Augmenting electronic health record data with social and environmental determinant of health measures to understand regional factors associated with asthma exacerbations

Abstract

Electronic health records (EHRs) provide rich data for diverse populations but often lack information on social and environmental determinants of health (SEDH) that are important for the study of complex conditions such as asthma, a chronic inflammatory lung disease. We integrated EHR data with seven SEDH datasets to conduct a retrospective cohort study of 6,656 adults with asthma. Using Penn Medicine encounter data from January 1, 2017 to December 31, 2020, we identified individual-level and spatially-varying factors associated with asthma exacerbations. Black race and prescription of an inhaled corticosteroid were strong risk factors for asthma exacerbations according to a logistic regression model of individual-level risk. A spatial generalized additive model (GAM) identified a hotspot of increased exacerbation risk (mean OR = 1.41, SD 0.14, p < 0.001), and inclusion of EHR-derived variables in the model attenuated the spatial variance in exacerbation odds by 34.0%, while additionally adjusting for the SEDH variables attenuated the spatial variance in exacerbation odds by 66.9%. Additional spatial GAMs adjusted one variable at a time revealed that neighborhood deprivation (OR = 1.05, 95% CI: 1.03, 1.07), Black race (OR = 1.66, 95% CI: 1.44, 1.91), and Medicaid health insurance (OR = 1.30, 95% CI: 1.15, 1.46) contributed most to the spatial variation in exacerbation odds. In spatial GAMs stratified by race, adjusting for neighborhood deprivation and health insurance type did not change the spatial distribution of exacerbation odds. Thus, while some EHR-derived and SEDH variables explained a large proportion of the spatial variance in asthma exacerbations across Philadelphia, a more detailed understanding of SEDH variables that vary by race is necessary to address asthma disparities. More broadly, our findings demonstrate how integration of information on SEDH with EHR data can improve understanding of the combination of risk factors that contribute to complex diseases.