EHR-Documented vs. Census-Derived Social Determinants for Type 2 Diabetes Outcomes

Health systems increasingly collect patient-reported social determinants of health (SDoH) at the point of care, motivated by the promise of more equitable clinical decision-making. But it remains unclear whether these EHR-documented measures actually capture something distinct from, or additive to, what Census-linked socioeconomic data already provide.

Using the UNC Health type 2 diabetes cohort, this project compares the predictive contribution of EHR-documented SDoH against Census-derived areal socioeconomic measures for two glycemic outcomes: continuous HbA1c and poor glycemic control (HbA1c 9% or above). Models are estimated across the full cohort and stratified by race and ethnicity to assess whether the relative value of each data source varies across subgroups. This work is in collaboration with Kahkoska and Vitale, and a manuscript is under revision.