High-Resolution Area-Level Measures as Predictors of Pediatric Individual-Level Social Risks

OBJECTIVE: Efficient social risk screening systems are needed to facilitate appropriate referrals and improve preventive care. This study evaluated whether high-resolution area-level measures could identify pediatric patients facing social risks and reduce clinic screening burden. METHODS: We analyzed electronic records for 5,237 pediatric patients who received care at a community safety-net hospital system between 2020 and 2022. Gold-standard social risk exposure was measured using a clinical screener. Three census block group-level predictors were evaluated: the Area Deprivation Index, Child Opportunity Index, and a measure of social risk prevalence generated through Bayesian aggregation of screening data. Receiver operating characteristic curve analysis identified optimal cutoffs for each measure when used as (1) a primary screener and (2) a prescreen to determine which families should receive full screening. Sensitivities, specificities, positive predictive values (PPV), and negative predictive values (NPV) were calculated overall and by language and race/ethnicity. RESULTS: All three area-level measures showed limited predictive value. As primary screeners, optimal cutoffs yielded PPVs of 0.23-0.28 and NPVs of 0.80-0.83, implying high false-positive rates and inefficient resource allocation. As prescreens, none reduced the proportion of children requiring direct screening by more than 10%. Predictive performance was poorer among Hispanic and Brazilian children, and among those receiving care in Spanish or Portuguese. CONCLUSION: In this safety-net pediatric population, high-resolution area-level measures could not reliably identify families facing social risks. While area-level measures are powerful tools for advancing public health equity in many contexts, direct, universal social risk screening remains essential in pediatric care.