Identifying Clusters of Risk Behavior and Change in Homeless Adults Transitioning to Permanent Supportive Housing

Background: Permanent supportive housing (PSH) has empirically demonstrated success in housing and retention for individuals with active substance use and high risk. Despite this, the use of a harm reduction approach within PSH does not reduce overall substance use, and little is known about HIV risk behavior once this population enters housing. This study identifies clusters of individuals with similar risk behaviors drawn from a sample of residents living in PSH in Los Angeles and attempts to understand the nature of individual risk profiles including the role of contextual and environmental factors.  

Methods: Participants are adults (n=437) transitioning from homelessness to PSH in Los Angeles. Baseline (prior to move in) and 3-month follow-up assessments captured substance use (e.g. binge drinking, injection drug use) and sexual risk behaviors (e.g. exchange sex, multiple partners, unprotected sex) in the past 3 months. Changes in these risk behaviors from baseline to follow-up were used to perform a prototype-based, partitional clustering technique to generate residential profiles that share similar changes in risk behaviors. A subset of these participants (n=30) was then selected to engage in ethnographic shadowing.  

Results: Five clusters emerged from the data. Cluster 1 (64%): low-risk sexual and substance use behavior at baseline that stayed low-risk at follow-up. Cluster 2 (8%): High-risk drug users (predominantly binge drinkers), with low-risk sex behaviors at baseline who decreased their substance use and increased their sexual risk behavior at follow-up. Cluster 3 (7%): low to moderate substance use and high sexual risk behavior (largely exchange sex) at baseline with 88% ceasing exchange sex and decreasing other risk behaviors at follow-up. Cluster 4 (11%): Low-risk substance use and low sexual risk at baseline, with low-risk substance use and increased risky sexual behavior at follow-up. Cluster 5 (10%): High-risk drug users and high-risk sexual behavior at both baseline and follow-up. Analysis of ethnographic field notes found variation in how participants managed risk even within the same cluster with some isolating and others engaging in their local context.

Conclusions: PSH has a varied impact on chronically homeless individuals. While the majority of residents were relatively low-risk at entry and remained so over time, 36% of the sample showed substantial behavior changes, with some increasing risk behavior over time. These findings indicate a need for targeted treatment for substance use and sexual risk behavior and the importance of identifying risk upon entry into PSH.