AIM: Using AI to Improve Residential Racial and Economic Segregation for Mortality Analysis

Residential racial and economic segregation has been associated with an increased risk for main causes of death. Quintiles of the Index of Concentration and the Extremes (ICE) are commonly used to measure residential segregation because of their robust and comprehensiveness. However, the high correlation between ICE and the proportion of population in poverty can cause the collinearity, and this issue was avoided by ignoring poverty rate or transforming the continuous variable to the category variable in previous research, which induced the missing information because poverty was one significantly crucial covariant in different models for the association of residential segregation and health outcomes. In this paper we developed one new methodology named AIM by exploring artificial intelligence (AI) techniques to extract the integrated information from poverty. The original algorithm of this methodology utilized the high correlation between ICE and poverty rate to implement balanced clustering, by which ICE quintiles were improved and the stronger association of the AI based segregation with age-adjusted mortality rate was observed.