An Empirical Analysis on Geographic Variations in the Prevalence of Diabetes

Abstract

We investigate how the regional prevalence of diabetes is affected by health-related and socioeconomic factors with a special emphasis on geographic variations. We focus on the likelihood of diabetes as function of various region-specific attributes. We construct a unique set of data at the level of 230 small administrative district collected from 2011 Annual Community Health Survey by Korea Centers for Disease Control and Prevention and other government agencies. To estimate, we use several methods including correlation analysis, multiple regression and decision tree model. We find that diabetes prevalence is more likely to be associated with hypertension prevalence, obesity and economic activity rate. Further findings using decision tree model suggest that hypertension prevalence, obesity, the rate of drinking and the density of population are more likely to affect the prevalence of diabetes than other regional attributes considered. More importantly, we find significant geographic variations in factors affecting diabetes prevalence across administrative districts. This topic are not just of academic interest, but have practical implications by helping policy makers to understand and identify important regional factors relating to the prevalence of diabetes by which they can implement more effective planning for promotion of health.