Uncovering geographical clustering of childhood obesity and childhood obesity disparities using hierarchical spatial models

Despite well documented race/ethnic disparities in childhood obesity, and studies suggesting differences in geographical clustering of obesity prevalence by race/ethnicity, no studies have rigorously tested if geographic patterns in obesity prevalence differ significantly by race/ethnicity owing, in part, to the lack of methods to do so. Significant race/ethnic differences in the geographical patterning of obesity trends can shed light into salient environmental causes of persisting disparities. We present multi-level logistic regression models with spatially-autocorrelated random effects to study the geographical patterning of obesity prevalence among children in the State of California, and to examine if geographical patterning in race-ethnic disparities in obesity rates exists. Although the models are complex, model results can be displayed using existing mapping methodologies and are therefore friendly to a wide variety of users.