Transcript Slide 1

Kansas Department of Health and Environment
Center for Health Disparities
2008 Health Disparities Conference
Topeka, Kansas, Apr. 1, 2008
Measuring Health Disparities
James P. Scanlan
Attorney at Law
Washington, DC
[email protected]
Objectives
1. Explain the problematic nature of
standard measures of differences
between rates (relative differences,
absolute differences, odds ratios)
2. Explain a plausible alternative
approach that avoids the problems with
standard measures
Part 1
Problematic Nature of Binary Measures of
Differences between Rates
References

Health Disparities Measurement tab on jpscanlan.com

Can We Actually Measure Health Disparities? Chance
(Spring 2006) (A12)

Race and Mortality, Society (Jan-Feb 2000) (A10)

The Misinterpretation of Health Inequalities in the United
Kingdom, British Society for Population Studies Conference
2006 (B7)

Measurement Problems in the National Healthcare
Disparities Report, American Public Health Association
Conference 2007 (B 12)

Items D23, D41, D43, D45, D46, D48, D52, D53
Four Binary Indicators of Differences Between Rates
Rates of experiencing some beneficial outcome:
Advantaged group (AG) = 50%
Disadvantaged group (DG) = 40%
1 Relative difference between rates of experiencing an outcome (in
terms of ratio of AG’s rate to DG’s rate (Ratio 1)): 1.25 (50/40)
2 Relative difference between rates of failing to experience the
outcome (Ratio 2): 1.20 (60/50)
3 Odds ratio (in terms of DG’s to AG’s odds of failing to experience the
outcome) : 1.50 ((60/40)/(50/50)0
4 Absolute differences between rates: 10 percentage points (50% 40%)
Table 1: Examples of Changing Rates and Changing
Differences Between Rates
Period
Yr 0 dir Yr 5 dir Yr 10 dir Yr 15
AG Rate
DG Rate
40%
23%
I
I
58% I
39% I
76% I
58% I
Ratio 1
Ratio 2
Odds Ratio
Absol Diff
1.77
1.29
2.29
.17
D
I
D
I
1.50
1.46
2.19
.19
1.31
1.75
2.28
.18
D
I
I
D
94%
85%
D 1.10
I 2.42
I 2.67
D .09
Question
In the prior slide, which measure provides the
most accurate information as to the change in
disparity?
Answer
None. There was no change in disparity. The patterns
are based on hypothetical test score data simulating the
situation where two groups have somewhat different
distributions of factors associated with some outcome.
Each measure changed in the manner that would occur
if, with no change in differences between averages, a
cutoff was lowered to allow everyone scoring just below
a cutoff now to pass the test (or if test performance
were improved such as to allow everyone between two
points to achieve the higher score)
Crucial Point

Not that various measures tend to support different
interpretations of the direction of a change in disparity
(though that is a matter of some consequence)

Rather, that no standard measure can alone provide
information as to whether there occurred a meaningful
change in disparity over time, because each measure
tends to change as the overall level of an outcome
changes

Caveat
Standard Patterns of Changes in Binary Measures as the Overall
Prevalence of an Outcome Changes
As an outcome increases from being very rare to being almost
universal:
1. Relative differences in experiencing it (Ratio 1) tend to
decrease
2. Relative differences in failing to experience it (Ratio 2) tend
to decline
3. Odds ratios tend to decrease until the approximate
intersection of Ratios 1 and 2 and thereafter increase
4. Absolute differences tend to move in the opposite direction of
odds ratios
Fig 1. Ratio of (1) AG Success Rate to DG Success Rate (Ratio
1) at Various Cutoffs Defined by AG Success Rate
5
Ratios
4
3
(1) AG Succ Rate/DG Succ Rate
2
1
0
1
3
5 10 20 30 40 50 60 70 80 90 95 97 99
Cutoffs Defined by AG Success Rate
Fig 2. Ratios of (1) AG Success Rate to DG Success
Rate (Ratio 1) and (2) DG Fail Rate to AG Fail Rate
(Ratio 2)
5
Zone A
Zone B
Ratios
4
3
Pt X
2
1
0
1
3
5
10 20 30 40 50 60 70 80 90 95 97 99
Cutoffs Defined by AG Success Rate
(1) AG Succ Rate/DG Succ Rate
(2) DG Fail Rate/AG Fail Rate
Fig 3. Ratios of (1) AG Success Rate to DG Success
Rate (Ratio 1), (2) DG Fail Rate to AG Fail Rate
(Ratio 2), and (3) DG Fail Odds to AG Fails Odds
5
Zone A
Zone B
Ratios
4
3
Pt X
2
1
0
1
3
5
10 20 30 40 50 60 70 80 90 95 97 99
Cutoffs Defined by AG Success Rate
(1) AG Succ Rate/DG Succ Rate
(2) Ratio DG Fail Rate/AG Fail Rate
(3) DG Fail Odds/AG Fail Odds
Fig 4. Ratios of (1) AG Success Rate to DG Success Rate, (2) DG Fail Rate to
AG Fail Rate, and (3) DG Fail Odds to AG Fails Odds; and Absolute Diff
Between Rates
5
Zone A
Ratios
4
3
Zone B
(1) AG Succ Rate/DG Succ Rate
(2) Ratio DG Fail Rate/AG Fail Rate
(3) DG Fail Odds/AG Fail Odds
Pt X
2
1
0
1
3
5 10 20 30 40 50 60 70 80 90 95 97 99
20
Percentage Points
Zone A
Zone B
Absolute Diff Betw Rates
10
0
1
3
5 10 20 30 40 50 60 70 80 90 95 97 99
Cutoffs Defined by AG Success Rate
Fig. 5. Ratios of (1) Wh to Bl Rate of Falling above Percentages
of the Poverty Line, (2) Bl to Wh Rate of Falling below the
Percentage, (3) Bl to Wh Odds of Falling Below the Percentage;
and (4)Absolute Difference Between Rates
4
Zone A
Zone B
Ratios
3
(1) Wh Rt Ab/Bl Rt Ab
(2) BL rt Bel/Wh Rt Bel
(3) Bl Odds Bel/Wh Odds Bel
Pt X
2
1
●
0
600 500 400 300 250 200 175 150 125 100 75 50
Percentage Points
30
Zone B
Zone A
20
(4) Absolute Diff betw Rates
Pt X
10
0
600 500 400 300 250 200 175 150 125 100
Percentage of the Poverty Line
75
50
Fig. 6. Ratio of (1) Wh to Wh Rate of Falling Above Various SBP Levels, (2)
Wh to Bl Rate of Falling below the Level, (3) Bl to Wh Odds of Falling Above
the Level; and (4) Absolute Difference Between Rates (NHANES 1999-2000,
2001-2002, Men 45-64)
5
Ratios
4
Zone A
Zone B
(1) Wh Rt Bel/Bl Rt Bel
(2) Bl Rate Ab/Wh Rate Ab
(3) Bl Odds Bel/Wh Odds Bel
3
Pt X
2
●
1
Percentage Points
110 120 130 140 150 160 170 180 190
Zone A
Zone B
20
(4) Absolute Diff betw Rates
Pt X
10
0
110
120
130
140
150
160
170
180
Systolic Blood Pressure
190
Interpretive Implications of Described Patterns of Change

Mortality and acute morbidity



declines in adverse outcomes tend to increase relative
differences in adverse outcomes but decrease relative
differences in favorable outcomes (mortality and survival)
since activity tends to be well into Zone B, reductions in adverse
outcomes tend to reduce absolute differences (increase odds
ratios)
Healthcare outcomes



improvements in care (e.g., increases in rates of receiving
procedures) tend to reduce relative differences in receipt of
procedures but increase relative differences in failure to receive
procedures
since (depending on the procedure) activity can be in Zone A or
B, improvements in care may tend to increase or decrease
absolute differences and odds ratios
issues with AHRQ and NCHS (A12, B12, D23a, D42, D52, D53)
Illustrations from Recent Journal Articles
Patterns of Black and White Rates of Adequate Hemodialysis
Sehgal AR. Impact of quality improvement efforts on race and sex
disparities in hemodialysis. JAMA 2003;289:996-1000
Rates of adequate hemodialysis:
Year
White
Black
1993
46%
36%
2000
87%
84%
Summary of changes in rate differences:
Absolute diff: decreased from 10 to 3 percentage points
Ratio 1 (adequate dialysis): decreased from 1.27 to 1.10
Ratio 2 (inadequate dialysis): increased from 1.19 to 1.23
See B12, D23, D23a, D42
Difference between means of hypothetical underlying distributions:
1993: .26 standard deviations
2000: .14 standard deviation
See Part 2 and D43
Two Contrasting Studies

Jha et al. Racial trends in the use of major procedures among
the elderly. N Engl J Med 2005;353:683-691: found (mainly)
increasing absolute differences during periods of increasing
prevalence of procedures

Trivedi et al. Trends in the quality of care and racial disparities
in Medicare managed care. N Engl J Med 2005;353:692-700:
found (mainly) declining absolute differences during periods of
increasing prevalence of appropriate care

Reconciliation: Jha et al. principally in Zone A; Trivedi et al.
principally in Zone B; see D23, D23a, D40, D40a, D41, D41a,
B11
Further examples

Pickett et al. Widening social inequalities in risk for sudden
infant death syndrome. Am J Public Health 2005;95:97-81.
(very successful “back-to-sleep” program seemed to increase
SES disparities in SIDS) See D3.

Morita et al. Effect of school-entry vaccination requirements
on racial and ethnic disparities in Hepatitis B immunization
coverage among public high school students. Pediatrics
2008;121:e547-e552. (very successful vaccination
requirement seems to reduce racial and ethnic disparities in
vaccination rates). See D52.

Baicker et al. Who you are and where you live: how race and
geography affect the treatment of Medicare beneficiaries.
Health Affairs 2004:Var-33-Var-44 (varied comparisons re
relative and absolute differences in procedures). See D53.
Pay for Performance and Healthcare Disparities

Werner et al. Racial profiling: The unintended consequences
of coronary artery bypass graft report cards. Circulation
2005;111:1257–63.


Casalino et al. Will pay-for-performance and quality
reporting affect health care disparities? Health Affairs
2007;26(3):405-414.


Increasingly cited as evidence the pay-for-performance will tend
to increase healthcare disparities
Recommends that pay-for-performance be tied to effects on
disparities as now being implement in Massachusetts
See D46, D48 (explaining Werner findins in light of
tendencies described above), D49, D51 (explaining patterns
one typically would observe in Massachusetts)
Implications of Focus Upon Subpopulation

Subpopulations that are truncated parts of overall populations tend
not to have normal distributions of factors associated with an
outcome when the distributions in the overall population are
perfectly normal

Nevertheless, since the truncated distributions tend to have regular
shapes, standard patterns of changes in binary measures (save for
odds ratios) tend to apply

Even so, there are interpretive implications of the fact that some
studies examine subpopulations defined by need for special
attention (e.g., hypertensive) rather than overall populations


Absolute differences in process outcome versus control outcomes
More serious implications with regard to “Approach 2”
Fig 7. Ratios (1) AG Success Rate to DG Success Rate, (2) DG Fail Rate to
AG Fail Rate, (3) DG Fail Odds to AG Fail Odds: and Absolute Differences
within Subpopulation Falling Below Point Defined by 30 Percent Fail Rate for
AG
4
3
Ratios
Zone A
Zone B
(1) AG Rate Above/DG Rate
Above
(2) DG Rate Bel/AG Rate Bel
(3) DG Odds Bel/AG Odds
Below
2
●
1
1
10 20 30 40 50 60 70 80 90 99
Percentage Points
20
Zone B
Zone A
(4) Absolute Diff betw Rates
10
0
1
10
20
30
40
50
60
70
80
90
99
Cutoffs Defined by AG Success Rate
w/in Truncated Subpopulation
Fig.8. Absolute Difference Between Rates within the Total
Population, and with Population Below the 30 Percent Fail Rate
for the AG, according to AG Fail Rate Within Each Population.
Zone A
Zone B
Absolute Diff betw Rates - All
10
0
1
10
20
30
40
50
60
70
80
90
99
●
Cutoffs Defined by AG Fail Rate - All
20
Zone A
Ratios
Ratios
20
Zone B
Absolute Diff betw Rate - Subpop
10
0
1
10
20
30
40
50
60
70
80
90
99
Cutoffs Defined by AG Fail Rate within
Universe Below AG Fail Ratio of 30
Fig. 9. Ratio of (1) Wh to Bl Rate of Falling below Various SBP Levels
(favorable outcome), (2) Bl to Wh Rate of Falling above the Level (adverse
outcome), (3) Bl to Wh Odds of Falling above the Level; and (4) Absolute
Difference between Rates (NHANES 1999-2000, 2001-2002, Men 45-64),
Limited to Population with SBP Above 139
5
Ratios
4
Zone A
Zone B
(1) Wh Rt Bel/Bl Rate Bel
(2) Bl Rt Ab/Wh Rt Ab
(3) Bl Odds Bel/Wh Odds Bel
3
2
●
14
4
14
8
15
2
15
6
16
0
16
4
16
8
17
2
17
6
18
0
18
4
18
8
1
Zone B
20
TADnum
10
Systolic Blood Pressure
18
8
18
4
18
0
17
6
17
2
16
8
16
4
16
0
15
6
15
2
14
8
0
14
4
Percentage Points
Zone A
Fig. 10. Absolute Differences Between Rates of Falling Above
Certain SBP Levels for Overall Population and Population with
SBP above 139
20
Zone B
Absolute Diff betw Rates
10
●
0
130
140
150
160
170
180
190
Systolic Blood Pressure
20
Zone A
Zone B
Absolute Diff betw Rates
10
Systolic Blood Pressure
19
0
18
8
18
4
18
0
17
6
17
2
16
8
16
4
16
0
15
6
15
2
0
14
8
120
14
4
110
Ratios
Ratios
Zone A
Part 2
Alternative Approaches to Measurement
Measurement Possibilities on a
Seemingly Continuous Scales









Longevity – no (see B7, B11)
SF 36 scores – no (see B11)
Metabolic syndrome measures – no (see B11)
Cardio risk indexes – no (see B11)
Allostatic load – possibly (see B11)
Components of allostatic load – possibly (see B9, B11)
Cortisol level – possibly (see B11)
Self rated health on a continuous scale - possibly (see
B7, B11)
Gini coefficient, concentration index etc (see A12, D43)
Measurement Possibilities Using
Outcome Rates

Approach 1 – departures from standard
patterns (A12, B7, D41, D43)

Approach 2 – identifying the difference
between means of hypothetical underlying
distributions based on group rates in
settings being compared (D43, D45, D46,
D48)
Table 2. Hypothetical Illustration of
Approach 2
Period
Yr 0
Yr 5
AG Rate
76%
94%
DG Rate EES
58%
.50
88%
.38
*Estimated effect size – difference between
hypothesized means in terms of
percentage of a standard deviation
Table 3: Illustration of Approach 2 Based on Data in Article to
which D48 Responds
Coronary angiogram
Year Wh Rate* Bl Rate
1988
86
43
1997 228
161
EES
Coronary angioplasty
Year Wh Rate
Bl Rate
1986
10
3
1997
26
16
Coronary artery bypass surgery
Year Wh Rate Bl Rate
EES
1986
31
8
1997
59
26
*All rates are per 10,000
.25
.14
EES
.32
.15
.41
.27
Conclusions Regarding Approach 2

Further examples on D43, D45, D46, D48, D52, D53

Procedure speculative because it rests on hypotheses
as to normality of underlying distributions (see D43)

Procedure unsuitable for truncated distributions, which
we know not to be normal (see D43, D46a)

Despite weaknesses, procedure is superior to standard
measures of differences between rates for evaluating
size of disparity in different settings

Where to go from here?
Other References

Keppel K., Pamuk E., Lynch J., et al. 2005. Methodological issues
in measuring health disparities. Vital Health Stat 2 (141)
(http://www.cdc.gov/nchs/data/series/sr_02/sr02_141.pdf) (see A12,
B12, D6)

Carr-Hill R, Chalmers-Dixon P. The Public Health Observatory
Handbook of Health Inequalities Measurement. Oxford: SEPHO;
2005 (http://www.sepho.org.uk/extras/rch_handbook.aspx) (see A7,
D8)

Houweling TAJ, Kunst AE, Huisman M, Mackenbach JP. Using
relative and absolute measures for monitoring health inequalities:
experiences from cross-national analyses on maternal and child
health. International Journal for Equity in Health 2007;6:15
(http://www.equityhealthj.com/content/6/1/15) (see D43, D50)