Predictors of Adolescents' First Episode of Homelessness Following Substance Use Treatment

Abstract

Purpose: A growing body of research has identified correlates (i.e., predictors) of youth homelessness. However, such risk and protective factors have not been identified for youth receiving substance use treatment. Using characteristics collected at treatment intake, the present study sought to identify predictors of youths' first episode of homelessness during the 12 months after substance use treatment entry.

Methods: Data come from a longitudinal study of adolescents (n = 17,911; aged 12-17 years) receiving substance use treatment throughout the U.S. Participants completed surveys at intake and at 3, 6, and 12 months later. Logistic regression and Lasso machine learning regression were used to predict participants' first episode of homelessness in the 12 months after treatment intake.

Results: After excluding adolescents reporting previous experiences of homelessness, 5.0% of adolescents reported their first episode of homelessness over the 12 months after treatment intake. The results from logistic and lasso models were generally consistent. Final models revealed that adolescents who were older, male, reported more victimization experiences, mental health problems, family problems, deviant peer relationships, and substance use problems (more treatment episodes and illicit drug dependence) were more likely to report experiencing homelessness. Hispanic/Latino adolescents were less likely to experience homelessness, compared with white adolescents.

Conclusions: The results point to the important risk and protective factors that can be assessed at treatment entry to identify adolescents at greater risk of experiencing their first episode of homelessness.

Keywords: Homelessness; Machine learning; Substance use; Victimization.