Predicting Change in an Integrated Dual Diagnosis Substance Abuse Intensive Outpatient Program

Research on routine outcome monitoring in psychotherapy settings is plentiful but not without implementation obstacles. In fact, there is a relative dearth of real-time outcome monitoring in substance use treatment settings. Numerous barriers to the development and implementation of clinical decision support tools and outcome monitoring of substance use patients, including the need to establish expected trajectories of change and use of reliable change indices have been identified (Goodman, McKay, & DePhilippis, 2013 ). The current study was undertaken to develop expected trajectories of change and to demonstrate the treatment effectiveness of a dual diagnosis intensive outpatient program. The expected trajectories of change for days of substance use and depression scores were developed using predictive equation models from derivation samples and then applied to cross-validation samples. Predictive equations to monitor substance use were developed and validated for all patients and for only patients who were actively using substance at the time of admission, as well as to monitor severity of their depression symptom on a weekly basis. Validation of the equations was assessed through the use of Cohen's kappa (κ), receiver operating characteristic curves, reliable change index, and percentage improvement. Large effect sizes for reductions in substance use (Cohen's d = .76) and depressive symptoms (d = 1.10) are reported. The best predictive models we developed had absolute accuracy rates ranging from 95 to 100%. The findings from this study indicate that predictive equations for depressive symptoms and days of substance use can be derived and validated on dual diagnosis samples.