Measurement Problems in the National Healthcare Disparities Report

1. Measurement Problems in the National Healthcare Disparities Report American Public Health Association 135th Annual Meeting & Exposition, Nov. 3-7,2007, Washington, DC James P. Scanlan Attorney at Law Washington, DC jps@jpscanlan.com

2. National Healthcare Disparities Report (NHDR) of Agency for Healthcare Research and Quality (AHRQ) • Yearly reports mandated by congress • 4 reports issued (2003-2006); 2007 forthcoming • 2005 and 2006 measure disparity changes over time • 2006 report: • 24 process quality core measures (e.g., prenatal care in first trimester) • 18 clinical outcome quality core measure (e.g., colorectal cancer) • 6 access core measures (e.g., insurance coverage) • For virtually all core measures the sizes of disparities are evaluated in terms of relative difference in rates of adverse outcomes (e.g., failure to receive prenatal care in first trimester, cancer rates, lack of insurance)

3. Key Measurement Problem in NHDR • NHDR fails to recognize that all measures of differences between two groups’ rates of experiencing or avoiding some outcome tend to change in certain ways solely as a result of changes in the prevalence of the outcome. • Without recognizing and attempting to account for these tendencies it is impossible to draw meaningful conclusions about changes in health or healthcare disparities over time. • According to the approach in the NHDR, as healthcare improves, and favorable outcomes become more common (adverse outcomes become less common), healthcare disparities will be perceived to be increasing.

4. Patterns of Changes in Four Measures of Differences Between Rates as an Outcome Goes from Being Very Rare to Being Almost Universal • Relative differences in experiencing the outcome tend to decrease. • Relative differences in failing to experience the outcome tend to increase. • Absolute differences initially increase then decline. • Odds ratios initially decline then increase.

5. References • jpscanlan.com – Measuring Health Disparities Tab (especially Section D) • “Can We Actually Measure Health Disparities?” Chance 2006 • “Measuring Health Disparities,” J Public Health Management and Practice 2006 • “The Misinterpretation of Health Inequalities in the United Kingdom,” British Society for Population Studies Conference 2006 • “Race and Mortality,” Society 2000

6. Specifications for Figures 1- 5 • Two normal distributions of factors associated with risks of experiencing or avoiding some outcome (e.g., scores on a paper and pencil test) • Mean of advantaged group (AG) is one half a standard deviation higher than mean of disadvantaged group (DG) • Distributions have same standard deviation

7. Fig 1: Ratio of (1) AG Success Rate to DG Success Rate (Ratio a) at Various Cutoffs Defined by AG Success Rate

8. Fig 2: Ratios of (1) AG Success Rate to DG Success Rate (Ratio a) and (2) DG Fail Rate to AG Fail Rate (Ratio b)

9. Fig 3: Ratios of (1) AG Success Rate to DG Success Rate (Ratio a), (2) DG Fail Rate to AG Fail Rate (Ratio b), and (3) DG Fail Odds to AG Fails Odds

10. Fig 4: Ratios of (1) AG Success Rate to DG Success Rate (Ratio a), (2) DG Fail Rate to AG Fail Rate (Ratio b), and (3) DG Fail Odds to AG Fails Odds; and (4) Absolute Difference Between Rates

11. Fig 5: Absolute Difference Between Success (or Failure) Rates of AG and DG at Various Cutoffs

12. Other Illustrative Data • Income data (Chance 2006) • NHANES data (D41, BSPS 2007) • Framingham calculator (do it yourself) • Published test score/proficiency data

13. A Timely Illustration (from Sunday’s Washington Post) From a chart styled “Remarkable Results” in De Vise D. Closing the gap. Washington Post Nov. 4, 2007:1,12. Math Proficiency Rates at Rock View Elementary: 2003 2007 Black 53% 81% White 81% 97% Summary of changes: Absolute diff: decreased from 28 to 16 Relative diff in proficiency: decreased from 53% to 20% Relative diff in non-proficiency: increased from 147% to 530%

14. Two Contrasting Studies of Changes in Absolute Differences Between Black and White Rates as Favorable Outcome Becomes More Common • Jha et al. Racial trends in the use of major procedures among the elderly. N Engl J Med 2005;353:683-691 • increasing overall rates of relatively uncommon outcomes • increasing absolute differences • Trivedi et al. Trends in the quality of care and racial disparities in Medicare managed care. N Engl J Med 2005;353:692-700 • increasing overall rates of more common outcomes • decreasing absolute differences • See D23, D40, D41

15. Illustration Based on Rates of Adequate Hemodialysis • Sehgal AR. Impact of quality improvement efforts on race and sex disparities in hemodialysis. JAMA 2003;289:996-1000 • Found that as rates of adequate dialysis increased, absolute differences declined. • Cited by AHRQ officials and others as showing how improving care reduces disparities. • Under measurement approach in the NHDR, however, the disparities would be deemed to have increased. • Rates of adequate dialysis White Black 1993 46% 36% 2000 87% 84% Summary of changes: Absolute diff: decreased from 10 to 3 Relative diff in adequate dialysis: decreased from 28% to 4% (cor. 3/12/08) Relative diff in inadequate dialysis: increased from 19% to 23%

16. Varied Implications of Measurement in Terms of Relative Differences in Adverse Outcome • As healthcare and health improve, disparities will seem to increase. • Most effective measures (even ones seemingly focused on the disadvantaged) will seem to increase disparities (see D3 re Back to Sleep Program). • Disparities will seem to be largest in areas or among subpopulations where adverse outcomes are rarest (see Chance 2006, Society 2000, BSPS 2006, Nordic Conf. 2006, D16,17). • Disparities tend to be largest with respect to adverse outcomes that are rarest (also pertinent to the National Healthcare Quality Report’s analyses of variance across states).

17. Which measure is best? • None of the four measures discussed alone can indicate whether a change between rates is other than solely a consequence of changes in prevalence. • Further, each measure can change in one direction even when there in fact is a meaningful change in the opposite direction. • Be mindful that it is not always the favorable outcome that is increasing – and that as adverse outcomes increase, relative differences in those outcomes will tend to decline.

18. Can we actually measure health or healthcare disparities? • Dichotomous variables while taking the described tendencies into account (Chance 2006, BSPS 2006) • Continuous variables (BSPS 2006, Lisbon 2007*, JSM 2007) • The problematic implications of changes in smoking prevalence (BSPS 2006) • Different possibilities for measuring health outcome disparities and health process disparities