Associations between Individual and Community Factors and Type 2 Diabetes Onset in Pennsylvania

Society for Epidemiologic Research (SER) Annual Meeting, June 18-21, 2019 Minneapolis, MN

Annemarie G. Hirsch, PhD, MPH; Melissa  N. Poulsen, PhD, MPH; Jonathan Pollak, MS; Joseph DeWalle, BS; Amy Poissant, BA; Brian S. Schwartz, MD, MS

In Pennsylvania (PA), 12.8% of the adult population have diabetes and an additional 35.8% of the adult population have pre-diabetes. Every year, an estimated 71,000 people in PA are diagnosed with diabetes. We conducted a case-control study of type 2 diabetes (T2D) onset of primary care patients using electronic health record (EHR) data from 2001 to 2016 to evaluate associations between individual and community-level factors and T2D onset in 37 counties in PA. Cases were required to have at least two diagnosis codes for T2D; or at least one T2D medication order; or at least one T2D diagnosis code and an abnormal glycated hemoglobin or glucose value. To confirm new onset, cases had to have been observed for at least 2 years in the EHR prior to diagnosis. We selected 5 controls for each case, matched on sex age, and year of clinical encounter. We conducted logistic regression analysis of sex, age, race/ethnicity (non-Hispanic white, other), history of Medicaid (used for at least 1/2 of encounters, less than 1/2), body mass index (BMI, linear and quadratic), and rural/urban residential address (urban area, urban cluster, rural) on T2D onset. A total of 24,836 cases and 124,180 controls were identified. The following factors were associated with T2D onset (odds ratio: 95% confidence interval): non-white vs. white (1.72: 1.54, 1.93); Medicaid coverage for at least 1/2 of encounters vs. less than 1/2 (1.55 (1.42, 1.96); living in an urban area (1.20: 1.16, 1.25) or an urban cluster vs. rural (1.07: 1.03, 1.12); and BMI (1.39: 1.36, 1.39). Removing BMI from the model did not substantially change associations, providing some evidence that BMI is not a mediator of these associations. In a geographically diverse area of PA, individuals living in urban areas had higher odds of T2D onset than those in rural areas, adjusting for individual sociodemographic characteristics and BMI. This finding contrasts national studies that report higher T2D prevalence in rural areas