Spatial association and vulnerability of obesity prevalence on county-level diabetes prevalence in the United States from 2004-2011

Over the past three decades, the prevalence of diagnosed diabetes in the U.S. increased from 2.5% to 6.9%. Clinical and epidemiological research has reported a strong association between diabetes and obesity. However, it is unknown whether the reason for this increase in diabetes prevalence is caused by geographic disparities of obesity. A spatiotemporal study was conducted to determine whether obesity prevalence is spatially associated with diabetes prevalence, and to identify whether areas with higher obesity prevalence are more likely clustered in some specific regions in the U.S. We analyzed adjusted prevalence data estimated from the CDC in 3,109 counties across 48 contiguous states from 2004 to 2011, while controlling for socioeconomic status variables from the American Community Survey. A Bayesian structured additive regression (STAR) modeling approach was applied to estimate the spatially interactive impact of obesity prevalence on diabetes prevalence, adjusting for socioeconomic conditions, health insurance coverage, and smoking rate at the county-level. An advanced STAR model was also built to accomplish spatial comparisons among categorized obesity prevalence. The spatial function in our STAR models was used to analyze county’s boundary data, and carried out by Markov random fields with a conditional autoregressive prior. We reveal that approximately 98.81% counties had a relative risk (RR) of diabetes significantly greater than 1, where greater RR concentrated in Southeast, Central, and South Regions. After categorizing obesity prevalence into quartiles, using the lowest level of obesity prevalence as the reference group, there are still 41.04% counties vulnerable to diabetes because of the low level of obesity prevalence. Median-high and high levels of obesity prevalence were expanded and clustered in counties in Southwest, Northwest and West Regions. In conclusion, this study identified vulnerable areas for diabetes in the U.S. and provides the possibility of establishing targeted surveillance systems to raise awareness of diabetes in areas with higher obesity prevalence.