Is socioeconomic segregation of the poor associated with higher premature mortality under the age of 60? A cross-sectional analysis of survey data in major Indian cities

OBJECTIVES: Although urbanisation is generally associated with poverty reduction in low-income and middle-income countries, it also results in increased socioeconomic segregation of the poor. Cities with higher levels of socioeconomic segregation tend to have higher mortality rates, although the evidence is based on ecological associations. The paper examines whether socioeconomic segregation of the poor is associated with higher under-60 years (‘premature’) mortality risk in Indian cities and whether this association is confounded by contextual and compositional sociodemographic and socioeconomic factors. SETTING AND PARTICIPANTS: A population representative sample of over one million from 39 427 households living in 1876 urban wards within 59 Indian districts (cities) from the third (2008) District Level Household Survey (DLHS-3). PRIMARY OUTCOME AND OTHER MEASURES: The outcome was any death under the age of 60 reported by households in the preceding 4years of the DLHS-3. Socioeconomic segregation, estimated at the district (city) level, was measured using an isolation index of the poor and the index of dissimilarity. RESULTS: Poor households living in cities where the poor were more isolated had higher probabilities of premature mortality than poor households living in cities where the poor were less isolated. In contrast, it did not matter whether rich households lived in more or less socioeconomically segregated cities. A 1 SD increase in the isolation index was associated with an absolute increase of 1.1% in the probability of premature mortality for the poorest households. CONCLUSION: Increasing segregation of the poor may result in higher premature mortality. As low-income and middle-income countries become increasingly urbanised, there is a risk that this may lead to increased segregation of the poor as well as increased premature mortality.