Mortality disparities between Black and White Americans mediated by income and health behaviors

Race disparities in health outcomes including mortality risk are well known, but mediating mechanisms that link race to mortality risk have rarely been formally tested.

We analyzed public NHANES III data from 1988 to 1994 linked to mortality outcomes prospectively through 2015. Participants included 10,460 non-Hispanic Black (40.5%, n = 4233) and non-Hispanic White (59.5%, n = 6227) adults. Proportional hazards regression models examined mortality risk in association with race, demographics, income, and an index of risky health behaviors including smoking, poor diet and low physical activity. A mediation approach under the counterfactual framework was used to test effects of income and risky health behaviors as mediators between race and mortality risk.

Considering only race, age and sex, Black participants had significantly higher mortality risk than Whites (HR = 1.46, 95% CI 1.35-1.58). When income and education were added, the race effect was lower but remained significant (HR = 1.15, 95% CI 1.02-1.30). In the subsequent model that also included risky behaviors the association between race and mortality was no longer significant (HR = 1.05, 95% CI 0.92-1.20); both higher income and healthier behaviors contributed to lower mortality risk. There was a significant indirect effect of race on mortality mediated through income, and the direct effect of race on mortality was not significant when the mediating effect of income was considered. Likewise, the risky behavior score significantly mediated the association between race with mortality, and the direct effect of race was not significant. In the separate models, income mediated 62% of the association between race and mortality and lifestyle mediated 61% of the relationship.

Efforts to reduce race-based mortality disparities may focus on policies to reduce income-based disparities and promote positive health behaviors that consider variations in socioeconomic resources and personal preferences.