Validity of Area-Based Social Risk Indices Used at Higher-Level Geographies and Clinic Locations

Importance: Area-based social risk indices allow researchers to analyze social exposures when individual measures are not available. However, small area location data are absent from many research datasets, which motivates researchers to use higher-level geographies and clinic addresses for social risk measurement. Objective: To assess the suitability of using social risk indices calculated with higher-level geographies and clinic addresses. Design, Setting, and Participants: This cohort study used primary care patient records from those seen at clinics included in the American Family Cohort from 2019 through 2021. Records were linked to area-based social risk indices using either patient's residential address or their primary care clinic's address in a cross-sectional cohort design that included geographic levels ranging from Census block group to 3-digit zip code tabulation area (ZCTA). Data were analyzed from December 2025 to February 2026. Exposures: Three area-based social risk indices: Reproducible Area Deprivation Index (ReADI), Social Deprivation Index (SDI), and Social Vulnerability Index (SVI). Main Outcomes and Measures: Using indices calculated with patient block group as the reference, correlation of indices across geographic levels and address types (residence vs clinic) were compared, as were the indices' associations with chronic kidney disease, hypertension, and diabetes. Results: We included 2 995 479 patients (1 667 673 [55.7%] female, 952 227 [34.0%] rural) seen at 809 clinics. Correlation of indices at the reference level (ie, block group) with the same index at the 3-digit ZCTA ranged from 0.34 to 0.48. ReADI had the highest correlation with all other geographic levels. Indices at clinic addresses were uncorrelated with patient reference values. Adjusted odds ratios (ORs) for reference indices and diabetes were between 1.21 (95% CI, 1.19-1.23) and 1.37 (95% CI, 1.34-1.40). ORs for other diseases were similar. These associations were primarily lower when higher-level geographies and clinic addresses were used. Conclusions and Relevance: In this cohort study of the association between patient outcomes and area-based social risk indices at different geographic levels and using clinic location as a proxy for home address, indices calculated using county and 3-digit ZCTA data available in popular claims databases generally misrepresented patient risk at more granular area levels of Census block group and tract. Clinic addresses were not useful for measuring patients' social risks.