Abstract 017: Impact of Gentrification on Cardiovascular Disease Surveillance Using Data from the Electronic Health Record

Introduction: Current static approaches of CVD surveillance may not capture the true health of neighborhoods as the influx of younger, healthier, and wealthier individuals (i.e., gentrification) can impact long term residents (LTR) who can be older and of lower socioeconomic status (SES), potentially biasing prevalence estimates. Hypothesis: LTR in gentrifying vs. LTR in non-gentrifying neighborhoods will have a lower prevalence of CVD related conditions. Methods: We quantified the census tract prevalence of diabetes (DM), hypertension (HTN), obesity, and CVD (myocardial infarction or stroke hospitalization) using electronic health record (EHR) data from the Duke Health System and Lincoln Community Health Center from 2008-2010 and 2014-2016. The EHR population was 116,760 patients living in Durham County in 2008 followed until 12/31/16. LTR were patients with the same 2010 and 2015 address; patients moving in/out (i.e., movers) did not have a recorded address in 2010 or 2015. A census tract was defined as gentrifying if 3 of 4 SES indicators (positive z-score for median household income, median rental price, % with bachelor’s degree, or a negative z-score for % living below the poverty line) improved. In difference-in-difference (DiD) analysis, we compared changes in prevalence of CVD health indicators between LTR in gentrifying and non-gentrifying neighborhoods. Results: At baseline, patients had a median age of 43 years, 42% were black and 61% were female. The overall prevalence of DM, HTN, obesity, and CVD was 12%, 32%, 19%, and 5% in 2010 & 14%, 31%, 32%, and 7% in 2016. The prevalence increased in gentrifying and non-gentrifying neighborhoods over time and was higher in LTR than comparable movers. Estimates from DiD models were not statistically significant (Figure 1). Conclusions: LTR vs. movers and not gentrification impacted the surveillance of CVD health indicators. Novel public health informatics methods are needed to address dynamic changes in neighborhoods that may bias the enumeration of CVD health indicators.