Predicting Drug Court Outcomes Using Assessments of Substance Use Severity and Criminogenic Risk

• In Event: Advances in Risk Assessment and Prediction

Abstract

Evidence-based drug court practices require psychometrically validated instruments. This study employs a modified multitrait-multimethod matrix to examine the validity of criminogenic risk (CR) and drug involvement (DI) measures based on official records and intake interviews. Participants were 240 New York City drug court participants (81% male). Interview measures of DI were: Level of Service Inventory-Revised (LSI-R) substance use, the Texas Christian University Drug Screen, and the Uniform Court Reporting system. Official records of prior drug arrests and convictions indexed DI. LSI-R general recidivism risk score, and prior non-drug arrests and convictions, indexed CR. Outcomes were percent positive UA, program failure, and 2-year re-arrest. Good convergent validity was found among CR measures, but records and interview measures of DI were not highly correlated with each other (r = .13). CR measures were better than DI measures at predicting re-arrest (mean r = .25 vs. .11). Drug priors, but not DI measures, predicted positive UA (mean r = .29 vs. r =.06). LSI-R-general risk was the best predictor of drug court failure (r = .26). In sum, this study provides evidence of adequate criterion and predictive validity for CR measures, but mixed convergent and inadequate discriminant and predictive validity for DI measures.

Author

• Warren Reich, Center for Court Innovation