Improved dropout prediction in group cognitive behavior therapy (CBT) using classification trees.

Abstract

Objective: Dropout is a major factor undermining the effectiveness of psychotherapy, however, it remains poorly anticipated in clinical practice. Classification trees may offer simple, accessible, and practical solutions to identifying patients at-risk of dropout by synthesizing potentially complex patterns of relationships among intake measures.

Method: Intake variables were collected from day-patients who attended a Cognitive Behavior Therapy (CBT) group program at a private psychiatric hospital between 2015 and 2019. Based on these variables, two classification trees were trained and tested to predict dropout in (1) a weekly group, and (2) an intensive daily program.

Results: Dropout was lower in the intensive treatment (Weekly CBT = 21.9%, Daily CBT = 13.2%), however, in both programs, the number of comorbid diagnoses was the most important factor predicting dropout. Overall balanced accuracy was comparable for both tree models, with the Weekly CBT model identifying 63.18% of dropouts successfully, and the Daily CBT model identifying dropouts with 62.06% accuracy.

Conclusion: Findings suggest that comorbidity may be the most important factor to consider when assessing dropout risk in CBT, and that dropout can be predicted with moderate accuracy early in therapy via simple models. Furthermore, findings suggest that condensed, intensive treatments may bolster patient retention.