Factors affecting exits from homelessness among persons with serious mental illness and substance use disorders

Abstract

Objective: We sought to understand the housing trajectories of homeless consumers with serious mental illness (SMI) and co-occurring substance use disorders (SUD) and to identify factors that best predicted achievement of independent housing.

Method: Using administrative data, we identified homeless persons with SMI and SUD admitted to a residential rehabilitation program from December 2008 to November 2011. Our primary outcome measure was independent housing status. On a random sample (N = 36), we assessed a range of potential predictors of housing outcomes, including symptoms, cognition, and social/community supports. We used the Residential Time-Line Follow-Back (TLFB) Inventory to gather housing histories since exiting rehabilitation and to identify housing outcomes. We used Recursive Partitioning (RP) to identify variables that best differentiated participants by these outcomes.

Results: We identified 3 housing trajectories: stable housing (n = 14), unstable housing (n = 15), and continuously engaged in housing services (n = 7). In RP analysis, 2 variables (Symbol Digit Modalities Test [SDMT], a neurocognitive speed of processing measure, and Behavior and Symptom Identification Scale [BASIS-24] Relationships subscale, which quantifies symptoms affecting relationships) were sufficient to capture information provided by 26 predictors to classify participants by housing outcome. Participants predicted to continuously engage in services had impaired processing speeds (SDMT score < 32.5). Among consumers with SDMT score ≥ 32.5, those predicted to achieve stable housing had fewer interpersonal symptoms (BASIS-24 Relationships subscale score < 0.81) than those predicted to have unstable housing. This model explains 57% of this sample's variability and 14% of this population's variability in housing outcomes.

Conclusions: Because cognition and symptoms influencing relationships predicted housing outcomes for homeless adults with SMI and SUD, cognitive and social skills training may be useful for this population.

Figure 1

Wilson and Cleary Conceptual Framework (adapted)

Figure 2. Results of recursive partitioning analyses for classifying consumers by housing trajectory

Noted below each housing trajectory are the number of participants with stable housing (S), unstable housing (U), and continuous engagement in housing programs (P). C = percentage of participants this model classified correctly. Estimation was based on 36 participants and 26 predictor variables, including: •Lifetime jail/prison days; Felony (yes/no); AUDIT-C (positive/negative); DAST-20 (positive/negative); BASIS-24® (overall score; depression and functioning; interpersonal problems; psychotic symptoms; alcohol or drug use; emotional lability; self-harm); PCLM; HVLT-R; SDMT; category fluency; TMT B; VR-12 (physical health component score, mental health component score); CIS (score out of 10 activities, overall involvement score); Social Capital Resource Generator (social network score, personal resources score); CIM; MOS-SSS; PDD; Number of outpatient mental health visits (past 6 months).