The construct and predictive validity of different approaches to combining urine and self-reported drug use measures among older adolescents after substance abuse treatment.

Reconciling urine results and self-reports is a classic challenge in substance abuse treatment research in general. For adolescents, the problems are compounded by the facts that they are more likely to use marijuana (which takes longer to metabolize) and to be coerced into treatment (which may increase lying). This article examines the construct and predictive validity of several different approaches for combining urine and self reported drug use including using common individual measures (urine tests and self-reported recency, frequency, and peak use), taking either as positive, using a summary scale, and using a latent model. Data are from 819 older adolescents 24 to 42 months after intake in seven sites. Days of use, the GAIN's substance frequency scale, and a latent model were the three best methods in terms of construct and predictive validity. Implications for treatment and longitudinal evaluation will be discussed.