Identifying high-risk neighborhoods using electronic medical records: A population-based approach for targeting diabetes prevention and treatment interventions

Abstract

Background Increasing attention is being paid to the marked disparities in diabetes prevalence and health outcomes in the United States. There is a need to identify the small-area geographic variation in diabetes risk and related outcomes, a task that current health surveillance methods, which often rely on a self-reported diagnosis of diabetes, are not detailed enough to achieve. Broad adoption of electronic health records (EHR) and routine centralized reporting of patient-level data offers a new way to examine diabetes risk and highlight hotspots for intervention. Methods and Findings We examined small-area geographic variation in hemoglobin A1c (HgbA1C) levels in three counties though a retrospective observational analysis of the complete population of diabetic patients receiving at least two ambulatory care visits for diabetes in three counties (two urban, one rural) in Minnesota in 2013, with clinical performance measures re-aggregated to patient home zip code area. Patient level performance measures included HgbA1c, blood pressure, low-density lipoprotein cholesterol and smoking. Diabetes care was provided to 63,053 patients out of a total population of 1.48 million people aged 18–74. Within each zip code area, on average 4.1% of the population received care for diabetes. There was significant and largely consistent geographic variation in the proportion of patients within their zip code area of residence attaining HgbA1C <8.0%, ranging from 59–90% of patients within each zip code area (interquartile range (IQR) 72.0%-78.1%). Attainment of performance measures for a zip code area were correlated with household income, educational attainment and insurance coverage for the same zip code area (all p < .001). Conclusions We identified small geographic areas with the least effective control of diabetes. Centrally-aggregated EHR provides a new means of identifying and targeting at-risk neighborhoods for community-based interventions.