American Journal of Epidemiology Advance Access originally published online on March 15, 2008
American Journal of Epidemiology 2008 167(8):905-907; doi:10.1093/aje/kwn015
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Harper et al. Respond to "Measuring Social Disparities in Health"
1 Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, Quebec, Canada
2 Division of Cancer Control and Population Sciences, National Cancer Institute, National Institutes of Health, Bethesda, MD
Correspondence to Dr. Sam Harper, Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, 1020 Pine Avenue West, Room 17B, Montreal, Quebec H3A 1A2, Canada (e-mail: sam.harper{at}mcgill.ca).
Received for publication January 9, 2008. Accepted for publication January 15, 2008.
Abbreviations: CHD, coronary heart disease
 |
INTRODUCTION |
|---|
 TOP INTRODUCTION ABSOLUTE MEASURESIMPROVING INTERPRETABILITY:...CAUSAL MODELSCONCLUDING REMARKSReferences |
|---|
We appreciate Messer's thoughtful comments (1) on our article (2) and, broadly speaking, we agree that health disparities research and policymaking would benefit from increased attention to the issues of scale, interpretability, and causal relations in the measurement of health disparities.
|
ABSOLUTE MEASURES |
|---|
|
TOPINTRODUCTIONABSOLUTE MEASURESIMPROVING INTERPRETABILITY:...CAUSAL MODELSCONCLUDING REMARKSReferences |
|---|
Like Messer (1), we have suggested using measures of absolute disparity, at least as a starting point for discussions of the size of health disparities, because they quantify the absolute burden of disease among disadvantaged populations and the potential gains to overall population health from reducing absolute disparities (3, 4). However, by way of clarification, her argument that absolute measures are preferable because they indicate the fraction of the disadvantaged population adversely affected is not necessarily true. Houweling et al. (5) show that absolute disparities and disease levels tend to have an inverse U-shaped association (differences tend to be larger when overall rates of disease are average and smaller at the extremes), whereas ratio measures generally decline with increasing overall prevalence.
Relative disparity indicators may be useful when comparing disparities across outcomes measured on different scales. For example, Keppel (6) recently ranked relative disparities across hundreds of Healthy People 2010 objectives to identify the 10 "largest" health disparities for five race-ethnic groups. However, his paper also highlights the potential utility of the summary measures of health disparity that we surveyed, as it is difficult to think about how to prioritize an ordering of 50 race-ethnic-specific disparity measures, even for a single time period. Despite the appeal of relative measures for comparing across outcomes, an exclusive focus on relative measures can nevertheless lead to counterintuitive conclusions. For instance, coronary heart disease (CHD) and stroke mortality do not appear on Keppel's top 10 list of health disparities for any race-ethnic group, despite the fact that they still account for one third of US deaths and make large contributions to race-ethnic differences in life expectancy (7).
|
IMPROVING INTERPRETABILITY: APPROACHES FROM THE RESIDENTIAL SEGREGATION LITERATURE |
|---|
|
TOPINTRODUCTIONABSOLUTE MEASURESIMPROVING INTERPRETABILITY:...CAUSAL MODELSCONCLUDING REMARKSReferences |
|---|
Messer (1) suggests that a fruitful approach for health disparities measurement may be to draw upon the literature on the measurement of residential segregation, and she cites a major reason being the concrete meanings embodied in the five dimensions developed by Massey and Denton (8). However, a closer look at the segregation literature reveals some ambiguity. The dimensions in Messer's table 1 (1) are explicitly designed to measure segregation between only two groups (e.g., Black/White, Hispanic/non-Hispanic, and so on), which simplifies their interpretation and may limit their applicability in an era of increasing diversity (9). As with summary measures of health disparity, attempts to develop multigroup summary measures of segregation reduce the "concreteness" of their interpretation. For example, Reardon and Firebaugh interpret their favored multigroup "H index" of segregation (derived from the same Theil index used in our paper (2)) as "one minus the ratio of the average within-unit population diversity to the diversity of the total population" (9, p. 42). One could argue that this is no less complicated an interpretation than any of the summary measures of health disparity that we considered. Thus, although we agree with Messer that clear interpretations of health disparity measures are desirable, there is likely some unavoidable trade-off between the amount of information summarized in a measure and the simplicity of its interpretation.
Measuring segregation over time also leads to interpretative difficulties not dissimilar to those we report. The attempt by Massey et al. (10) to replicate the original analysis a decade later found only four, not five, dimensions of segregation, while a recent investigation (11) failed to replicate the initial findings and concluded, as have others (12), that segregation has only two dimensions. Despite the apparent conceptual distinctness of segregation measures, in the end demographers are likely to face the same problem as health disparity researchers, and that is deciding which aspects of segregation (or disparity) are important. From 1980 to 2000, Hispanic segregation from Whites increased by 22 percent when measured by the isolation index but decreased by 5 percent when measured by the absolute centralization index (13). Thus, the answer to the question "Are Hispanics now more or less segregated from Whites?" depends on the value one attaches to either isolation or centralization.
|
CAUSAL MODELS |
|---|
|
TOPINTRODUCTIONABSOLUTE MEASURESIMPROVING INTERPRETABILITY:...CAUSAL MODELSCONCLUDING REMARKSReferences |
|---|
We agree with Messer (1) on the importance of developing conceptual models for particular health disparities as a means to inform the choice of disparity indictor. This is particularly true because causal models for relative disparity may be different from causal models for absolute disparity. For instance, over 200 risk factors for CHD have been identified (14), and while conventional cardiovascular risk factors (cholesterol, smoking, hypertension, and diabetes) may statistically "explain" only a third of relative disparity in CHD across social groups, their contribution to absolute disparity is larger because they cause many cases in all social groups (15). Thus, the difficulty lies not in simply developing a causal model but in deciding what might be more and less "important" causal pathways for the question at hand. If the goal was to explain relative disparity in CHD, then the arrows on conventional risk factors would only be weighted about 30 percent, and there would be justification to examine pathways involving the multitude of other suspected risk factors. If the goal was to explain absolute disparity in CHD, then in some populations there would seem to be little justification to go beyond the four major risk factors.
|
CONCLUDING REMARKS |
|---|
|
TOPINTRODUCTIONABSOLUTE MEASURESIMPROVING INTERPRETABILITY:...CAUSAL MODELSCONCLUDING REMARKSReferences |
|---|
Finally, Messer (1) states that a major critique of our paper (2) is the failure to provide guidance as to how one might combine multiple disparity measures or on what basis to choose one measure over another. We have discussed these issues at greater length elsewhere (3, 4), but nevertheless we do, in fact, provide some guidance in our paper for selecting indicators, noting specifically (2, p. 889) that researchers who believe that disparity measures should reflect changes in the distribution of social groups over time should use population-weighted rather than unweighted measures. Researchers or policymakers with different purposes may take an approach that suggests other measures. For instance, someone focused on meeting the needs of a small minority group, for example, for a cancer control plan, may take an approach different from that of someone focused on national total burden of disease. The purpose of our article (2) was precisely to draw attention to the importance of answering such questions prior to undertaking any analysis. Ultimately, the suite of indicators that researchers use to measure disparity should depend on which aspects of disparity they think are important and not on which ones the data show to be most supportive of their favored hypotheses.
|
ACKNOWLEDGMENTS |
|---|
This project was carried out under contract with the National Cancer Institute (contract 263-MQ-611198).
The content of this publication does not necessarily reflect the views or policies of the National Cancer Institute.
Conflict of interest: none declared.
|
References |
|---|
|
TOPINTRODUCTIONABSOLUTE MEASURESIMPROVING INTERPRETABILITY:...CAUSAL MODELSCONCLUDING REMARKSReferences |
|---|
- Messer LC. Invited commentary: measuring social disparities in health—what was the question again? Am J Epidemiol (2008) 167:900–4.
[Abstract/Free Full Text] - Harper S, Lynch J, Meersman SC, et al. An overview of methods for monitoring social disparities in cancer with an example using trends in lung cancer incidence by area-socioeconomic position and race-ethnicity, 1992 –2004. Am J Epidemiol (2008) 167:889–99.
[Abstract/Free Full Text] - Harper S, Lynch J. Methods for measuring cancer disparities: a review using data relevant to Healthy People 2010 cancer-related objectives. In: NCI cancer surveillance monograph series, no. 6. (2005) Bethesda, MD: National Cancer Institute. (NIH publication no. 05-5777).
- Harper S, Lynch J. Selected comparisons of measures of health disparities: a review using databases relevant to Healthy People 2010 cancer-related objectives. In: NCI cancer surveillance monograph series, no. 7. (2008) Bethesda, MD: National Cancer Institute. (NIH publication no. 07-6281).
- Houweling TA, Kunst AE, Huisman M, et al. Using relative and absolute measures for monitoring health inequalities: experiences from cross-national analyses on maternal and child health. (Electronic article). Int J Equity Health (2007) 6:15.[CrossRef][Medline]
- Keppel KG. Ten largest racial and ethnic health disparities in the United States based on Healthy People 2010 objectives. Am J Epidemiol (2007) 166:97–103.
[Abstract/Free Full Text] - Wong MD, Shapiro MF, Boscardin WJ, et al. Contribution of major diseases to disparities in mortality. N Engl J Med (2002) 347:1585–92.
[Abstract/Free Full Text] - Massey DS, Denton NA. The dimensions of residential segregation. Soc Forces (1988) 67:281–315.[CrossRef][Web of Science]
- Reardon SF, Firebaugh G. Measures of multigroup segregation. Sociol Methodol (2002) 32:33–67.[CrossRef][Web of Science]
- Massey DS, White MJ, Phua VC. The dimensions of segregation revisited. Sociol Methods Res (1996) 25:172–206.[Abstract]
- Johnston R, Poulsen M, Forrest J. Ethnic and racial segregation in U.S. metropolitan areas, 1980 –2000: the dimensions of segregation revisited. Urban Aff Rev Thousand Oaks Calif (2007) 42:479–504.[CrossRef]
- Reardon SF, O'Sullivan D. Measures of spatial segregation. Sociol Methodol (2004) 34:121–62.[CrossRef]
- Iceland J, Weinberg DH, Steinmetz E, et al. Racial and ethnic residential segregation in the United States 1980–2000. (2002) Washington, DC: US Census Bureau.
- Hopkins PN, Williams RR. A survey of 246 suggested coronary risk factors. Atherosclerosis (1981) 40:1–52.[Medline]
- Lynch J, Davey Smith G, Harper S, et al. Explaining the social gradient in coronary heart disease: comparing relative and absolute risk approaches. J Epidemiol Community Health (2006) 60:436–41.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
Related articles in Am. J. Epidemiol.:
- An Overview of Methods for Monitoring Social Disparities in Cancer with an Example Using Trends in Lung Cancer Incidence by Area-Socioeconomic Position and Race-Ethnicity, 1992–2004
- Sam Harper, John Lynch, Stephen C. Meersman, Nancy Breen, William W. Davis, and Marsha E. Reichman
Am. J. Epidemiol. 2008 167: 889-899.[Abstract] [FREE Full Text]
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||