An analysis of marginalization, substance abuse and longitudinal service use among homeless adults using a Bayesian framework

Background: This NIDA funded study examined the complex longitudinal relationships between housing status, substance abuse, and service use among 400 homeless adults. The primary aim was to better understand how urban homeless adults engage in emergency and routine services based on changing indicators of marginalization, substance abuse, and housing status. Methods: Survey data from three interviews was matched with administrative service data collected from local and regional health, mental health, housing, and substance abuse service providers. To explore the proposed relationships and accurately account for high levels of over dispersion and zero inflation in the service count data, a flexible Bayesian framework was adopted, using a zero inflated Poisson Gamma model structure. Results: Key findings include consistent relationships between living on the streets, alcohol use, and increased marginalization, notably legal problems, shadow work, and victimization. The prevailing assumption that more marginalized individuals (legally, socially, and economically) would use more emergency services was not supported. Living on the streets and engaging in shadow work together predicted lower levels of both routine and emergency types of service use. In a similar manner, social isolation (i.e. not being able to count on family for help) predicted lower rates of routine service use. Discussion: These results underscore the importance of addressing legal problems, economic, and safety issues at a community level for homeless adults rather than simply focusing on adding more service options. Additional results and implications are discussed as are the benefits of Bayesian modeling strategies in exploring complex social problems.