Multilevel Exploration of Health Disparity and Segregation

Objectives: To employ a recently described multilevel design to determine whether the risk of premature mortality is explained by between-city differences in racial segregation, and whether this effect is due to the grouping of neighborhood-specific risk in racially segregated cities. Methods: A multilevel analysis of Los Angeles County premature mortality data for the years 1989 through 1991 was modeled as the proportion of deaths in 24,826 cells (deaths and denominators cross-tabulated by age, gender, and race/ethnicity) at level 1 with 1,552 census tracts at level 2, and 105 cities at level 3. Racial segregation was characterized as black isolation. Neighborhood-specific risk was characterized as contextual (alcohol outlet density) and compositional (percent in poverty, family structure, and education) exposures across census tracts. Results: Accounting for the between-city and between-census tract variance in premature mortality reduced the odds of mortality for blacks compared to whites (adjusted Odds Ratios, aOR=1.59 fixed model; aOR=0.83 random model). Comparing cities in the lowest and highest quartiles for Black isolation was associated with an increase in mortality risk (aOR = 1.74; 95% Confidence Interval, CI= 1.16, 2.13). Including census tract exposures reduced but did not eliminate the effect (aOR = 1.48; 95% CI 1.33, 1.65). Conclusions: Accounting for the between-city and between-census tract level variance in premature mortality explained health disparities for blacks in Los Angeles County. Partial mediation of black isolation's effect on mortality by risk exposures at the census tract level supports a model of institutional racism to account for health disparities. Likely endogenous effects limit interpretation.