Predicting Dropout From Cognitive Behavioral Therapy for Panic Disorder Using Machine Learning Algorithms

Abstract

Background: Attrition is an important problem in clinical practice and research. However, the predictors of dropping out from cognitive behavioral therapy (CBT) for panic disorder (PD) are not fully understood. In this study, we aimed to build a dropout prediction model for CBT for PD using machine learning (ML) algorithms.

Methods: We treated 208 patients with PD applying group CBT. From baseline data, the prediction analysis was carried out using two ML algorithms, random forest and light gradient boosting machine. The baseline data included five personality dimensions in NEO Five Factor Index, depression subscale of Symptom Checklist-90 Revised, age, sex, and Panic Disorder Severity Scale.

Results: Random forest identified dropout during CBT for PD showing that the accuracy of prediction was 88%. Light gradient boosting machine showed that the accuracy was 85%.

Conclusions: The ML algorithms could detect dropout after CBT for PD with relatively high accuracy. For the purpose of clinical decision-making, we could use this ML method. This study was conducted as a naturalistic study in a routine clinical setting. Therefore, our results in ML approach could be generalized to regular clinical settings.

Keywords: Cognitive behavioral therapy; Machine learning; Panic disorder; Predictor.

Figure 1

Process flow diagram for predictive models. LightGBM: light gradient boosting machine; SMOTE: synthetic minority oversampling technique.

Figure 2

The predictive performance of machine learning models. LightGBM: light gradient boosting machine.

Figure 3

The feature importance of machine learning models. LightGBM: light gradient boosting machine; NEO-FFI: NEO Five Factor Index; PD: panic disorder.