Introduction Segregation means enforced separation of disadvantaged social groups. The combination of poverty and spatial segregation favours the reproduction of poverty and affects the health of segregated groups as well the health of the population. A number of dimensions of segregation can now be measured through indices of spatial clustering, isolation and exposure to other social groups.

Objectives To analyse the association between socioeconomic urban segregation and population health in Brazil.

Methods This is a cross-sectional and ecological study. Secondary data for 20 of the biggest Brazilian cities was obtained from the census and from the national information systems. The district level health outcome variables were cardiovascular, cancer, external causes and total mortality rates. The explanatory variables were: census tract and district level scores of several spatial and local socio-economic segregation indices, and income. Regression analysis, testing for spatial autocorrelation, and spatial regression were used to check the association between segregation indices and health outcomes.

Results Districts with the poorest health were also districts with the highest segregation of the poorest groups. In contrast, districts with the best health were ones where the rich were isolated and lacked exposure to the poor.

Conclusions Within Brazilian cities, the disadvantaged social groups are spatially segregated. The data suggest that segregation is bad of the health of poor districts and good for the health of the rich districts. This process of segregation leading to divergent health outcomes depending on the socioeconomic profile of communities may intensify health inequalities.