Existing evidence demonstrating a relationship between racial residential segregation and health has been based on aggregate analysis. Using a multilevel analytical framework, we assess the extent of geographic variation in black/white disparities in self-rated health across US metropolitan areas, and whether racial residential segregation accounts for such variation. We estimated multilevel regression models of poor self-rated health among 51,316 non-Hispanic white and non-Hispanic black adults nested within 207 metropolitan areas to assess the multilevel relationship between segregation and racial disparities in health. We found statistically significant variation in the black/white disparity in poor self-rated health across metropolitan areas, after controlling for individual level factors (age, sex, marital status, education and income) and residential segregation. High black isolation was associated with increased odds of reporting poor health among blacks (p<0.05). While a similar pattern was observed for white/black dissimilarity and white isolation, they were not statistically significant. Our multilevel analysis only partially supports the previously reported aggregate findings linking segregation to health. Additional multilevel statistical investigations across different health outcomes are required to draw firmer conclusions regarding the adverse effects of segregation on health.