS. GASTROENTERITIS&
366 IBACK AND NECK PROCEDURES WITHO
437 IOTHEU DERMATOLOGIC PROCEDURES
438 ~OTHERDERMATOLOGIC PROCEDURES
504 MAJOR URINARY TRACT PROCEDURES
508 MINOR UPPER URINARY TRACT PROC
510 TRANSURETHRAL PROSTATECTOMY WI
513 ~OTHERTRANSURETHRAL OR BIOPSY
554 IMISCELLANEOUS MALE REPRODUCTIV
703 OTHER O.R. PROCEDURES OF THE B
711 RETICULOENDOTHELIALAND IMMUN1
734 ILL-DEFINED NEOPLASM WlTH OTHE
750 MULTlSYSTEMlC OR UNSPECIFIED S

8
210.7

1
1

261

1
1

Cost Model Results

LOS
Case
-

Cases

1
5
17
18
22 1
70 1
IO
2

5
2

24.29

3.73
87.00
1.O0
2.00
3.00
6.15
1.oo
4.33
5.60
10.06
6.83
2.32
I.06
1.O0
1 O0
1.33
1.O0

42.7

$884.0

2.0

--

ripie

cost

063
000
040
127
294
295

1
Direct

3.0

42.7

Direct
Cost
$884.0

Msk

I
I

756
801
803
804
805
821
DO0
D35
D61
000
185
350
352
353
354
356
357
362
366
368
370
371
372
374
377
378
379
380
391
437
438

IPOST-OPERATIVE
AND POST-TRAUMA
~OUND
DEBRIDEMENT FOR DlAGNOSl
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
COMPLICATIONS OF TREATMENT AGE
OTHER DEFINED BY MODEL
BLADDER & URETHRAL PROCEDURES
BIOPSY
OTHER DEFINED BY MODEL
,AMPUTATION EXCEPT UPPER LlMB A
MULTIPLE OR BILATERAL JOINT RE
HIP REPLACEMENT WlTH CC
HIP REPLACEMENT WITHOUT CC
KNEE REPLACEMENT
FRACTURED FEMUR PROCEDURES WIT
FRACTURED FEMUR PROCEDURES WIT
MUSCULOSKELETALBIOPSY WITHOUT
BACK AND NECK PROCEDURES WlTHO
MAJOR HIP AND KNEE PROCEDURES
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR UPPER EXTREMITY PROCEDUR
MlNOR LOWER EXTREMITY PROCEDUR
WOUND DEBRIDEMENT AND SKiN GRA
SOFT TISSUE PROCEDURES (MNRH)
OTHER MUSCULOSKELETAL PROCEOUR
OTHER LOWER EXTREMITY PROCEDUR
SECONOARY NEOPLASMS AND PATHOL
OTHER DERMATOLOGIC PROCEDURES
~OTHERDERMATOLOGIC PROCEDURES

Volume Model Results

CMG
Description

Cases

9
A

1
10
264
1,284.0
32
142
6
6
130
88

135
41
34
4
13
5
14.2
71
3
42

5
10
56

5.27
19.00
1t .O0
9.50
27.00
2.50

Case

50.2
192.2
180.2
59.8

4.52
26.72
8.80
8.01
7.25
6.46
11.02
9.58
4.64
5.25
6.77
7.20
3.83
7.50
3.46
1.67
4.83
2.57
2.00
3.00
2.50
6.00 1

93
2

LOS

44.3
21.8
35.6
66.4
84.6
159
177.9
166.2
147.4 118.4
107.9
85.2
149.7
138.4
121.9
114.5
142.2
80.3
92.5
100.2
92.5
74.6
,

76.7
55.2 1

Direct

1 Cases 1

$1,338.91
11.0
$10,272.0)1.O
$4,345.0
1.O
$3,873.0
2.0
$5,874.5
4.0
$721.5
6.0
223.0
$222.5
$194.1 1,088.0
$342.5
40.0
$1,261.2
174.0
$6,269.9
7.0
$4,604.1
15.0
$3,898.3
130.0
$3,535.0
88.0
$2,859.3 - 135.0
$3,201.11
48.0
39.0
$2,488.0 1
$2,498.9
11.0
$1,197.4
16.0
$2,077.9
13.0
$2,047.0
25.0
$1,241.4
60.0
$2,282.5
2.0
$1,049.0
26.0
$1,1253
3.0
$1,216.8
6.0
$923.6
35.0
$798.1
58.O
1.O
$666.0 (
$1,067.51
2.O
$1.389.01
1.0 1

Cost Model Results

LOS
Case

5.3
19.0
11.0
9.5
27.0
2.5

4.0
26.7
8.8
8.0
7.3
6.5
11,O
8.8
4.6
5.3
6.8
7.2
3.4
7.5
3.5
1.7
4.8
2.6
2.0
3.0
2.5
6.0 1

50.2
192.2
180.2
59.8
44.3
21.8
35.6
66.4
84.6
159.0
177.9
166.2
147.4
118.4
107.9
85.2
149.7
138.4
121.9
114.5
142.2
80.3

02.5
100.2
92.5
74.6
76.7
55.2 1

Direct
$1,338.9
$10,272.0
$4,345.0
$3,873.0
$5,874.5
$721.5
$196.5
$159.2
$342.5
$1,159.4
$6,269.9
$4,604.1
$3,583.6
$3,249.6
$2,628.5
$2,917.7
$2,231.3
$2,498.9
$1,197.4
$2,077.9
$2,047.0
$1,141.2
$2,282.5
$1,049.0
$1,125.3
$1,216.8
$923.6
$733.7
$666.0
$1,067.5
$1.389.0

Team

l 1l
cG
:

728
734
750
756
801
803
804
805
821
DO0
Thoracic 000
25
126
127
184
187
192
250
251
252
292
294
295
296

703
904
000
028

Oncology 000

CMG
Description

LYMPHOMA AND CHRONIC LEUKEMIA

ILL-DEFINED NEOPLASM WlTH OTHE
MULTISYSTEMIC OR UNSPEClFlED S
POST-OPERATIVE AND POST-TRAUMA
WOUND DEBRIDEMENT FOR DlAGNOSl
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
COMPLICATIONS OF TREATMENT AGE
OTHER DEFINED BY MODEL
OTHER DEFINED BY MODEL
MAJOR CHEST PROCEDURES
OTHER RESPIRATORY PROCEDURES W
OTHER RESPIRATORY PROCEDURESW
MAJOR RECONSTRUCTIVE VASCULAR
OTHER VASCULAR PROCEDURES
VEIN LlGATlON AND STRIPPING
EXTENSIVE GASTROINTESTINAL PRO
GASTROSTOMY & COLOSTOMY PROCED
MAJOR ESOPHAGEAL, STOMACH & DU
G.I. OBSTRUCTION AGE 18-69 WIT
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITIS, GASTROENTERITIS &
OTHER O.R. PROCEDURES OF THE B
NON-EXTENSIVE UNRELATED O.R. P
OTHER DEFINED BY MODEL
GI PROCEDURES (ENDOSCOPY)
OTHER DEFINED BY MODEL

Volume Model Results

LOS 1 Case

Cases

I
2
2
10

13
4
2
1
392
57
17
8.6
2
27

2
269
124
498.2

3.83

2 ,

2
4
16

148.5
230.5
27.4
26.2
125.6
93.7
123.6
274.8
56.3
115
490.7
127.2
222.2
47.2
49.4

133.6
120.2
10.8
36.7
92.9

1 Cases 1

$1,3f 4.0
$1,002.0
$1,449.2
$3,599.5
$6,009.6
$1,404.8
$3,624.2
$3,996.9
$3,156.3
$149.4
$1,127.9
$4,925.8,
$1,076.3
$3,610.4
$6,922.7
$2,652.5
$943.8
$24,788.5
$3,778.0
$5,059.0
$732.0,
$553.0
$1,030.9
$12,000,0
$1,608.6
$4,763.0
$37.1
$81.5
$1,302.9

86.3

rn

3.58
10.82
1.86
6.00
13.78
7.13
1.84
47.50
11.00
13.00
3.00
3.00
3.08
31.00
2.56
24.00

77

60.2
77.2
62.3
24.6
128.8

3.00
1.O0
5.00
19.50
34.00
4.35
12.50
10.14
14.33

1
Direct

1.O
1.O
6.0
2.0
5.0

17.0
12.0
7.0
3.0
245.0
48.0
17.0
7.0
5.O
27.0
15.0
94.0
2.O
1.O
1.O
1.O
2.0
13.0
1.O
9.0
1.O
168.0
105.0
406.0

Cost Model Results

LOS
Case

3.O
1.O
5.0
19.5
34.0
4.4
12.5
10.1
14.3
2.5
10.8
1.5
6.0
9.9
7.1
1.7
42.9
11.0
13.0
3.0
3.0
3.1
31.O
2.6
24.0

2.4

60.2
77.2
62.3
24.6
128.8
86.3
148.5
230.5
27.4
26.2
125.6
93.7
123.6
274.8
56.3
115.0
490.7
127.2
222.2
47.2
49.4
133.6
120.2
10.8
36.7
92.9

Direct
$1,314.0
$1,002.0
$1,449.2
$3,599.5
$6,009.6
$1,404.8
$3,624.2
$3,996.9
$3,156.3
$137.4
$924.9
$4,039.2
$882.6
$2,9605
$5,676.7
$2,175.1
$773.9
$23,502.3
$3,778.0
$5,059.0
$732.0
$553.0
$845.3
$12,000.0
$1,319.0
$4,763.0
$30.4
$66.0
$727.6

CMGl
#

CMG
Description

Volume Mode1 Results

1 Los 1

Cases

ORBITAL PROCEDURES
RADICAL LARYNGECTOMY AND GLOSS

076

MAJOR HEAD AND NECK PROCEDURES

077
081
084
087
088
089
125
126
127
131
187
250

LESS EXTENSIVE HEAD AND NECK P

SALIVARY GLAND PROCEDURES
MISCELLANEOUS AR, NOSE AND TH
SINUS PROCEDURES
ETHMOIDECTOMY (MNRH)
DENTAL EX TRACTION/RESTORATION
MAJOR CHEST PROCEOURES
OTHER RESPIRATORY PROCEDURES W
OTHER RESPIRATORY PROCEDURES W
RESPIRATORY NEOPLASMS
OTHER VASCULAR PROCEDURES
EXTENSIVE GASTROINTEST~NAL-PRO

252
254
255
256
257
258
267
268
280

MAJOR ESOPHAGEAL, STOMACH & DU

MAJOR INTESTINAL AND RECTAL PR
MAJOR INTESTINAL AND RECTAL PR
LESS EXTENSIVE ESOPHAGEAL, STO
LESS EXTENSIVE ESOPHAGEAL,STOM
LESS EXTENSIVE ESOPHAGEAL, STO
LAPAROTOMYWlTH CC
ANUS AND STOMAL PROCEDURES, AG
ANUS AND STOMAL PROCEDURES, AG
DIGESTIVE SYSTEM MALIGNANCY WI

7
57
43

300
311
,312
1314

OTHER G.I. DIAGNOSES AGE > 18

MAJOR PANCREATIC PROCEDURES
MAJOR HEPATOBILIARY PROCEDURES
IOTHER
HEPATOBlLlARY AND PANCRE

253

5
3
13
11

110.9
525.5
309.4
123.7
140.6
62.9
161
68.7
69.2
217.1
97.2
144.8
25.8
91.1
402.7

8
2
43

12.50
12.22
8.90
20.00
22.00
5.00
12.50
2.80
3.33
10.31

241.4
199.9
206.6
153.2
223.2
210.2
54.4
88.6
88,5
57.5

$5,231.81
$4,414.61
$3,712.5
$9,002.0
$5,355.5
$3,662.7
$2,721.9
$1,148.0
$1,273.0
$2,586.6

18
4
8
10 1

5.46
47.90
11.67
11.56 1

59.3
365.9
424.9
162.8 1

$1,131.l
$22,961.5
$5,247.8
$3,229.1 1

15
3

4
2
40
31
9

70
5
2

Cost Model Resull

Direct 1 Cases 1
Cost
1
$942.4 -1
4.0
8.0
$16,047.4
34.0
$4,474.7
14.0
$1,574.8
$1,170.8
35.0
9.O
$627.3
8. O
$1,548.6
$577.1
2.0
$657.4
1.O
$4,299.9
102.0
$1,310.6
26.0
24.0
$4,113.9
$2,239.6
59.0
$1.446.5
3.0
4.0
$12,987.1

2.00
35.38
12.30
5.07
2.52
1.33
3.88
1.O0
1.O0
9.45
3.04
10.79
9.15
4.67
24.00-

42

TipPe

050
075

Case

Los

Case

2.0
35.4

110.9
525.5

10.8

309.4

4.5
2.2
1.3
3.9
0.9
1.O
8.7
2.7
9.5
9.2
4.7
24.0

123.7
140.6
62.9
161.0
68.7
69.2
217.1
97.2
144.8
25.8
91.1
402.7

4.0
57.0
43.0
1.O
1.O
1.O
4.0
5.0
3.0
36.0

12.5
12.2
8.9
20.0
22.0
5.0
10.8
2.4
2.9
8.9

241.4
199.9
206.6
153.2
223.2
210.2
54.4
88.6
88.5
57.5

10.0
21.0
9.0

28.3
10.2
10.0 1

365.9
424.9
162.8

Direct

Cost
$866.3

Team

I l
324
350
352
353
354
355
356
357
361
362
366

368
370
371
372
374
377
378
379
380
383
391
428
432
433
437
438
477
479
,504
508

CMG
Description

PANCREATIC CANCER OR OTHER MAL

MULTIPLE OR BILATERAL JOINT RE
HIP REPLACEMENT WlTH CC
HIP REPLACEMENT WlTHOUT CC
KNEE REPLACEMENT
REATTACHMENT PROCEDURES OR LOW
FRACTURED FEMUR PROCEDURES WIT
FRACTURED FEMUR PROCEDURES WIT
MUSCULOSKELETAL BlOPSY FOR MAL
MUSCULOSKELETAL 6iOPSY WITHOUT
BACK AND NECK PROCEDURES WlTHO
MAJOR HIP AND KNEE PROCEDURES
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR LOWER EXTREMITY PROCEDUR
MAJOR UPPER EXTREMIN PROCEDUR
MINOR LOWER EXTREMITY PROCEDUR
WOUND DEBRIDEMENT AND SKlN GRA
SOFT TISSUE PROCEDURES (MNRH)
OTHER MUSCULOSKELETAL PROCEDUR
OTHER LOWER EXTREMITY PROCEDUR
JOINT REPLACEMENT FOR MALlGNAN
SECONOARY NEOPLASMS AND PATHOL
BREAST PROCEDURES EXCEPT BlOPS
SUBTOTAL MASTECTOMY & OTHER BR
SUBTOTAL MASTECTOMY & OTHER BR
OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES
PARATHYROlD PROCEDURES
THYROID PROCEDURES
MAJOR URINARY TRACT PROCEDURES
MINOR UPPER URINARY TRACT PROC

Volume Model Results

LOS
Case

Cases

36

3
5
6
6
29
37
4
12
23
183

13.33
2.00
29.50
5.00
8.00
15.77
14.50
11.50
7.66
3.85
10.00
10.85
19.50
7.63
11.23
7.77
12.81
5.24
4.78
3.00
15.03
15.74
0.94

165

1.64

101
50
47

1.93
3.00
3.38
3.11
2.56
7.00
48.00

1
1
6
3
-

2
46

47 ,

10

1
-

22

129

T
m
15.8
368.2
227.7
163.2
139.8
253.8
316.4
205.3
73.3
71.2
375.2
122
326.9
180.2
241
162.7
196.4
120
110.3
80.2
202.5
27.9
57.1
93.4
100.5
79.2
97.1
133.9

137.1
178.6

1
Direct

1 Cases 1

Cost
$2,414.6
$3,302.2
$5,236.0
$3,224.1
$3,759.8
$6,149.0
$4,022.6
$3,132.3 $2,332.9

$1,231.4
$3,552.7
$2,344.2
$4,566.5
$2,221$5
$3,519.9
$1,940.0
$2,967.7
$1,965.6
$1,747.9
$1,226.8
$6,775.0
$2,640.8

$535.2
$908.6
$977.9
$1,020.8
$1,206.1
$1,411.4
$1,324.4
$2,670.0)
$6,678.1 l

30.0
1.O
2.0
2.0
3.0
17.0
8.0
6.0
46.0
47.0
1.O
13.0
4.0
8.0
13.0
17.0
16.0
36.0
46.0
3.0
31.0
19.0
114.0
103.0
63.0
31.0
29.0
18.0

109.0
19.0
2.0

Cost Model Resulb

LOS
Case

11.4
2.0
26.1
4.5
7.1
15.8
12.5
1O. 1

7.7
3.9
10.0
9.5
12.3
6.2
9.4
5.6
9.7
4.5
4.0
2.5
13.1
13.7
0.9 .
1.5
1.8
2.6
3.0
2.7
2.6
6.2
48.0

15.8
368.2
227.7
163.2
139.8
253.8
316.4
205.3
73.3
71.2
375.1
122.0
326.9
180.2
241.O
162.7
196.4
120.0
110.3
80.2

202.5
27.9
57.1
93.4
100.5
79.2
97.1
133.9
137.1
178.6

Direct
$2,219.7
$3,035.6
$4,813.4
$2,963.9
$3,456.3
$5,652.7
$3,697.9
$2,879.4
$2,144.6
$1,132.0
$2,399.9
$2,155.0
$3,392.6
$1,953.5
$3,198.1
$1,569.9
$2,474.2
$1,806.9
$1,590.6
$1,127.8
$6,228.1
$2,427.7
$492.0
$835.3
$898.9
$938.4
$1,108.8
$1,207.5
$1,217.5
$2,454.5
$4.456.5

CMG
#
510
513
578
,579
580
703
711
725
727
728

729
731
732
734
735
736
750
756
801
803
,804
805
821
900
904
DO0
004
DO8
028

035
D61

CMG
Description

TRANSURETHRAL PROSTATECTOMY WI
OTHER TRANSURETHRAL OR BIOPSY
MAJOR GYNECOLOGICAL PROCEDURES
MAJOR GYNECOLOGICAL PROCEDURES
MAJOR GYNECOLOGICAL PROCEDURES
OTHER O.R. PROCEDURES OF THE B
RETICULOENDOTHELIAL AND IMMUN1
MAJOR LEUKEMJA AND LYMPHOMA PR
ACUTE LEUKEMIA WITHOUT MAJOR P
LYMPHOMA AND CHRONIC LEUKEMIA
LYMPHOMA AND CHRONIC LEUKEMIA
LYMPHOMA AND CHRONIC LEUKEMIA
LYMPHOMA AND CHRONJC LEUKEMIA
ILL-DEFINED NEOPLASM WlTH OTHE
RADIATION THERAPY
,CHEMOTHERAPY
MULTlSYSTEMlC OR UNSPECIFIED S
POST-OPERATIVEAND POST-TRAUMA
WOUND DEBRIDEMENT FOR DIAGNOSi
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
COMPLICATIONS OF TREATMENT AGE
EXTENSIVE UNRELATED O.R. PROCE
1NON-EXTENSIVE UNRECATEO O.R. P
OTHER DEFINED BY MODEL
ORBITAL AND OTHER EYE PROCEDUR
EXTERNAL EYE PROCEDURES
GI PROCEDURES (ENDOSCOPY)
BLADDER & URETHRAL PROCEDURES
BIOPSY

Volume Model Results

Cases

LOS

15
66
29

32 ..
90
1
48
11
4

26
10
19
12
9

104
317

5
2
2
2
7
3
516
15
57
631.2
115.2
2

Direct
Cost
$979.8
$662.1
$1,440.3
$1,406.5
$1,132.5
$2,065.7
$2,592.0
$4,743.1
$7,860.1

Case

68.3

2.89
1.50
4.92
4.04
2.60
4.00
7-57
10.82
26.40

56.9

120.9
132.8
125.5
147.7
2.8
149

3.62
20.85
14.83
17.08
8.74
3.02
2.84
44.00
7.33
3.00
7.00
12.80
10.00
22.00
14.00
22.29

69

$1,492.7

91.3
10.4

$5,190,9
$3,442.2
$4,184.5
$2,773.2
$2,036.2
$1,037.9
$18,733.9
$1,412.3
$739.5
$1,807.2
$3,821.2
$2,415.5
$3,815.1
$3,804.8
$6,667.6
$201.3
$263.4
$306.7
$97.5
$210.6
$1 15.1

90.7

165.7
20.1
43.2
105.2
233.8
190.2
60.2
42.5
71.7
36.4
38.9
53.9
44.5
25.5
40

Cases
9.0
41.O
24.0
27.0
56.0
4.0
40.0
28.0
5.0

26.0
20.0
23.0
13.0
23.0
65.0
198.0
2.0
3.0
1.O
1.O
5.0

1.O
1.O
4.0
7.0
322.0
9.0
48.0
428.0
88.0
1.O

Cost Model Results

LOS
Case
2.5
1.5
4.3
3.6
2.6
4.0
7.6
9.4
26.4
3.0
17.5
12.3
14.5
7.5
3.0
2.6
44.0
6.3
3. O

6.1
12.8
10.0
22.0
14.0
22.3

68.3
56.9
120.9
132.8
125.5
147.7
2.8
149.0
69.0
91.3

10.4
90.7

165.7
20.1
43.2
105.2
233.8
190.2
60.2
42.5
71.7
36.4
38.9
53.9
44.5
25.5
40.0

Direct
$900.7
$608.6
$1,324.1
$1,292.9
$1,041.1
$1,898.9,
$2,382.8
$4,360.2
$7,225.7
$1,353.8
$4,771.9
$3,164.3
$3,846.8
$2,549.3
$1,871.9
$954.1
$13,815.9
$1,298.3
$679.8
$1,661.3
$3,512.7
$2,220.6
$2,953.1
$3,497.7
$6,120.4
$185.1
$215.2
$282.0
$60.9
$193.6
$105.8

CMG
Description

Cases

Volume Model Results

LOS
Case

Direct

1 Cases

Cost h ode1 Results

LOS

Case

Direct

D63
Dental 000
E Y ~ 050
ENT
052
055
056
057
076
077
084
087
089
091

126
1

127
257
294
295
437
438
477
479
703
729

803
804
805
904
DO0

TRANSFUSIONS
OTHER DEFINED BY MODEL
ORBITAL PROCEDURES
RETINAL PROCEDURES
LENS INSERTION WITH CC (MNRH)
LENS INSERTION WITHOUT CC (MNR
OTHER OPHTHALMIC PROCEDURES (M
MAJOR HEAD AND NECK PROCEDURES
LESS EXTENSIVE HEAD AND NECK P
MlSCELiANEOUS EAR, NOSE AND TH

(DENTAL EXTRACTIONIRESTORATION
NASAL PROCEDURES (MNRH)
OTHER RESPIRATORY PROCEDURES W
OTHER RESPIRATORY PROCEDURES W
LESS EXTENSIVE ESOPHAGEAL, STO
ESOPHAGITIS, GASTROENTERITIS &
ESOPHAGITIS, GASTROENTERITIS &
OTHER DERMATOLOGIC PROCEDURES
OTHER DERMATOLOGIC PROCEDURES
PARATHYROID PROCEDURES
"TYROID
PROCEDURES
OTHER O.R. PROCEDURES OF THE B
LYMPHOMA AND CHRONIC LEUKEMIA
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
OTHER PROCEDURES FOR DlAGNOSlS
NON-EXTENSIVE UNRELATED O.R. P
OTHER DEFINED BY MODEL

1.96
1.72
1.37
1.22
1.O2
1.O8
3.33
250

124.4
103.1

24

1.33
1.43

50.9
70.7

1.30 1
1.09
1.33
4.00
4.00
3.00
10.00
1.20
6.00
2.83
2.46
2.00
3.00
1.10
2.56
3.25
1.O0

120.3 1
66.7

SINUS PROCEDURES

148
50
66.2
62.5
85.3
72.5

200
390
52
123
68
227
157
8
11

22 1
118
4
5
2
4

6
2
1247
2
2
16
15
7
2
129

68.1

48.3
58.3
94.3
16.1
34.6
58.3
55.4
118.3
109.9
110.8
114.3
56.3
107.2
95.8
50.3
50.7

$264.4
$717.8
$706.9
$582.3
$624.4
$513.0
$642.5
$1,604.0
$906.8
$620.21
$731.01
$887.1 1
$704.5
$902.0
$938.7
$1,196.0
$571.5
$1,640.0
$686.2
$1,717.3
$1,372.8 $1,230.4
$1,169.3
$1,984.0
$468.5
$1,155.1
$1,051.3
$176.0
$212.5

169.0
330.0
64.0
150.0
84.0
278.0
192.0
6.0
14.0

1.4

1.7
1.4
1.2
1.O
0.9
-

3.3
2.5

48.0
30.0

1.3
1.4

56.0 1
144.0
6.O
3.0
1.O
2.0
2.0
5.0
6.0
30.0
119.0
6.0
1.0
10.0
9.0
4.0
1.O
158.0

1.3

1.1
1.3
4.0
4.0
3.0
10.0
1.2
6.0
2.8
2.5
2.0
3.0
1.1
2.6
3.3
1.O

148.0
50.0

$243.1
$588.6

LVNLUAI

IMHbL

TEST TARGET (QA-3)

APPLIED

I W G E . lnc

-.
----,

1653 East Main Street

Rochester. NY 14609 USA
Phone: 7W482-0300
Fax: i l W88-5989

O 1993. Applied Image. lm.. All Rihts Reserwd