Individual Submission Summary
Share...

Direct link:

How do social needs cluster? Evidence from Medicaid beneficiaries in New York before 1115 waiver implementation

Thursday, November 5, 1:45 to 3:15pm, Property: Boston Marriott Copley Place, Floor: 3rd Floor, Room: Clarendon

Abstract

New York State’s Section 1115 waiver demonstration aims to address Medicaid beneficiaries’ health-related social needs (HRSNs). HRSNs are an individual’s unmet, adverse social conditions that contribute to poor health (e.g. housing instability, food insecurity). HRSNs tend to cluster together, particularly among low-income populations, compounding disease burden and driving avoidable health care utilization. Existing research on HRSN clustering relies on screening instruments and self-reported surveys; ICD-10-CM Z-codes in Medicaid claims offer an alternative for the full population but under-document HRSNs relative to screening. This study uses 2024 claims from before the state’s 1115 waiver implementation to describe documented HRSNs among Medicaid beneficiaries in New York state, identify distinct subgroups of beneficiaries with co-occurring documented HRSNs, and compare them with classes found in screening-based research.

We analyzed 2024 Medicaid claims data, including fee-for-service claims and managed care encounters, from New York State. We identified HRSNs by extracting claims with at least one ICD-10-CM diagnosis code in the Z55.x-Z65.x range and mapping individual Z-codes to the following domains: Social and Emotional Support, Housing, Employment, Financial Insecurity, Food, Transportation, Education, Unspecified Economic, Legal, and Interpersonal Safety. We characterized pairwise co-occurrence using phi coefficients. Among beneficiaries with two or more documented domains, we used latent class analysis (LCA) to identify distinct subgroups of beneficiaries based on their patterns of HRSN co-occurrence, selecting the number of classes using BIC, entropy, and interpretability.

Only 4.1% of 8.9 million beneficiaries had at least one documented HRSN. Among the ten domains, Social and Emotional Support was the most-commonly documented, followed by Housing and Employment.

About one third of beneficiaries with any HRSN documented had more than one HRSN. The LCA identified five distinct subgroups among these beneficiaries: Social Support (30.0%), defined by universal social and emotional support needs with low-level needs across various domains; Unspecified Economic and Transportation (26.9%); Financial-dominant (20.9%); Employment and Housing (13.4%); and Complex Polyneeds (8.7%), with high to moderate probabilities across nearly every domain. The subgroups differed in their demographic and clinical characteristics — particularly on age, race/ethnicity, geography, and the prevalence of serious mental illness and substance use disorder. Complex Polyneeds, Financial-dominant, and Employment and Housing correspond to archetypes reported in existing screening-based research, whereas Social Support and Unspecified Economic and Transportation have no clear screening analog.

HRSNs documented in Medicaid claims cluster into distinct subgroups; some classes mirror screening-based archetypes while others appear to reflect local coding practices, such as the concentration of Complex Polyneeds in New York City. Interventions tailored to distinct clusters of co-occurring HRSNs may be needed to effectively address the heterogeneous social needs of Medicaid beneficiaries. As New York's NYHER 1115 waiver moves into full implementation, this typology of co-occurring HRSNs can inform how Social Care Networks prioritize and tailor services across the state.

Authors