This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike License.

Your use of this

material constitutes acceptance of that license and the conditions of use of materials on this site.

Copyright 2008, The Johns Hopkins University Marie Diener-West, and Sukon Kanchanaraksa. All rights reserved.
Use of these materials permitted only in accordance with license rights granted. Materials provided “AS IS”; no
representations or warranties provided. User assumes all responsibility for use, and all liability related thereto, and
must independently review all materials for accuracy and efficacy. May contain materials owned by others. User is
responsible for obtaining permissions for use from third parties as needed.
The Direct and Indirect Methods of Adjustment

Marie Diener-West, PhD

Sukon Kanchanaraksa, PhD
Johns Hopkins University
Section A: Introduction

Marie Diener-West, PhD

Johns Hopkins University
Comparing Rates Among Different Populations
„ Stratum-specific rates (e.g., age-specific rates)
„ Crude rates
„ Stratum-adjusted rates (e.g. age-adjusted rates)

4
Comparison of Stratum-Specific Rates
„ If a population can be stratified (divided into groups),
appropriate comparisons may be made of stratum-specific
rates such as:
− Age-specific rates
− Cause-specific rates
− Age-cause-specific rates
− Age-gender-race-specific rates

5
Comparison of Crude Rates
„ A crude rate (overall rate) is a weighted average of stratum-
specific rates (the weights are the population totals of the
strata)
„ The difference between crude rates of two populations
involves differences in both the stratum-specific rates and
population composition (distribution of characteristics)
„ Comparison of crude rates is often confounded by these
differences and not appropriate

6
Notation
„ Let
− j = stratum
− x j = number of events in stratum j of the observed
population
− Nj = number of persons in stratum j of the observed
population
− p j = event rate in stratum j of the observed population
− C = crude rate of the observed population
∑xj total deaths
=
− C= ∑Nj totalpopulation

Xj deaths in stratum j
=
„ Stratum-specific rate = Nj populationin stratum j
7
Example: Mortality by Age in Population 1

Population 1
Stratum i Total
Age (years) Deaths (Xi)
(Age group) population (Ni )
1 0–4 100 5
2 5–14 90 10
3 15–19 110 15
Total 300 30

„ Crude death rate = 30 deaths/300 population = 0.10 = 10

deaths per 100 population

8
Example: Mortality by Age in Population 2

Population 2
Stratum i Total
Age (years) Deaths (Xi)
(Age group) population (Ni )
1 0–4 165 10
2 5–14 75 10
3 15–19 60 10
Total 300 30

„ Crude death rate = 30 deaths/300 population = 0.10 = 10

deaths per 100 population

9
Example: Comparison of Crude Death Rates
„ The crude death rate (CDR) for each population is 10 deaths
per 100 population
„ Is the risk of dying the same in the two populations?

10
Example: Mortality by Age in Population 1

Population 1
Stratum i Age Total Deaths Death Rate per
(Age group) (years) Population (N1i ) (X1i) 100 (X1j/N1j)*100
1 0–4 100 5 5
2 5–14 90 10 11
3 15–19 110 15 14
Total 300 30

11
Example: Mortality by Age in Population 2

Population 2
Stratum i Age Total Deaths Death Rate per
(Age group) (years) Population (N2i ) (X2i) 100 (X2j/N2j)*100
1 0–4 165 10 6
2 5–14 75 10 13
3 15–19 60 10 17
Total 300 30

12
Example: Comparison of Age-Specific Death Rates
„ Age-specific death rates (per 100 population)

Stratum i Age Population 1 Population 2

1 0–4 5 6
2 5–14 11 13
3 15–19 14 17

13
Conclusion: Age-Specific vs. Crude Death Rates?
„ Inspection of the age-specific rates reveals a higher death
rate in each age group for Population 2 as compared to
Population 1
„ Why are the crude rates the same in the two populations?
− Hint: look at the age composition (population
distribution by age)!

14
Population 1 Distribution

Population 1
Stratum i Age Total % Death rate per 100
1 0–4 100 33 5
2 5–14 90 30 11
3 15–19 110 37 14

15
Population 2 Distribution

Population 2
Stratum i Age Total % Death rate per 100
1 0–4 165 55 6
2 5–14 75 25 13
3 15–19 60 20 17

16
Crude Death Rates for Populations 1 vs. 2
„ The Crude Death Rate (CDR) is a weighted average of
the age-specific death rates.

„ For Population 1:
∑p1jN1j (0.05×100)+(0.11×90)+(0.14×110)
CDR1 = = = .10
N1j 300

„ For Population 2:

∑ p2jN2j (0.06 ×165)+(0.13×75)+(0.17× 60)

CDR 2 = = = .10
N2j 300

17
What Is the Appropriate Comparison?
„ The comparison of crude death rates is confounded by the
differences in population composition (age distribution)
between the two populations
„ Population 2 has a younger age distribution
„ A comparison of age-specific death rates between the two
populations reflects the risk of dying in each age group
„ An adjustment procedure is needed to make an
appropriate comparison of the overall risk of dying between
the two populations

18
Review
„ Why might crude death rates be misleading?
„ What is an alternative to comparing crude death rates?

19
Next Steps
„ The two most common adjustment procedures for rates:
− Direct method of adjustment
− Indirect method of adjustment
„ Commonly used in vital statistics and epidemiology.

20
Section B: Direct Method of Adjustment

Marie Diener-West, PhD

Johns Hopkins University
Adjustment Procedures
„ Adjustment procedures are any of a variety of procedures
performed during data analysis to attempt to remove the
effect of one or more extraneous sources of variation that
could affect (or are believed to affect) a particular result
(Meinert, 1996)

22
Goals of Adjustment Procedures
„ Appropriately combine data
„ Make appropriate comparisons among groups
„ Reach appropriate conclusions and inferences

23
Types of Adjustment Procedures
„ Stratified or subgroup analyses
„ Direct or indirect standardization of rates
„ Life tables
„ Multivariable statistical analyses

24
Adjusted or Standardized Rates
„ Calculation of adjusted (standardized) rates allows
comparison of summary event rates between populations
when there are differences in characteristics between the
populations that may influence the event of interest
− For example, age, race, gender, disease status

25
Methods of Adjustment of Rates
„ Direct method
− Apply stratum-specific rates observed in the populations
of interest to a reference or standard population in
order to obtain the number of deaths expected in the
reference population
− Calculate an adjusted rate based on expected number of
deaths in the reference population
„ Indirect method
− Apply stratum-specific reference rates to the
populations of interest to obtain the number of expected
deaths in each of those populations
− Compare the observed number of deaths to the expected
number of deaths for each population of interest

26
Notation for Populations of Interest

N 1j = Number of individuals in stratum j of Population 1

N 2j = Number of individuals in stratum j of Population 2
Number of individuals in stratum j of Population 1 who have
X 1j =
the event
Number of individuals in stratum j of Population 2 who have
X 2j
the event

P 1j = X 1j / N 1j = Rate in stratum j of Population 1

P 2j = X 2j / N 2j = Rate in stratum j of Population 2

27
Notation for Reference Population

Nj = Number of individuals in stratum j of a reference population

Pj = Rate for stratum j of the reference populations
C = Crude rate of the reference population

28
Adjusted Rates by the Direct Method
„ Use the event rates (experience) of the population of
interest to calculate the number of deaths expected in the
reference population
„ The adjusted rate in the population of interest is the
expected number of deaths divided by the total reference
population

29
Adjusted Rates by the Direct Method
„ Adjusted rate for Population 1

=
∑ p ⋅N 1j j

∑N j

„ Adjusted rate for Population 2

=
∑ p ⋅N 2j j

∑N j

30
Example: U.S. Mortality—Single vs. Married Males

Single Males Married Males

Column 1 Column 2 Column 3 Column 4 Column 5 Column 6
N1j X1j P1j N2j X2j P2j
Population Number of Death Rate Population Number of Death Rate
J Age (thousands) Deaths Per 1,000 (thousands) Deaths Per 1,000

1 15–24 17,724 39,745 2.24 3,427 7,327 2.14

2 25–44 5,390 26,372 4.89 23,083 60,337 2.61
3 45–64 1,210 37,125 30.68 18,088 286,535 15.84
4 65–74 364 33,679 92.52 5,500 275,818 50.15
5 75+ 199 64,386 323.55 2,331 275,135 118.03
Total 24,887 201,307 8.09 52,429 905,152 17.26

31
Crude Mortality Rates: Single vs. Married Males
„ Crude death rates per 1,000 population are:
− Single: 8.09
− Married: 17.26
„ A comparison of crude rates gives the impression that the risk
of dying is twice as high in married males as single males

32
U.S. Males, Death Rates by Marital Status and Age

Single males

Married males

33
Summary: U.S. Mortality, Single vs. Married Males
„ The crude death rate is higher for married males than single
males
„ However, age-specific death rates per 1,000 are lower for
married males in each age stratum
„ Why is this?
„ The two populations have very different age distributions:
− Single: 90% are aged < 45
− Married: 51% are aged < 45

34
Age-Specific Death Rates: Single vs. Married Males

Single Males Married Males

Column 1 Column 2 Column 3 Column 4 Column 5 Column 6
N1j X1j P1j N2j X2j P2j
Population Number of Death Rate Population Number of Death Rate
J Age (thousands) Deaths Per 1,000 (thousands) Deaths Per 1,000

1 15–24 17,724 39,745 2.24 3,427 7,327 2.14

2 25–44 5,390 26,372 4.89 23,083 60,337 2.61
3 45–64 1,210 37,125 30.68 18,088 286,535 15.84
4 65–74 364 33,679 92.52 5,500 275,818 50.15
5 75+ 199 64,386 323.55 2,331 275,135 118.03
Total 24,887 201,307 8.09 52,429 905,152 17.26

35
 Calculating Expected Deaths: U.S. Mortality, Males

Column 3 x Column 6 x
Column 3 Column 6 Column 7
Column 7 Column 7
Nj
Expected Expected
P1j Single Death P2j Married Death Reference
J Age Deaths— Deaths—
Rate per 1000 Rate per 1000 Population
Single Married
(Thousands)

1 15-24 2.24 2.14 21,151 47,430 45,221

2 25-44 4.89 2.61 28,473 139,312 74,426

3 45-64 30.68 15.84 19,298 592,097 305,703

4 65-74 92.52 50.15 5,864 542,565 294,072

5 75+ 323.55 118.03 2,530 818,576 298,624

Total 77,316 2,139,980 1,018,046

Note: rates used to calculate expected deaths were carried to more decimal places
Reference population: total of single and married populations

36
 Calculating Expected Deaths: U.S. Mortality, Males

Column 3 x Column 6 x
Column 3 Column 6 Column 7
Column 7 Column 7
Nj
Expected Expected
P1j Single Death P2j Married Death Reference
J Age Deaths— Deaths—
Rate per 1000 Rate per 1000 Population
Single Married
(Thousands)

1 15-24 2.24 2.14 21,151 47,430 45,221

2 25-44 4.89 2.61 28,473 139,312 74,426

3 45-64 30.68 15.84 19,298 592,097 305,703

4 65-74 92.52 50.15 5,864 542,565 294,072

5 75+ 323.55 118.03 2,530 818,576 298,624

Total 77,316 2,139,980 1,018,046

Note: rates used to calculate expected deaths were carried to more decimal places
Reference population: total of single and married populations

37
 Calculating Expected Deaths: U.S. Mortality, Males

Column 3 x Column 6 x
Column 3 Column 6 Column 7
Column 7 Column 7
Nj
Expected Expected
P1j Single Death P2j Married Death Reference
J Age Deaths— Deaths—
Rate per 1000 Rate per 1000 Population
Single Married
(Thousands)

1 15-24 2.24 2.14 21,151 47,430 45,221

2 25-44 4.89 2.61 28,473 139,312 74,426

3 45-64 30.68 15.84 19,298 592,097 305,703

4 65-74 92.52 50.15 5,864 542,565 294,072

5 75+ 323.55 118.03 2,530 818,576 298,624

Total 77,316 2,139,980 1,018,046

Note: rates used to calculate expected deaths were carried to more decimal places
Reference population: total of single and married populations

38
Directly Age-Adjusted Death Rate: Single Males
„ Age-adjusted rate per 1,000 for single males

= expected deaths inreference populationusing single rates

totalreference population
= 2,139,980
77,316

„ DAR s = 27.68 deaths per 1,000 population

39
Directly Age-Adjusted Death Rate: Married Males
„ Age-adjusted rate per 1,000 for married males

= expected deaths inreference populationusingmarriedrates

totalreference population
= 1,018,046
77,316

„ DARm = 13.17 deaths per 1,000 population

40
Comparing Adjusted Rates: Married vs. Single Males
„ The directly age-adjusted death rates per 1,000 are:
− Single: DAR s = 27.68
− Married: DARm = 13.17
„ A comparison of the directly age-adjusted rates reveals that
the risk of dying is twice as high for single versus married
males after adjusting for the differences in population age
distribution between the two groups

41
Quick Check
„ Why do the crude death rates suggest higher risk of dying in
married males than in single males?

42
Notation

Let j = stratum
xj = number of events in stratum j of the observed
population
Nj = number of persons in stratum j of the observed
population
pj = event rate in stratum j of the observed population
DARi =
∑ p 1j ⋅N j

∑Nj
DARi =
n
∑ expected deaths inreference population (usingPop irates)
totalreference population

43
Another Example of Age-Adjustment

Comparison of Crude Death Rates in a Population

at Two Different Time Periods
Early Period Later Period
No. of Rate per No. of Rate per
Population Population
Deaths 100,000 Deaths 100,000
900,000 862 96 900,000 1,130 126

44
An Example of Age-Adjustment

Comparison of Age-Specific Death Rates

in the Two Time Periods
Early Period Later Period
Death Death
Age No. of Rate per No. of Rate per
Group Population Deaths 100,000 Population Deaths 100,000
30–49 500,000 60 12 300,000 30 10
50–69 300,000 396 132 400,000 400 100
70+ 100,000 406 406 200,000 700 350
Total 900,000 862 96 900,000 1,130 126

45
An Example of Age-Adjustment

Carrying out an Age Adjustment Using the Total

of the Two Populations as the Standard
Early Period Later Period
Age Standard Death Rate Expected No. Death Rate Expected No.
Group Population per 100,000 of Deaths per 100,000 of Deaths
30–49 800,000 12 96 10 80
50–69 700,000 132 924 100 700
70+ 300,000 406 1,218 350 1,050
Total 1,800,000 2,238 1,830
Age-Adjusted Rate 2,238 1,830
1,800,000 1,800,000

46
Summary of the Direct Method of Adjustment
„ The age-adjusted rate is an index measure, the magnitude
of which has no intrinsic value
„ The actual rates are only meaningful when directly compared
to each other
„ The adjusted rates are useful for comparison purposes only
„ The choice of the reference population is important
− It should not be abnormal or unnatural
„ Adjustment (standardization) is not a substitute for the
examination of age-specific rates in the populations of
interest

47
Review Questions
„ What is an age-adjusted rate?
„ Can directly-adjusted rates be compared?
„ Using the direct method of adjustment, what data are needed
from the reference (standard) population?

48
Section C: Indirect Adjustment

Sukon Kanchanaraksa, PhD

Johns Hopkins University
Direct versus Indirect Method of Adjustment
„ Direct method
− Assume that the observed population had the same
distribution of characteristics as the reference population
− Apply the experience of the observed population to the
reference population
„ Indirect method
− Assume that the observed population had the experience
of the reference population
− Apply the experience of the reference population to the
observed population

Experience
Experience==morbidity
morbidityrate
rateor
ormortality
mortalityrate
rate
Characteristics
Characteristics==age,
age,gender,
gender,……

50
Indirect Method of Adjustment
„ Apply the experience (e.g., mortality rate) of the reference
(standard) population to the population of interest (observed
population) by strata (age group)
„ Sum the values to obtain the expected number of deaths
„ Divide the observed number of deaths by the expected
number of deaths in the population of interest to get a value
called SMR (Standardized Mortality Ratio)
„ Multiply SMR by the crude rate (C) of the reference
population to get the Indirect Adjusted Rate (IAR)

51
Notation
Let j = strata (e.g., age group)
xj = number of events in stratum j of the observed
population (e.g., observed number of deaths)
Nj = number of persons in stratum j of the observed
population
pj = rate in stratum j of the reference population (e.g.,
mortality rate)
C = crude rate of the reference population
SMR = ∑Xj
∑ (p jN j )

IAR = SMR x C
IAR
Thus: SMR =
C

52
Example of an Indirect Adjustment
„ In a population of 534,533 White male miners, 436 died from
tuberculosis (TBC) in 1950
„ Is this mortality experience from TBC greater than, less than,
or about the same as that which you would expect in White
males of the same ages in the general population?
„ And, what is the IAR of the White male miners compared to
the general population of White males?

53
 Computation of an SMR for Tuberculosis
Computation of an SMR for Tuberculosis, All Forms (TBC),
for White Male Miners Ages 20–59 yrs, U.S., 1950
(pj) (pj .Nj) (Xj)
(Nj) Death Rate Per Expected Deaths Observed
100,000 for TBC in from TBC in White Deaths
Estimated White Males Miners (If Same Risk from TBC in
Population of (General as General White
(j) White Miners Population) Population Miners
Age (Yrs) (1) (2) (3)=(1)X(2) (4)
20–24 74,598 12.26 9.14 10
25–29 85,077 16.12 13.71 20
30–34 80,845 21.54 17.41 22
35–44 148,870 33.96 50.55 98
45–54 102,649 56.82 58.32 174
55–59 42,494 75.23 31.96 112
Total 181.09 436

Adapted from: Vital Statistics - Special Reports, DHEW, 53(5), 1963 54

Computation for SMR and IAR
observed deaths
„ SMR =
expected deaths
436
= x 100
181.09
= 241% or 2.41

„ IAR = SMR x C
= 2.41 x 10.9 per 1000
= 26.3 per 1000

55
Interpretation of SMR
„ SMR = 1
− Risk is the same in both the observed population and the
reference population
„ SMR < 1
− Risk is lower in the observed population compared to the
reference population
„ SMR > 1
− Risk is higher in the observed population compared to
the reference population

SMR 2.41 ⇒
SMR==2.41 ⇒White
Whiteminers
minershad
had2.41
2.41times
timesthe
therisk
riskof
of
mortality
mortalityof
ofthe
theU.S.
U.S.White
Whitemale
malepopulation
population

The
Theindirect-adjusted
indirect-adjustedmortality
mortalityrate
ratefor
forWhite
Whiteminers
minerswas
was
26.3
26.3per
per1000
1000
56
Comparison of SMRs
„ Since the number of deaths in a population depends on age
distribution, the number of observed deaths and the
calculation of the expected deaths must depend on the age
distribution of the population of interest
− Consequently, SMR must also depend on the age
distribution of population of interest
„ Therefore, when using the same reference population in the
calculation of SMR, the SMR from one population cannot be
compared to the SMR from another population unless the
two populations are similar in age distribution

57
Quick Check
„ If the SMR from a textile worker industry were 4.0 or 400%
and the SMR from miners were 2.4 …
− Could we conclude that the risk of death of the textile
workers was 4/2.4=1.7 times higher than the miners?

58
Example of SMR by Occupation

SMR by Occupation
Men Aged 20–64 in the United States, 1950
SMR

Disease of
Occupation Level Respiratory System Asthma
All occupations 100 100
Professional workers 72 71
Technical, administrative and
52 79
management workers
Clinical sales and skilled workers 87 104
Semiskilled workers 149 99
Laborers 157 145
Agricultural workers 75 95

Source: Higgins, 1974 59

Merits of Indirect Adjustment
„ No need to know the age-specific (mortality) rates of the
population of interest
− These rates may be difficult to obtain (in a developing
country or an industry)
− The direct method of adjustment cannot be conducted
without these rates
„ The rates of the standard population are often based on large
population, while the rates calculated from the observed
population may be based on small number and be unstable

60
Review Questions
„ What is SMR?
„ Can SMR of one population be compared to SMR of another
population?
„ In the indirect method, what data from the reference
(standard) population is used?

61