To facilitate targeted intervention in substance abuse treatment programs, a scoring methodology is developed to identify clients at risk of premature program exit. Designed to be simple enough for a clinician to easily apply in practice, the risk score is derived from self-reported and observable client characteristics collected at program intake. Our motivating problem is improving a residential substance abuse treatment program for military veterans, and we demonstrate the risk score applied to data from 680 veterans who exited from a San Diego-based rehabilitation program from 2009 to 2011. For this program, the existence of a mental health condition, chronic physical health condition, and the client’s residence prior to program admission were predictive of successfully completing 150 days of treatment. Length of stay and residence prior to program admission were predictive of successful program completion. The risk score methodology is generalizable and can be customized for any treatment program.