The study examined the ability of several baseline variables to predict retention and treatment outcome in a cohort study that included 137 participants with heroin and cocaine addiction attending a public addiction service in Italy. Predictors included past drug use, intravenous use, initial urine drug screen results, cocaine and heroin craving and selfreported psychiatric symptoms. Severe depression and, in a lesser extent, anxiety symptoms, predicted lower retention. Only intravenous use was a significant predictor of cocaine use in the whole sample. Positive urine for cocaine and intravenous use predicted worse outcome among heroin addicts. Three Symptoms Checklist-90 (SCL-90) scales scores were associated with lower cocaine and heroin use. Multiple logistic regression showed that cocaine positivity and more structured psychosocial treatment predicted successful outcome. Results highlight the prognostic value of carefully assessing patterns of cocaine use and co-use, routes of administration and psychiatric symptoms at the beginning of treatment, especially in heroin addicts.For clinicians that aim to identify predictors of treatments outcomes, usefulness of knowledge from careful academic research depends on two factors: how much acquired knowledge can be exported to the context in which they work and how simple and cost effective are procedures and tools with which were produced. Regarding the first factor, unlike in research settings, public addiction services treat heterogeneous clinical samples. Mostly, patients use more than Int J Ment Health Addiction