Duration of Postoperative Mechanical

Ventilation as a Quality Metric for Pediatric

Cardiac Surgical Programs
Michael Gaies, MD, MPH,* David K. Werho, MD,* Wenying Zhang, MS,
Janet E. Donohue, MPH, Sarah Tabbutt, MD, PhD, Nancy S. Ghanayem, MD,
Mark A. Scheurer, MD, MS, John M. Costello, MD, MPH, J. William Gaynor, MD,
Sara K. Pasquali, MD, MHS, Justin B. Dimick, MD, Mousumi Banerjee, PhD, and
Steven M. Schwartz, MD
Division of Cardiology, Department of Pediatrics, C.S. Mott Children’s Hospital and Center for Healthcare Outcomes & Policy,
University of Michigan, Ann Arbor, Michigan; Division of Critical Care Medicine, Department of Pediatrics, Lucile Packard Children’s
Hospital, Stanford University, Palo Alto, California; Michigan Congenital Heart Outcomes Research and Discovery Unit, University of
Michigan, Ann Arbor, Michigan; Division of Critical Care Medicine, Department of Pediatrics, Benioff Children’s Hospital, University
of California San Francisco, San Francisco, California; Division of Critical Care Medicine, Department of Pediatrics, Texas Children’s
Hospital, Baylor College of Medicine, Houston, Texas; Division of Cardiology, Department of Pediatrics, Medical University of South
Carolina, Charleston, South Carolina; Division of Cardiology, Department of Pediatrics, Ann & Robert H. Lurie Children’s Hospital of
Chicago, Northwestern University Feinberg School of Medicine, Chicago, Illinois; Division of Cardiac Surgery, The Children’s Hospital
of Philadelphia, Philadelphia, Pennsylvania; Department of Surgery and Department of Biostatistics, School of Public Health,
University of Michigan, Ann Arbor, Michigan; and Departments of Critical Care Medicine and Paediatrics, The Hospital for Sick
Children, University of Toronto School of Medicine, Toronto, Ontario, Canada

Background. Few metrics exist to assess quality of care Results. Overall, 3,108 patients (65.6%) received
at pediatric cardiac surgical programs, limiting opportu- POMV; the remainder were extubated intraoperatively.
nities for benchmarking and quality improvement. Post- Our model was well calibrated across groups; neonatal
operative duration of mechanical ventilation (POMV) age had the largest effect on predicted POMV. These
may be an important quality metric because of its asso- comparisons suggested clinically and statistically impor-
ciation with complications and resource utilization. In tant variation in POMV duration across centers with a
this study we modelled case-mix–adjusted POMV dura- threefold difference observed in O/E ratios (0.6 to 1.7).
tion and explored hospital performance across POMV We identified 1 hospital with better-than-expected
metrics. and 3 hospitals with worse-than-expected performance
Methods. This study used the Pediatric Cardiac Crit- (p < 0.05) based on the O/E ratio.
ical Care Consortium clinical registry to analyze 4,739 Conclusions. We developed a novel case-mix–adjusted
hospitalizations from 15 hospitals (October 2013 to model to predict POMV duration after congenital heart
August 2015). All patients admitted to pediatric cardiac operations. We report variation across hospitals on metrics
intensive care units after an index cardiac operation of O/E duration of POMV that may be suitable for bench-
were included. We fitted a model to predict duration of marking quality of care. Identifying high-performing
POMV accounting for patient characteristics. Robust centers and practices that safely limit the duration of
estimates of SEs were obtained using bootstrap POMV could stimulate quality improvement efforts.
resampling. We created performance metrics based
on observed-to-expected (O/E) POMV to compare (Ann Thorac Surg 2017;-:-–-)
hospitals. Ó 2017 by The Society of Thoracic Surgeons

Q uality assessment in pediatric cardiac operations

suffers from a paucity of valuable metrics suitable
for benchmarking. Death remains the predominant
Dr Dimick discloses a financial relationship with
metric by which hospitals are compared, but many
ArborMetrix, Inc.
stakeholders have called for development of new quality

Accepted for publication June 6, 2017.

The Appendix and Supplemental Table can be viewed
*Drs Gaies and Werho are co-first authors and contributed equally to this in the online version of this article [http://dx.doi.
work. org/10.1016/j.athoracsur.2017.06.027] on http://www.
Address correspondence to Dr Gaies, Mott Children’s Hospital, 1540 E annalsthoracicsurgery.org.
Hospital Dr, Ann Arbor, MI 48109; email: [email protected].

Ó 2017 by The Society of Thoracic Surgeons 0003-4975/$36.00

Published by Elsevier Inc. http://dx.doi.org/10.1016/j.athoracsur.2017.06.027
2 GAIES ET AL Ann Thorac Surg
POSTOPERATIVE VENTILATION METRICS 2017;-:-–-

measures that assess morbidity and resource utilization Cardiac Code [19] and The Society of Thoracic Surgeons
[1, 2]. Duration of postoperative mechanical ventilation Congenital Heart Surgery Database, as previously
(POMV) may be one such important measure of quality described [20]; each site ensures that these data match
care after pediatric cardiac operations. Although many across databases. Trained data managers, who pass a
children and adults undergoing operations for congenital certification examination, abstract cases. We previously
or acquired heart disease wean from mechanical venti- published results of the rigorous audit process demon-
lation early in the postoperative period, prolonged strating excellent data integrity within the registry [21].
ventilation may be necessary for complex patients [3, 4]. Submission of data to PC4 is considered quality
Data suggest that mechanical ventilation poses increased improvement activity and is not subject to ongoing
risk for infection, airway and lung injury, and failed Institutional Review Board oversight. The University of
extubation [3, 5, 6]. Increased duration of POMV prolongs Michigan Institutional Review Board oversees the PC4
the critical care period and further exposes patients to Data Coordinating Center; this study was approved with
potential risks [4, 7, 8], so it is reasonable to conclude that waiver of informed consent.
shortening POMV duration may represent higher-
quality, more cost-effective care. The National Quality Inclusion and Exclusion Criteria
Forum endorsed an adjusted prolonged mechanical We considered all surgical hospitalizations for patients
ventilation hospital quality metric for coronary artery admitted to PC4 hospitals that included at least one CICU
bypass operations [9]. encounter between October 1, 2013, and August 31, 2015.
However, the heterogeneity of pediatric cardiovascular Each hospitalization included an “index” surgical pro-
surgical procedures and case-mix differences across cedure as defined by The Society of Thoracic Surgeons
hospitals [10] create challenges in assessing duration of [22]. Although the unit of analysis is the surgical hospi-
POMV as a measure of hospital performance. An ideal talization, we refer to these episodes as “patients”
quality metric of POMV would account for the patient throughout this report. We excluded patients with a tra-
factors and operative complexity that differ across hos- cheostomy in situ at the index operation, who weighed
pitals. Previous investigators used this approach to mea- less than 2.5 kg undergoing isolated patent ductus arte-
sure ventilation time for critically ill adults [11–13], but riosus repair, who underwent index operations not
these models do not apply to children with congenital classifiable into one of The Society of Thoracic Surgeons–
heart disease. Some factors associated with POMV European Association for Cardiothoracic Surgery (STAT)
duration in children with cardiac disease are known, but mortality categories [23], or who died before post-
no validated tool exists to predict the duration of POMV operative day 7.
in this population [3, 4, 14–18]. Metrics of adjusted POMV
duration would allow hospitals to measure performance Model Development: Outcomes and Predictor Variables
and benchmark against peer institutions. These data We developed a case-mix–adjustment model to predict
could motivate improvement initiatives by collaborative duration of POMV after a congenital cardiac operation.
learning with high-performing centers. Total duration of POMV was the dependent variable
In this context, we aimed to develop hospital perfor- in the model. We calculated duration of POMV by adding
mance metrics for duration of POMV after pediatric car- the length of each episode of mechanical ventilation after
diac operations using data from the Pediatric Cardiac the index operation. The registry includes the exact start
Critical Care Consortium (PC4) clinical registry. We and stop time of each mechanical ventilation episode in
developed and validated a model to predict duration of the CICU, including episodes in other critical care areas.
POMV and then created several metrics based on the Patients extubated in the operating room after an index
observed-to-expected (O/E) duration of POMV. We procedure who never require mechanical ventilation
describe differences in performance across hospitals for were included with a POMV duration of 0 hours. Dura-
each metric and in aggregate. tion of POMV was capped at 60 days to eliminate the
effect of extreme outliers on the model.
Our aim was to create a population-level case-mix–
Patients and Methods adjustment model. We did not attempt to predict dura-
tion of ventilation at the patient level. Thus, we explored
Data Source candidate variables for our model present preoperatively
PC4 is a voluntary quality improvement collaborative or intraoperatively. We excluded postoperative variables
among hospitals across North America, and 15 hospitals that may be important for predicting the duration of
were submitting data at the time of the analysis. All car- mechanical ventilation for an individual patient such as
diac intensive care unit (CICU) encounters at participant vasoactive-inotropic score, complications, and critical
hospitals are submitted to the PC4 clinical registry. Each care therapies. These variables could also reflect surgical
case record includes demographics, patient comorbid- or critical care quality and would therefore not be
ities, data on cardiac surgical procedures and other in- appropriate for a case-mix–adjustment model used in
terventions, critical care therapies, and complications, all hospital quality assessment.
with standardized definitions. Preoperative patient factors, comorbidities, and opera-
The registry shares common terminology and defini- tive complexity variables that affect outcomes after
tions with the International Pediatric and Congenital congenital cardiac operations were selected a priori.
Ann Thorac Surg GAIES ET AL 3
2017;-:-–- POSTOPERATIVE VENTILATION METRICS

had POMV of less than 24 hours, we calculated this

pseudo-C statistic for patients within subgroups of ex-
pected POMV of 24 or less hours or more than 24 hours.

Creating Metrics of Hospital Performance Using

O/E Duration of POMV
From the final model, we calculated O/E ratios of dura-
tion of POMV for each center. The O/E ratio was calcu-
lated by dividing total observed hours of mechanical
ventilation by total expected hours. Expected hours were
derived for each patient from our multivariable model.
For this analysis, we excluded any patient with an ex-
pected duration of POMV exceeding 168 hours (w10% of
the entire sample) because the model did not fit as well at
Fig 1. Plot of the mean expected duration of postoperative mechan- the upper tail of the distribution (Fig 1). We empirically
ical ventilation (POMV) in equal rank-ordered groups (red squares) derived the 95% CI around the O/E ratio from our boot-
vs mean observed duration within the groups (black triangles). strap resampling. Statistically better-than-expected or
worse-than-expected performance was defined as an
O/E ratio of 1 or less, respectively, with the 95% CI not
Patients were categorized as neonate (<30 days), infant
crossing 1.
(30 to 365 days), child (>365 days to 18 years), or adult
Although the O/E ratio is a clear performance metric,
(>18 years). Other demographic variables included sex,
we explored alternative ways to express the data that
prematurity (<37 weeks gestational age), and the presence
might have value to clinicians and quality researchers.
of any extracardiac or chromosomal anomalies, syn-
We created 4 additional metrics based on O/E duration of
dromes, or airway anomalies [19]. We recorded preoper-
POMV (Table 1). Expected duration of POMV for each
ative mechanical ventilation only if it was present on CICU
metric was determined from the model as described.
admission. STAT category and cardiopulmonary bypass
Total days lost/saved per 100 patients is the O/E differ-
times were included as measures of surgical complexity.
ence rather than ratio, which may be more easily inter-
Model Derivation and Validation pretable for quality improvement initiatives. We took the
difference of total days observed and total days expected
We used zero-inflated negative binomial regression to
at each hospital and normalized the difference to per 100
model duration of POMV as a function of patient char-
hospitalizations. If observed was less than expected (dif-
acteristics and operative complexity. As determined by
ference <0), we considered those days of mechanical
the generalized information criterion, zero-inflated nega-
ventilation “saved.” If the difference was more than 0, we
tive binomial regression fit the data better than Poisson,
considered those days “lost.” We then ranked each hos-
negative binomial, or zero-inflated Poisson models. Var-
pital on every metric (1 ¼ best, 13 ¼ worst). Two hospitals
iables associated with duration of POMV at p of less than
that had recently joined the collaborative were dropped
0.05 were retained in the final model. We derived bias-
for these analyses because of low case numbers (n < 50).
corrected confidence intervals (CIs) for the regression
Finally, we summed the ranks across each metric to
coefficients based on 1,000 bootstrap resamples.
determine an average hospital rank, weighing each
To test goodness of fit, we separated the patients into
metric equally.
equal rank-ordered groups (237 patients per group) and
plotted the mean observed duration of mechanical
ventilation within each group against the mean duration
Results
predicted by the model. Model discrimination was
assessed using a pseudo-C statistic (see details of the The study included 4,739 hospitalizations at 15 partici-
method in the Appendix). Because most of the patients pating CICUs. Overall, 3,108 patients (66%) required

Table 1. Metrics of Hospital Performance on Duration of Postoperative Mechanical Ventilation

Metric Numerator Denominator Unit

Early extubation success Hospitalizations with observed Hospitalizations with predicted %

MV <6 hours MV >12 hours
Early extubation failure Hospitalizations with observed Hospitalizations with predicted %
MV >12 hours MV <6 hours
Positive deviance in Hospitalizations with observed All hospitalizations %
reduction of MV MV <50% predicted duration
Total days of MV lost/ (total days observed – total days All hospitalizations with Days lost/saved per 100
saved predicted) predicted MV <7 days hospitalizations

MV ¼ mechanical ventilation.
4 GAIES ET AL Ann Thorac Surg
POSTOPERATIVE VENTILATION METRICS 2017;-:-–-

Table 2. Multivariable Model Predicting Duration of Postoperative Mechanical Ventilation (N ¼ 4,739)

Characteristic Frequency (%) or Median (IQR) IRR (95% CI)a p Valueb

Male gender 2,602 (55) 1.10 (0.92–1.30) 0.046

Age group
Neonate 1,007 (21) 3.81 (2.87–4.79) <0.001
Infant 1,528 (32) 1.87 (1.40–2.38) <0.001
Child 1,925 (41) Reference group
Adult 279 (6) 1.35 (0.60–2.57) 0.014
Weight category
Underweightc 1,070 (23) 1.19 (0.96–1.49) 0.005
Normal weight 3,515 (74) Reference group
Overweightd 159 (3) 0.52 (0.35–0.76) <0.001
Prematurity (<37 weeks) 509 (11) 1.26 (1.02–1.57) 0.002
Airway anomaly 145 (3) 2.76 (2.03–3.61) <0.001
Extracardiac or genetic anomaly 1,307 (28) 1.44 (1.19–1.73) <0.001
Preoperative mechanical ventilation 355 (7) 1.88 (1.45–2.39) <0.001
STAT category 1–3 3,595 (76) Reference group
STAT category 4–5 1,030 (22) 1.96 (1.62–2.39) <0.001
CPB time, min 78 (43–124) 1.00 (1.00–1.01) <0.001
a b c
Bias-corrected CIs derived from 1,000 bootstrap resamples. The p values are from zero-inflated negative binomial regression. Weight-for-age z
score <–2 (age 20 years old) or body mass index <18.5 kg/m2 (age >20 years old). d
Weight-for-age z score >2 (age 20 years old) or body mass index
>30 kg/m (age >20 years old).
2

CI ¼ confidence interval; CPB ¼ cardiopulmonary bypass; IQR ¼ interquartile range; IRR ¼ incidence rate ratios; STAT ¼ Society of
Thoracic Surgeons-European Association for Cardiothoracic Surgery.

POMV; of these, 53% were infants, and 22% had a high- performance (O/E ratio, 0.59; 95% CI, 0.4 to 0.8), and 3
complexity operation (STAT categories 4 or 5). The me- hospitals had worse-than-expected performance (O/E > 1).
dian duration of POMV was 0.3 days (10th to 90th Figures 3, 4, and 5 demonstrate the variation in hospital
percentile range, 0 to 5.9 days). The patient characteristics performance on each metric of the adjusted POMV
and all candidate predictor variables are reported in described in Table 1. For the metric of total days lost –
Supplemental Table 1. saved per 100 hospitalizations (Fig 5), the top-ranked
The 15 hospitals contributing data to this analysis hospital saved 173 days of POMV compared with the
perform the full spectrum of congenital heart operations lowest-ranked hospital. The aggregate rankings for each
(complex neonatal repairs, STAT 5 procedures), and the hospital on all 4 metrics are reported in Table 3 (1 ¼ best).
range of annual index operations across the hospitals is
206 to 601 per year. Every hospital has a dedicated CICU,
and all but 4 have a 24/7 intensivist coverage model.

Model Performance
Table 2 reports the final predictor variables in our case-
mix–adjustment model. Independent predictors of longer
duration of POMV were age, prematurity, extracardiac/
genetic anomalies, underweight, preoperative mechanical
ventilation, higher STAT category, and cardiopulmonary
bypass time. Neonate status had the largest effect on
duration of POMV (incidence rate ratio of 3.8 vs child;
95% CI, 2.9 to 4.8). The goodness-of-fit plot showed a
well-fitted model predicting mean duration of POMV for
groups of patients (Fig 1). The pseudo-C statistic for this
model was 0.73 for patients with expected duration of
POMV of less than 24 hours and 0.63 for patients with
expected duration exceeding 24 hours.
Fig 2. Observed-to-expected (O/E) ratio of duration of postoperative
mechanical ventilation (POMV). Hospitals are rank ordered with
Hospital Performance on Metrics of Case-Mix–Adjusted
A ¼ lowest O/E and M ¼ highest. Hospital A has statistically
Duration of POMV significant less-than-expected duration of POMV, and hospitals K
Figure 2 shows O/E ratios of duration of POMV for each through M have significantly greater-than-expected POMV. The
hospital (n ¼ 13). One hospital had better-than-expected range bars show the 95% confidence interval.
Ann Thorac Surg GAIES ET AL 5
2017;-:-–- POSTOPERATIVE VENTILATION METRICS

Fig 3. Percentage of patients with expected postoperative mechanical

ventilation (POMV) >12 hours who had observed POMV <6 hours
(success; black), and those with an expected duration of POMV
<6 hours who had observed POMV >12 hours (failure; red).
Fig 5. Total days of postoperative mechanical ventilation (POMV)
saved (black) or lost (red) at each center, per 100 hospitalizations.
The average rankings suggested 2 positive outliers Hospitals are rank ordered with A ¼ best performance (days saved)
and M ¼ worst performance (days lost). *Denotes hospitals with
(average rank, 1.75), a middle group (average rank, 4.5 to
statistically significant days saved/lost (p < 0.05).
7.25), and 5 hospitals in a lower group (average rank, 8.25
to 10.75) and ranking more than 10 on 2 or more metrics.
some cases, early extubation is not feasible as a result of
unique microsystem features of each hospital, such as
need for operating room turnover or ICU staffing [5, 24],
Comment
although we can only speculate on whether that is more
We developed a case-mix–adjusted model to predict or less true among hospitals in this cohort. These per-
duration of POMV in pediatric cardiac surgical patients formance metrics may reflect those practice differences
using a parsimonious set of easily collected predictor rather than differences in quality of care. Further, un-
variables. The model was well calibrated to predict the derstanding how adjusted duration of POMV correlates
duration of POMV across groups of patients, a charac- with other resource utilization, such as adjusted intensive
teristic that aligns with our aim to use this model to assess care and hospital length of stay, will be important.
hospital performance. We also created a novel set of Despite these caveats, our POMV metrics could provide
quality metrics based on O/E duration of POMV. Analysis actionable information to hospitals for determining
of hospital performance on these metrics suggested whether improvement initiatives are warranted.
clinically important variation across 13 hospitals. To our Mahle and colleagues [25], from the Pediatric Heart
knowledge, this is the first such analysis in pediatric Network, demonstrated the value of understanding
cardiac surgical outcomes research. duration of POMV for quality improvement efforts in
Efforts to reduce duration of POMV probably represent pediatric cardiac operations. These investigators suc-
high-quality care as long as this is accomplished with low cessfully reduced POMV after two infant cardiac surgical
reintubation rates, avoidance of periextubation compli- procedures through collaborative learning. A critical
cations, such as cardiac arrest, and a low mortality rate. aspect of this process involved presenting POMV metrics
Strategies for POMV practices differ across hospitals; in to the participating study teams. Comparative reports—
although not adjusted—across the hospitals helped to
identify positive and negative outliers, and the key prac-
tices at higher-performing hospitals were identified and
implemented at hospitals with a longer duration of
POMV.
This prior experience illustrates how we hope PC4
hospitals will use the metrics developed in the current
study. The rigorous case-mix–adjustment method allows
meaningful cross-hospital comparison. Transparent data
sharing within the PC4 collaborative facilitates identifi-
cation of top performers and discussions about practice
differences that could lead to collaborative learning ap-
proaches to reduce the duration of POMV across a wider
patient population [25, 26].
Fig 4. Percentage of patients by hospital whose observed duration of We created metrics that assess performance of the
postoperative mechanical ventilation (POMV) was <50% of the entire perioperative team and, therefore, only included
expected duration of POMV. quality-independent preoperative factors. Determining
6 GAIES ET AL Ann Thorac Surg
POSTOPERATIVE VENTILATION METRICS 2017;-:-–-

Table 3. Aggregate Hospital Ranking Across Metricsa

Center EE Success EE Failure POMV Reduction Days POMV Saved/Lost Average Rank

A 3 1 2 1 1.75
C 1 2 1 3 1.75
F 2 1 9 6 4.5
D 7 5 3 4 4.75
H 4 1 6 8 4.75
E 5 4 10 5 6
J 6 7 4 10 6.75
B 13 9 5 2 7.25
G 12 1 13 7 8.25
K 9 6 8 11 8.5
M 8 3 11 13 8.75
I 10 10 7 9 9
L 11 8 12 12 10.75
a
By metric ranking, 1 ¼ best and 13 ¼ worst. Hospitals are designated A to M to correspond with hospital designations in Figures 2 to 5.
EE ¼ early extubation; POMV ¼ postoperative mechanical ventilation.

the relative contributions of surgical and CICU care to Finally, other potential metrics of hospital quality
adjusted duration of POMV metrics represents the next related to limiting POMV may exist [16, 28]. Although the
key step in this investigation. To do so will require the quality metrics we evaluated provide insight into hospital
addition of postoperative severity of illness variables to performance, we analyzed these out of context to other
the model to isolate and measure the quality of CICU care important quality indicators.
[4, 15, 17]. Assessing the drivers of prolonged POMV— In conclusion, we developed a case-mix–adjusted
surgical vs CICU quality—will help hospitals to identify model to predict duration of POMV after congenital
the key levers for improvement. heart operations, accounting for baseline patient and
Our analytic strategy works well for assessing the operative characteristics. We used this model to derive
duration of POMV for groups of pediatric cardiac surgical several novel metrics of hospital performance pertaining
patients and for hospital quality assessment. The model to the duration of POMV and highlight differences
fit across groups of patients demonstrates that our across centers. Identifying high-performing hospitals
modeling approach achieves the stated aim for that safely limit the duration of POMV is a necessary
population-level case-mix adjustment. The model was step to elucidate key practices that underlie better out-
not created or optimized to predict the duration of POMV comes. We aim to translate this knowledge into quality
for individual patients. As such, the C statistic for our improvement initiatives and reduce the duration of
model is low relative to other risk-adjustment models [27] POMV across hospitals through collaborative learning
commonly used in this population because of our within PC4.
approach.
Beyond those already mentioned, there are important
The authors acknowledge the data collection teams at all of the
limitations to consider. Our investigation was limited to participating centers and the generous donors to the University
candidate variables included in the clinical registry, and of Michigan Congenital Heart Center and CHAMPS for Mott for
there may be other important predictors not measured. their support of PC4. Funding for this study was provided by
For instance, at the time of this analysis, the variable the National Heart, Lung, and Blood Institute (K08-HL116639;
PI: Gaies).
“prior operations” was not collected in the database. The
database has since been expanded, and we will have the
opportunity to refine our model in future analyses.
This analysis was performed at 15 hospitals performing References
more than 200 index operations per year and all with 1. Jacobs JP, Jacobs ML, Austin EH, 3rd, et al. Quality measures
dedicated CICUs; whether the model would include for congenital and pediatric cardiac surgery. World J Pediatr
different variables if derived from a larger, more diverse Congenit Heart Surg 2012;3:32–47.
set of hospitals, is unclear. 2. Jensen HA, Brown KL, Pagel C, Barron DJ, Franklin RC.
Mortality as a measure of quality of care in infants with
It is important to note that our model allows a hospital congenital cardiovascular malformations following surgery.
to understand how its observed duration of ventilation Br Med Bull 2014;111:5–15.
compares with what would be expected by the case-mix. 3. Polito A, Patorno E, Costello JM, et al. Perioperative factors
Our methods, like all those that use indirect standardi- associated with prolonged mechanical ventilation after
complex congenital heart surgery. Pediatr Crit Care Med
zation approaches, do not facilitate direct comparisons 2011;12:e122–6.
of quality between hospitals that may have a different 4. Szekely A, Sapi E, Kiraly L, Szatmari A, Dinya E. Intra-
case-mix. operative and postoperative risk factors for prolonged
Ann Thorac Surg GAIES ET AL 7
2017;-:-–- POSTOPERATIVE VENTILATION METRICS

mechanical ventilation after pediatric cardiac surgery. Pae- 17. Hassinger AB, Wald EL, Goodman DM. Early postoperative
diatr Anaesth 2006;16:1166–75. fluid overload precedes acute kidney injury and is associated
5. Gaies M, Tabbutt S, Schwartz SM, et al. Clinical epidemi- with higher morbidity in pediatric cardiac surgery patients.
ology of extubation failure in the pediatric cardiac ICU: a Pediatr Crit Care Med 2014;15:131–8.
report from the Pediatric Cardiac Critical Care Consortium. 18. Shi S, Zhao Z, Liu X, et al. Perioperative risk factors for
Pediatr Crit Care Med 2015;16:837–45. prolonged mechanical ventilation following cardiac surgery
6. Harris KC, Holowachuk S, Pitfield S, et al. Should early in neonates and young infants. Chest 2008;134:768–74.
extubation be the goal for children after congenital cardiac 19. The International Pediatric and Congenital Cardiac Code.
surgery? J Thorac Cardiovasc Surg 2014;148:2642–7. Available at: http://ipccc.net/. Accessed January 18, 2017.
7. Crawford TC, Magruder JT, Grimm JC, et al. Early extuba- 20. Gaies M, Cooper DS, Tabbutt S, et al. Collaborative quality
tion: a proposed new metric. Semin Thorac Cardiovasc Surg improvement in the cardiac intensive care unit: development
2016;28:290–9. of the Paediatric Cardiac Critical Care Consortium (PC4).
8. Gupta S, Boville BM, Blanton R, et al. A multicentered pro- Cardiol Young 2015;25:951–7.
spective analysis of diagnosis, risk factors, and outcomes 21. Gaies M, Donohue JE, Willis GM, et al. Data integrity of the
associated with pediatric ventilator-associated pneumonia. Pediatric Cardiac Critical Care Consortium (PC4) clinical
Pediatr Crit Care Med 2015;16:e65–73. registry. Cardiol Young 2016;26:1090–6.
9. National Quality Forum. Available at: https://www. 22. The Society of Thoracic Surgeons National Database. Avail-
qualityforum.org/. Accessed January 18, 2017. able at: http://www.sts.org/. Accessed January 18, 2017.
10. Pasquali SK, Jacobs ML, O’Brien SM, et al. Impact of patient 23. O’Brien SM, Clarke DR, Jacobs JP, et al. An empirically based
characteristics on hospital-level outcomes assessment in tool for analyzing mortality associated with congenital heart
congenital heart surgery. Ann Thorac Surg 2015;100:1071–6; surgery. J Thorac Cardiovasc Surg 2009;138:1139–53.
discussion 1077. 24. Mahle WT, Jacobs JP, Jacobs ML, et al. Early extubation after
11. Clark PA, Lettieri CJ. Clinical model for predicting prolonged repair of tetralogy of Fallot and the Fontan procedure: an
mechanical ventilation. J Crit Care 2013;28:880.e1–e7. analysis of The Society of Thoracic Surgeons Congenital
12. Dunning J, Aub J, Kalkatc M, Levine A. A validated rule for Heart Surgery Database. Ann Thorac Surg 2016;102:850–8.
predicting patients who require prolonged ventilation post 25. Mahle WT, Nicolson SC, Hollenbeck-Pringle D, et al, Pedi-
cardiac surgery. Eur J Cardiothorac Surg 2003;24:270–6. atric Heart Network Investigators. Utilizing a collaborative
13. Kramer AA, Gershengorn HB, Wunsch H, Zimmerman JE. learning model to promote early extubation following infant
Variations in case-mix-adjusted duration of mechanical heart surgery. Pediatr Crit Care Med 2016;17:939–47.
ventilation among ICUs. Crit Care Med 2016;44:1042–8. 26. Wolf MJ, Lee EK, Nicolson SC, et al, Pediatric Heart Network
14. Bandla HP, Hopkins RL, Beckerman RC, Gozal D. Pulmo- Investigators. Rationale and methodology of a collaborative
nary risk factors compromising postoperative recovery after learning project in congenital cardiac care. Am Heart J
surgical repair for congenital heart disease. Chest 1999;116: 2016;174:129–37.
740–7. 27. O’Brien SM, Jacobs JP, Pasquali SK, et al. The Society of
15. Crow SS, Robinson JA, Burkhart HM, Dearani JA, Thoracic Surgeons Congenital Heart Surgery Database
Golden AW. Duration and magnitude of vasopressor sup- mortality risk model: part 1-statistical methodology. Ann
port predicts poor outcome after infant cardiac operations. Thorac Surg 2015;100:1054–62.
Ann Thorac Surg 2014;98:655–61. 28. Blackwood B, Clarke M, McAuley DF, McGuigan PJ,
16. Gupta P, Rettiganti M, Gossett JM, et al. Risk factors for Marshall JC, Rose L. How outcomes are defined in clinical
mechanical ventilation and reintubation after pediatric heart trials of mechanically ventilated adults and children. Am J
surgery. J Thorac Cardiovasc Surg 2016;151:451–8.e3. Respir Crit Care Med 2014;189:886–93.