Factors That Predict Leaving Treatment Against Professional Advice Among Homeless Heroin Users

Background/Purpose: Heroin use has increased across the country, and heroin related deaths have more than tripled since 2010. Individuals who experience homelessness or have experienced homelessness have a greater chance of being injection drug users, and have a greater chance of death and exposure to infectious diseases such as HIV/AIDS and Hepatitis C. The social context in which homeless individuals use heroin by injection may pose significant risks to themselves and community residents.

In response to the heroin crisis among homeless individuals and to decrease the potential exposure to used syringes, greater attention ought to be paid to the factors that contribute to successful treatment outcomes of homeless individuals. This secondary data analysis examines the factors that predict treatment outcomes among heroin injection users.   

Methods: Data and samples: We use the Treatment Episode Discharge Data Set (2006-2011), a national census data of discharges from substance abuse treatment facilities. Inclusion criteria were those who were homeless at the time of discharge from a rehab center offering 30 days or more of treatment, primary substance was heroin, and primary route of administration was injection (n=1405). Predictor variables were: gender, race, length of stay, education, age, health insurance, days waiting to enter treatment, number of arrests in 30 days prior to treatment admission, principal source of the referral, number of prior treatment episodes, and psychiatric problem along with alcohol/drug abuse. The outcome variable was selected from the discharge categories indicating completion of treatment or left against medical advice (did not complete treatment). These were recoded into two categories.

Results: Logistic regression was conducted on all 11 predictors to determine whether those discharged as homeless completed treatment or left against professional advice. When all predictors are considered together, they significantly predicted whether a homeless individual who reported injection heroin use would complete treatment or leave against professional advice. The likelihood ratio test of the model was statistically significant (χ² = 540.47, df  = 11, p < .001).  Gender, length of stay, number of prior treatment episodes, and age separately significantly predicted leaving against professional advice. The odds of leaving against professional advice become increasingly greater when the number of prior treatment episodes increases. In addition, an increase in the length of stay in treatment decreases the odds of leaving against professional advice. Among older homeless individuals, the odds of leaving against professional advice decreases.

Conclusions and Implications: Long-term re-hab facilities may need to differentiate treatment for homeless youth to increase their odds of completing treatment. Future research is needed that focuses on strategies at the onset of treatment for those at greater risk of not completing treatment, to determine which methods increase their likelihood of completing treatment.