Subpopulation Differences in Connecting to Resources to Address Social Needs.

Abstract

Evaluations of healthcare-based interventions to address social needs are flourishing, but few assess for which subpopulations such interventions are comparatively more effective than others. Yet, this is an additional lens for understanding what works and where improvements are needed. Ad hoc subgroup analyses were conducted in 2024-2025 using data from a 3-arm intent-to-treat RCT (2021-2023). The control arm (Arm 1) received a postcard listing generic Veteran Health Administration resources (e.g., National Call Center for Homeless Veterans). Intervention arms received the same plus tailored resource sheets (Arm 2) and tailored resource sheets plus navigation assistance from a social worker (Arm 3). The primary outcome was social need resource connection. Generalized linear models were used with random effects to account for intervention sites to conduct subgroup analyses on the basis of veteran sociodemographic and clinical characteristics (e.g., age, comorbidity, income). As-treated analysis was also conducted. Among veterans who were low income but ineligible for Veteran Health Administration disability benefits and veterans with ≥4 Elixhauser comorbidities, those randomized to Arm 3 were more likely to connect to ≥1 resource than those randomized to Arm 1 (OR=6.47, 95% CI=1.87, 22.37 and OR=2.18, 95% CI=1.16, 4.10, respectively). Intervention Arm 3 was also associated with higher odds of resource connection in the as-treated analysis than Arm 1 (OR=1.77, 95% CI=1.04, 2.99). Specific veteran subgroups may benefit most from interventions designed to connect patients to social need resources. Given limitations of adhoc analyses, it will be important for future study to explore the generalizability of these findings.