A Summary Measure of Health Disparities Yukiko Asada, PhD AcademyHealth, Methods Workshop Health Disparities: Concepts and Methods June 28, 2010 1 Health disparities research and policy: Current situation • • • Lots of description Increasing call for action Not satisfactory results 2 Health disparities research and policy: Current situation • • • Lots of description Increasing call for action Not satisfactory results May partly due to the way we typically measure health disparities 3 Typical way to measure health disparities: Illustration Better Health ... Summary number Worse Group A Group B Group C Attribute 4 '%# Sex Neighbourhood SES Race/ethnicity (male) Race/ethnicity (female) &"# &%# $"# $%# "# 9# 3: ;< ?;@ :. +/ 5* A= 6# 1 .? ;@+ /# 0 =5 :; *. ;+ # /D B; 5C E+ +/ @;F ;@# @ #G => 5,+ ?;@ /H +/ .? A= # 1 .? ;@+ /# 0 =5 :; *. ;+ # /D B; 5C E+ +/ @;F ;@# @ #G 5,+ /H .? # => 8# 7# !"# () *+ ,# ./ # 0 )1 ./ 2# # 3,) 4 .5 *6# Life expectancy at birth Typical way to measure health disparities: Example (life expectancy, California, 1999-2001) Source: Clarke et al. 2010 5 '%# Sex Neighbourhood SES Race/ethnicity (male) Race/ethnicity (female) &"# &%# $"# $%# "# 9# 3: ;< ?;@ :. +/ 5* A= 6# 1 .? ;@+ /# 0 =5 :; *. ;+ # /D B; 5C E+ +/ @;F ;@# @ #G => 5,+ ?;@ /H +/ .? A= # 1 .? ;@+ /# 0 =5 :; *. ;+ # /D B; 5C E+ +/ @;F ;@# @ #G 5,+ /H .? # 8# 7# !"# () *+ ,# ./ # 0 )1 ./ 2# # 3,) 4 .5 *6# Many pieces of information, but hard to obtain overall picture => Life expectancy at birth Typical way to measure health disparities: Example (life expectancy, California, 1999-2001) 6 Bivariate vs. univariate health disparity • Bivariate health disparity - measuring a joint distribution of health and another attribute, such as income, education, social class, gender, race/ethnicity, or rurality • Univariate health disparity - measuring a distribution of health only (Wolfson & Rowe 2001) 7 Univariate health disparity: Illustration Pop. freq. Summary number Worse Better Health 8 Univariate health disparity: Example Source: The World Health Report 2000 9 Univariate health disparity: Example Overall picture, but no information on bivariate health disparities Source: The World Health Report 2000 10 Country A pop. freq. Country B pop. freq. Life expectancy Life expectancy Source: Asada & Hedemann (2002) 11 Country A Country B pop. freq. pop. freq. pop. pop. freq. freq. pop. pop. freq. freq. Group Group Group Group 1111 Group Group Group 111 Group Group Group 222 Group Group Group 22 Health Life expectancy expectancy Health Life expectancy expectancy 12 A summary measure of health disparities: A proposal • What it measures - overall picture + many bivariate health disparities • Purpose - to measure and track health disparities to assist policy to reducing them 13 Approaches • Bottom-up - measures many bivariate health disparities separately - combines them into a summary index • Top-down - estimates the overall health disparity - breaks it down into health disparities related to different attributes 14 Bottom-up approach: Health of Wisconsin Report Card 2007 • Examined 4 attributes - gender, education, rurality, and race/ethnicity • Calculated the Index of Disparity - a modified coefficient of variation developed by Pearcy and Keppel for Healthy People 2010 - using all 14 groups - 2 gender, 3 education, 4 rurality, and 5 race/ ethnicity groups 15 Examples Income - rich / poor Education - low / high education Gender - male / female 16 Healthy 17 Rich Female Healthy High education Male Low education Poor 18 Rich Female Healthy High education Male Low education Poor 19 Rich Female Healthy High education Male Low education Poor 20 Rich Female Healthy High education Low education Male Distance Poor # group - 1 21 Bottom-up approach: Health of Wisconsin Report Card 2007 Attributes examined: gender, education, rurality, and race/ethnicity 22 Issues for the bottom-up approach • “Overall picture” is a composite, not univariate health disparity • Each bivariate disparity is “unadjusted” 23 Top-down approach • Relevant methods available, but no direct attempt in the past • Possible with the regression-based inequality decomposition methods - Univariate health disparity measured by a measure of choice (e.g., the Gini coefficient) - Decomposition tells us how much each attribute (bivariate health disparity) explains the univariate health disparity 24 Preliminary analysis using the top-down approach • Data: Joint Canada/United States Survey of Health (2002/03) - “clean” (no missing) data - n=2887 (Canada), n=3780 (US) • Examined disparities in the Health Utilities Index - 0 = dead; 1= full health • Attributes (bivariate health disparities) considered - sex, race, education, and income • Country-specific analysis: - calculated univariate disparity in HUI by the Gini coefficient (Stata: ineqdec0) - run regression based decomposition by the four attributes, adjusting for age (Stata: ineqrbd) 25 Top-down approach: A preliminary result, disparities in HUI Canada Univariate health disparity (Gini) 0.089 Attribute contribution (%) Sex 0.10 Race 0.01 Income 4.13 Education 2.28 Age (adjustment) 3.18 Unexplained (residual) 90.30 US 0.102 0.07 0.14 5.56 3.87 3.87 86.50 Data: clean (no missing) weighted JCUSH 26 Summary • A summary measure of health disparities provides an overall picture of health disparities along with information on many bivariate health disparities • The bottom-up and top-up approaches are possible • Lots of work need to be done, but development worthwhile as a potential tool to assist policy 27 Contact information Yukiko Asada, PhD Department of Community Health & Epidemiology, Dalhousie University, Canada yukiko.asada@dal.ca / tel: +1-902-494-1421 Funding Canadian Institutes of Health Research New Investigator Award Dalhousie Faculty of Medicine Clinical Research Scholar Award 28 References • The idea presented here appears in: - Y. Asada. Prev Chronic Dis 7 (2010) 4 • Work cited: - Y. Asada & T. Henemann. International Journal for Equity in Health (May 2002) - B.C. Booske et al. Health of Wisconsin Report Card. University of Wisconsin Population Health Institute (2007) - B.C. Booske et al. Prev Chronic Dis 7 (2010) 1 - C.A. Clarke et al. Soc Sci & Med 79 (2010) 1373-1380 - J. Pearcy & K. Keppel. Public Health Rep 117 (2002) 273-380 - M. Wolfson & G. Rowe. Bull World Health Organ 79 (2001) 553-60 - World Health Organization. The World Health Report 2000 (2010) 29