Mapping and Testing Spatial Clusters of Diabetes in the U.S.

Ronald E. Cossman, Mississippi State University
Jeralynn S. Cossman, Mississippi State University
Wesley James, Mississippi State University
Troy Blanchard, Mississippi State University
Richard Thomas, University of Tennessee Health Science Center
Louis Pol, University of Nebraska, Omaha
Erdenechimeg Eldev-Ochir, Mississippi State University

Geographic variations in chronic illness prevalence could create new analysis issues. We use prescriptions-filled at the county-level as a proxy for the prevalence of diabetes in the resident population. We map diabetes prescription rates and there appears to be geographic clusters of high and low prescription rate counties. We use two spatial statistical tests to confirm these patterns. The Global Moran’s I (a test for spatial autocorrelation across all counties) is 0.603 (0 = no spatial correlation and 1 = perfect correlation), thus there is statistically significant spatial autocorrelation in diabetes prescriptions-filled rates across counties. We then map the Local Moran’s I for high and low spots and find numerous patterns of self-defining clusters. Low prescription rates are clustered in the West, while high prescription rates are clustered in the upper Mid-West, Appalachia and the Carolinas. The presence of spatial autocorrelation suggests that non-spatial models may produce biased results.