Development of Perinatal Social Determinants of Health Composite Risk Score in Infants Born Preterm.

Abstract

To develop and validate a perinatal social determinant of health (SDoH) risk score for predicting neurodevelopmental outcomes in infants born preterm. Prospective cohort study of 395 infants born at ≤32 weeks' gestation recruited from 5 level III/IV NICUs in Greater Cincinnati (2016-2019); 328 (83%) with complete data were analyzed. Five prenatal social risk factors-maternal age, education, household income, employment, and family structure-were combined into composite SDoH scores using supervised and unsupervised machine learning. Outcomes were Bayley-III cognitive, language, and motor composite scores at 2 years corrected age; neurodevelopmental impairment was defined as ≥1 subscale <85. Among 328 infants (median gestational age 29.6 weeks), 52.1% of mothers had a college degree or higher, 47.3% were low income, and 29.3% were single parents. The supervised ridge model explained R²=0.19 (SD 0.13) and the unsupervised multiple correspondence analysis (MCA) explained R²=0.20 (SD 0.14) of cognitive variance. Infants in the highest-risk tertile scored 13.1 points lower in cognition than those in the lowest-risk tertile, with similar gradients for language (19 points) and motor outcomes (7.8 points). Cross-domain consistency was high (r=0.97-1.00). Temporal validation at age 5 (n=291) showed correlations of 0.33-0.35 with general cognitive ability (all p<.001). Discrimination for impairment was moderate (AUC 0.67-0.68). This validated perinatal SDoH score quantifies cumulative social disadvantage in infants born preterm and supports early identification, targeted intervention, and standardized case-mix adjustment across NICUs.