Visibly Homeless with Less Visible Needs: Concordance of Medicaid & HMIS Indications of Serious Mental Illness & Substance Use
*Names in bold indicate Presenter
The opportunity: This paper draws on a statewide, linked, longitudinal Medicaid and homelessness services administrative records (HMIS) dataset (2014-2016), which offers a unique opportunity to empirically assess how often, under what circumstances, and for whom housing caseworkers identify mental health (MH), SUD, or both among chronically homeless adults known to have these conditions. The dataset was constructed by the Rutgers Center for State Health Policy, with support from The Nicholson Foundation and the New Jersey Division of Medical Assistance & Health Services, as part of a broader study of the Medicaid service use and spending among persons using homeless services in New Jersey.
The research questions:
- How thoroughly are behavioral health conditions (MH, SUD) recorded in the housing services encounter administrative records (HMIS)?
- How does the likelihood of HMIS recording of behavioral health conditions vary by client demographics, history of homelessness, use of homeless services, other health conditions, and health utilization/spending?
The analysis: The linked administrative dataset enables us to identify a subset of participants with both of these qualifying eligibility conditions for supportive housing services—chronic homelessness and a disabling condition.
From longitudinal HMIS records we identify subgroups of participants who are either:
- Already identified in HMIS as chronically homeless (N= 849)
- Has a disability and homeless history that likely qualifies them as chronically homeless but not identified as such in the HMIS (N= 1,355)
- At-risk of chronic homelessness, etc. (N= 2,160)
We further divide each of these homeless subgroups into four disability subgroups, based on Medicaid claims-derived indicators of the following:
- Any serious mental illness (SMI)
- Any substance use disorder (SUD)
- Both SMI & SUD
- Neither SMI nor SUD
For each subgroup we assess how often and for whom housing services case managers identified the condition. Next, we employ logistic regression models to predict identification based on participant demographic characteristics, current housing situation, other identified health barriers, and Medicaid claims-derived indicators of eligibility group, cumulative health services use, and cumulative health services spending.
Policy Implications: Chronically homeless adults with SMI or SUD cannot be placed into PSH unless these conditions are accurately identified. Our analysis offers the first empirical evidence of how often housing case managers identify SMI and SUD, under what circumstances, and for which types of participants. In doing so, it illustrates how better information sharing between Medicaid and homelessness services could improve the effectiveness and equity of services to some of the most vulnerable homeless adults.