Relapse outcomes at 6-, 12-, and 18-month intervals were compared between clients randomly assigned to day (n=114) versus residential (n=147) drug abuse treatment. Day clients were more likely than residential clients to relapse 6 months post-admission (OR=3.06, p<0.001); however, no setting differences at 12 or 18 months were found. Few baseline predictors were prospectively related to relapse at 12 and 18 months. These predictors were usual employment status (part-time OR=17.47, p<0.001; full-time OR=2.54, p<0.001), history of drug injecting (OR=5.39, p<0.01), multiple sex partners (OR=1.16, p<0.01), and not having a gay sexual partner (OR=0.05, p<0.03) during 6 months prior to admission. Still, these baseline predictors, together with the existing literature, could be used by drug treatment professionals to identify individuals who may be at high risk for relapse over time, and to offer specialized treatment and aftercare resources as intervention and prevention measures.