Modeling service access in a homeless population

Previous research has shown use of greater amounts and more types of services to be robustly associated with positive outcomes. However, research has neither adequately explored nor conceptualized the complex interactions among demographics, homelessness, mental illness, and service use. Subjects (N=396) were systematically sampled. Service use was divided into four sectors: homeless amelioration, homeless maintenance, mental health, and chemical dependency services. Three nested conceptual models were analyzed using logistic regressions predicting service use for each service sector (30 day and lifetime): (1) need factors, (2) need factors, comorbidity, and demographics; and (3) all of the above plus other sector service use. Improvements in predictive ability were tested for each pair of nested equations. A fourth set of analyses including all variables used in any analysis was tested for further improvement. In most analyses, the third conceptual model demonstrated significant advantage over simpler models. Models including all variables did not add significant improvement. Mental health and addiction service use were relatively strongly predicted by diagnosis. Despite limitations, the study has implications for service use: (a) services should facilitate individual level cross-sector use; (b) homelessness sectors require attending to need factors; (c) service providers should be aware of barriers unintentionally imposed by service restrictions.