Background: Environmental attributes (e.g., greenness, air pollution, walkability) have been associated with type 2 diabetes (T2D) incidence, but this literature has notable biases (e.g., residential self-selection, ascertainment bias). We examine whether place-based T2D associations replicate across two complementary population health data sources and test whether BMI mediates these associations.

Methods: We used the Utah Population Database to follow a parent and offspring cohort residing in the four-county Wasatch Front region in Utah. Incident T2D was identified independently using the Healthcare Facility Database (HCFD; encounter data from all state-licensed hospitals/emergency departments/surgery centers; 1996-2019; n = 896,204) and the Utah All-Payer Claims Database (APCD; insurance claim data, including from Medicaid; 2013-2019; n = 550,275). To document the validity of each data source and potential limitations, we compared diagnosis timing and comorbidity profiles among cases observed in both sources, assessed screening-related ascertainment bias due to overweight status, and documented the magnitude of socioeconomic disparities. Cox models were used to estimate associations between environmental attributes and T2D and test body mass index as a mediator.

Results: Among 22,032 individuals diagnosed in both sources from 2013-2019, APCD recorded T2D 1.2 years earlier whereas HCFD diagnoses had more disease comorbidities (i.e., kidney disease in 9.1% vs 5.7% and cardiovascular disease in 24.1% vs 18.4%). Higher tract-level greenness and active commuting rate were protective (HR=0.94 per 0.1 NDVI; HR=0.96 per 10% active commuting rate), while greater PM10 increased risk (HR=1.10 per 10 ug/m3). BMI minimally mediated greenness (~5%) but accounted for at least half of the active commuting and PM10 associations.

Conclusions: Triangulating across health data sources and testing mechanisms can strengthen confidence in T2D evidence, informing potential place-based interventions.