Neighborhood racial/ethnic segregation and BMI: A longitudinal analysis of the Multi-ethnic Study of Atherosclerosis

Background Current knowledge regarding the relationship between segregation and body weight is derived mainly from cross-sectional data. Longitudinal studies are needed to provide stronger causal inference. Methods We use longitudinal data from the Multi-Ethnic Study of Atherosclerosis and apply an econometric fixed-effect strategy, which accounts for all time-invariant confounders, and compare results to conventional cross-sectional analyses. We examine the relationship between neighborhood-level racial/ethnic segregation, neighborhood poverty, and body mass index (BMI) separately for blacks, Hispanics, and whites. Segregation*gender interactions are included in all models. Neighborhood segregation was operationalized by the local G(i)(*) statistic, which assesses the extent to which a neighborhood's racial/ethnic composition is under ( G(i)(*) statistic < 0) or over ( G(i)(*) statistic > 0) represented, given the composition in the broader (e.g., county) area. For black, Hispanic, and white stratified models, the G(i)(*) statistic reflects the level of black, Hispanic, and white segregation, respectively. The G(i)(*) statistic was scaled such that a unit change represents a 1.96 difference in the score. Results Cross-sectional models indicated higher segregation to be negatively associated with BMI for white females and positively associated for Hispanic females. No association was found for black females or males in general. In contrast, fixed-effect models adjusting for neighborhood poverty, higher segregation was positively associated with BMI for black females (coeff = 0.25 kg/m(2); 95% CI = [0.03, 0.46]; p-value = 0.03) but negatively associated for Hispanic females (coeff = -0.17 kg/m 2 ; 95% CI = [-0.33, -0.01]; p-value = 0.04) and Hispanic males (coeff = -0.20; 95% CI = [-0.39, -0.01]; p-value = 0.04). Further controls for socioeconomic factors fully explained the associations for Hispanics but not for black females. Conclusions Fixed-effect results suggest that segregation's impacts might not be universally harmful, with possible null or beneficial impacts, depending on race/ethnicity. The persistent associations after accounting for neighborhood poverty indicate that the segregation-BMI link may operate through different pathways other than neighborhood poverty.