Abstract
Context: Disease burden may vary substantively across neighborhoods in an urban setting. Yet, data available for monitoring chronic conditions at the neighborhood level are scarce. Large health care data sets have potential to complement population health surveillance. Few studies have examined the utility of health care data for neighborhood-level surveillance.
Objective: We examined the use of primary care electronic health records (EHRs) and emergency department (ED) claims for identifying neighborhoods with higher chronic disease burden and neighborhood-level prevalence estimation.
Design: Comparison of hypertension and diabetes estimates from EHRs and ED claims with survey-based estimates.
Setting: Forty-two United Hospital Fund neighborhoods in New York City.
Participants: The EHR sample comprised 708 452 patients from the Hub Population Health System (the Hub) in 2015, and the ED claim sample comprised 1 567 870 patients from the Statewide Planning and Research Cooperative System in 2015. We derived survey-based estimates from 2012 to 2016 Community Health Survey (n = 44 189).
Main Outcome Measure: We calculated hypertension and diabetes prevalence estimates by neighborhood from each data source. We obtained Pearson correlation and absolute difference between EHR-based or claims-based estimates and survey-based estimates.
Results: Both EHR-based and claims-based estimates correlated strongly with survey-based estimates for hypertension (0.91 and 0.72, respectively) and diabetes (0.83 and 0.82, respectively) and identified similar neighborhoods of higher burden. For hypertension, 10 and 17 neighborhoods from the EHRs and ED claims, respectively, had an absolute difference of more than 5 percentage points from the survey-based estimate. For diabetes, 15 and 4 neighborhoods from the EHRs and ED claims, respectively, differed from the survey-based estimate by more than 5 percentage points.
Conclusions: Both EHRs and ED claims data are useful for identifying neighborhoods with greater disease burden and have potential for monitoring chronic conditions at the neighborhood level.