Poverty and mortality in hemodialysis patients.

A vexing observation among individuals with ESRD is the decreased mortality associated with attributes that, in healthy populations, confer increased risk of death. Examples include the apparent protective benefit of obesity, higher levels of LDL cholesterol, elevated systolic anddiastolicBP, and increased levels of parathyroid hormone. Potential explanations for this survival paradox include unmeasured confounders, misspecification of exposures, selective survival with informative censoring, selection bias due to conditioning on a diseased population, and a true, unbiased protective benefit of the otherwise harmful attribute. Self-reported black race is an interesting example of this conundrum. As summarized by Kimmel and colleagues in this issue of JASN, increased risk of mortality is consistently reported for black populations in the United States, a disadvantage that stands in stark contrast to their longer survival after the onset of hemodialysis. The authors sought to clarify this unexpected survival advantage by exploring the degree to which spatial measures of material disadvantage and racial segregation might mediate these survival differences. Material disadvantage was estimated using the Gini coefficient for a personal income in a county, whichmeasures of the degree of inequality of the income distribution in the population. It varies from 0, which represents perfect equality of income distribution, to 1, which represents maximal income inequality. The race-specificmedian income of an individual patient’s zip code of residence was used as a surrogate for individual income. A third measure, the dissimilarity index, was used to measure racial segregation in the county of residence. The index varies from 0, an equal distribution of blacks and whites in a county, to 1, indicative of a racially homogeneous population. The degree of racial segregation was used to assess perceived discrimination and related stress that an individual might experience in a community. It was hypothesized that these spatial measures would explain part of the observed excess mortality observed inwhite, comparedwith black, ESRDpatients. The results of this interesting article are provocative and instructive. First, neither measures of income inequality nor racial segregation attenuated the reduction in mortality among blacks compared with whites. After controlling for individual risk factors at the start of ESRD, including two surrogates for individual socioeconomic status (health insurance status and employment status), blacks were 73% less likely to die during follow-up (hazard ratio [HR], 0.73; 95% confidence interval [95% CI], 0.72, 0.75). Serial addition of income inequality and segregation measures actually accentuated the protective benefit of race, and after accounting for individual-level attributes and spatial characteristics, blacks were 30% less likely to die (HR, 0.70; 95%CI, 0.69, 0.71) during follow-up. Race-specificmedian income, income inequality, and segregation were independently associated with mortality after controlling for race and other covariates. Finally, in multivariable models stratified by race, there was a graded, inverse association betweenmedian zip code income and mortality such that HRs were lower as income level increased in both races. Taken together, these results lead to the conclusion that community measures of income, income inequality, and racial segregation as measured in this study cannot explain racial differences in survival among ESRD patients. This interesting article raised important additional issues. Mortality in blacks and whites appears to be influenced differently by income inequality and racial segregation, because statistically significant interactions between race and both county income inequality and residential segregation were found. Kimmel et al. found that higher levels of income inequality were associated with higher mortality among whites but not blacks, whereas higher levels of racial segregation were associated with substantially higher mortality in blacks and not whites. Although such interactionmay be artifactual, it may also signal modification of mortality risk among blacks and whites by unappreciated mechanisms that should be further studied. The effect of income inequality on mortality in blacks is of similar magnitude as that reported by a recent meta-analysis. These meta-analytics support a posited threshold effect for income inequality, with Gini values ,0.3 less likely to confer increased risk of adverse outcomes. It is interesting to speculate this may be one mechanism through which race-specific income inequality may have increasedmortality risk inwhites, but not blacks, in this study. Because interactions may exist between individual and county measures of income inequality that could not be examined, it is possible that black/white mortality differentials may be modified by individual incomes and these effects might differ by county income inequality. Another consideration is the complexity inherent in measures of income inequality like the Gini index. The influence of income inequality onmortality may bemediated by either direct effects Published online ahead of print. Publication date available at www.cjasn.org.