Leveraging Innovative Electronic Health Record Data to Characterize Social Determinants of Health Among Survivors of Cancer in Persistent Poverty Areas: Cross-Sectional Study

Background: Residents in persistent poverty areas experience higher cancer mortality due to social determinants of health that negatively affect multiple factors, including health behaviors. Objective: This study aimed to characterize demographic, clinical, and social determinant of health factors among survivors of cancer in persistent poverty areas using electronic health record (EHR) data-including an embedded social risk screener and natural language processing (NLP) of social work notes-to inform community-engaged adaptation of lifestyle interventions. Methods: EHR data from a large multispecialty group practice were extracted for patients with cancer residing in zip codes inclusive of persistent poverty areas targeted for a health behavior intervention and receiving care between January 2018 and November 2023. Self-reported social determinant of health data were obtained using the Protocol for Responding to and Assessing Patients' Assets, Risks, and Experiences (PRAPARE) and through NLP of social histories from a social work visit. Results: We identified 2672 unique patients with cancer, of whom 578 (21.6%) had PRAPARE data and 1597 (59.8%) had social history data available for analysis. The most common cancers among survivors (n=1420, 53.1% female; n=1299, 48.6% Black individuals; mean age 65.2, SD 13.7 years) included breast (n=536, 20.1%), prostate (n=400, 15%), and lymphoid or hematopoietic (n=323, 12.1%) cancer. Among survivors in persistent poverty areas (n=509, 19%; all with a high Social Vulnerability Index), 34.6% (176/509) were single, 55.4% (282/509) had Medicare coverage (with only 73/509, 14.3% having private insurance), 36.5% (186/509) had obesity, 63.9% (325/509) had hypertension, and 31.2% (159/509) had diabetes. Of survivors in persistent poverty areas with PRAPARE data, 15.8% (19/120) lacked transportation, 4.2% (5/120) lived with housing insecurity, and 6.7% (8/120) felt unsafe where they lived. Conclusions: Innovative EHR and NLP approaches identified several socioeconomic and safety-related challenges along with opportunities for health behavior interventions to leverage Medicare coverage and target multiple comorbidities when adapting interventions for survivors of cancer living in persistent poverty areas.