Abstract
This article examines the explanatory power of programmatic and individual-level variables, using the first nationally representative sample of both treatment facilities and clients. Multivariate statistical models accounted for almost half the variance in post- treatment heroin, marijuana, and any illicit drug use and about one- third the variance in crime. Use of any illicit drug during treatment, even after controlling for pretreatment use. was predictive of every type of drug use after treatment and three of the four major crimes analyzed. In addition, longer lengths of stay decreased the likelihood of alcohol/illicit drug use and criminal behavior after treatment. However, treatment variables added little explanatory power beyond the sociodemographic and pretreatment variables.