Predicting individual change during the course of treatment

Abstract

Objective: An empirically derived prediction model was developed in a private practice setting to monitor on-track and off-track weekly treatment progress in an intensive outpatient program (IOP). Method: The predictive equation was derived as a function of the baseline measure and time. The formulae for the predictive equations were derived from two groups of psychiatric patients (N = 400 each) in an IOP diagnosed with major depression. Each equation was cross-validated between these two psychiatric IOP samples and a dual diagnosis sample (N = 198) using κ, the reliable change index (RCI), receiver operating characteristic curves, and Youden's J. Results: Using varying RCI classifications, approximately 66–75% of both samples reliably improved, 23–24% were indeterminant, and only 1–3% deteriorated. Of patients identified as off-track, which included patients classified as indeterminant and deteriorated, 83% were correctly identified. Of those identified as on-track, 85% were correctly classified. Those identified as on-track (85%) are highly likely to respond to treatment as expected. Conclusions: The overall efficiency index (hit rate) for the correct classification of all patients was 85%. Implications for using this predictive model as a clinical support decision tool with relatively homogeneous populations in other practice settings are discussed.