Treatment attrition is a major problem for programs treating adolescent substance abusers. To isolate and cross validate factors which are predictive of addiction treatment attrition among adolescent substance abusers, screening interview and diagnostic variables from 119 adolescent in-patients were submitted to a discriminant equation analysis. The results showed that the equation was able to successfully predict program completion/non-completion for 82 percent of the derivation sample. The variables which distinguished the two groups were a diagnosis of affective disorder, two or more previous psychiatric treatments, a history of family pathology, and minimal (or no) family involvement in the patient's treatment. Use of this formula on a second cross-validation sample (N=45) yielded a hit rate of 76 percent. These findings support the use of the discriminant equation analysis model, and suggest that treatment dropouts can be identified even among high risk groups such as adolescents in substance abuse treatment. (BL)