Of the 382 million people worldwide with diabetes, and if current trends continue, nearly half a billion people worldwide will have diabetes by 2035. Two-thirds of current diabetics are living in urban centers and the urban concentration of individuals with diabetes is on the rise. The problem is that in the absence of widespread clinical testing, there is no reliable way to predict which segments of the population are the most vulnerable to the onset of diabetes. Knowing who the most vulnerable are, and where they live, can guide the efficient allocation of prevention resources. Toward this end, we introduce the concept of composite vulnerability, which includes both group and individual-level attributes, and we provide a demonstration of its application to a large urban setting. The components of composite vulnerability are estimated using a novel, population-based, procedure that relies on sample survey data and nonparametric statistical techniques. First, cluster analysis identified three multivariate profiles of adult residents with type 2 diabetes, based on 35 socioeconomic indicators. Second, the undiagnosed population was screened for vulnerability based on their resemblance or fit to these multivariate profiles. Geographic neighborhoods with high concentrations of "vulnerables" could then be identified. In parallel, recursive partitioning found the best predictors of type 2 diabetes in this urban population, combined them with indicators of disadvantage, and applied them to residents in the selected neighborhoods to establish relative levels of composite vulnerability. Neighborhoods with high concentrations of residents manifesting composite vulnerability can be easily identified for targeting community-based prevention measures.
Keywords: biological factors; cluster analysis; diabetes; socioeconomic factors.