Predicting music therapy recommendation intentions in insomnia patients: A decision tree approach based on the kab model

Background: Music therapy is a safe, low-cost, non-pharmacological intervention for insomnia, yet patient adoption rates remain low. Traditional statistical methods often fail to capture nonlinear relationships and lack interpretability. Methods: This study analyzed 1,002 survey responses from insomnia patients based on the Knowledge-Attitude-Behavior (KAB) model. Core features, including knowledge score, attitude score, musical background, and prior experience with music for sleep, were constructed to predict binary willingness to recommend music therapy (high vs. moderate/low). A decision tree model, optimized via GridSearchCV, was compared with random forest and logistic regression. Results: The Decision Tree achieved an AUC of 0.535 (95% CI: 0.461 – 0.610), while the Random Forest achieved 0.555 (95% CI: 0.473 – 0.639) and Logistic Regression achieved 0.529 (95% CI: 0.439 – 0.613). Attitude score, knowledge score, and prior experience with music for sleep emerged as the strongest predictors. Visual decision rules and feature importance analysis provided clear clinical interpretability. Conclusions: The decision tree model offers high interpretability for assessing patients’ willingness to adopt music therapy, despite moderate predictive performance. These findings provide an exploratory, data-driven tool with transparent rules that may support preliminary clinical screening and personalized promotion of music therapy. Limitations include moderate AUC values and the need for external validation.