Identifying Predictors of Therapy Responsiveness from Meta-Emotion Beliefs and Cognitive Flexibility Using ML

Objective: This study aimed to identify transdiagnostic cognitive and affective predictors of therapy responsiveness by applying machine learning algorithms to evaluate meta-emotion beliefs and cognitive flexibility in an outpatient clinical sample.Methods and Materials: A prospective, longitudinal predictive design was employed with a sample of N=۵۱۴ adult participants from Malaysia. Baseline data were collected utilizing the Meta-Emotion Scale and the Cognitive Flexibility Inventory, while treatment outcomes were measured using the Outcome Questionnaire-۴۵.۲. Data analysis was conducted in Python using Scikit-Learn, which involved handling missing values, Z-score standardization, and an ۸۰/۲۰train-test split. Machine learning models underwent hyperparameter optimization via ۱۰-fold cross-validation and were thoroughly evaluated using accuracy, precision, recall, F۱-score, and ROC-AUC metrics, with SHAP (Shapley Additive Explanations) values utilized to determine explicit feature interpretability.Findings: Results indicated that ۶۲.۴%of the participants (N=۵۱۴) demonstrated clinically significant improvement following therapeutic intervention. Therapy responsiveness exhibited significant positive correlations with meta-emotion facets (acceptability: r=〖۰.۴۱〗^(**); controllability: r=〖۰.۵۳〗^(**)) and cognitive flexibility domains (alternatives: r=〖۰.۶۱〗^(**); control: r=〖۰.۴۸〗^(**)). Among the evaluated machine learning classifiers, the XGBoost model achieved the highest predictive performance on the test set (n=۱۰۳), yielding an overall accuracy of ۸۵.۴%and an ROC-AUC of ۰.۹۱. Furthermore, SHAP value analysis explicitly identified the Alternatives facet of Cognitive Flexibility (Mean Abs SHAP: ۱.۲۴) and the Controllability facet of Meta-Emotion (Mean Abs SHAP: ۰.۹۸) as the most highly significant positive predictors of successful therapeutic outcomes.Conclusion: Assessing cognitive flexibility and meta-emotion utilizing advanced algorithmic modeling provides a highly accurate framework for predicting therapy responsiveness, thereby directly facilitating the crucial transition toward proactive, personalized mental health care. Objective: This study aimed to identify transdiagnostic cognitive and affective predictors of therapy responsiveness by applying machine learning algorithms to evaluate meta-emotion beliefs and cognitive flexibility in an outpatient clinical sample. Methods and Materials: A prospective, longitudinal predictive design was employed with a sample of N=۵۱۴ adult participants from Malaysia. Baseline data were collected utilizing the Meta-Emotion Scale and the Cognitive Flexibility Inventory, while treatment outcomes were measured using the Outcome Questionnaire-۴۵.۲. Data analysis was conducted in Python using Scikit-Learn, which involved handling missing values, Z-score standardization, and an ۸۰/۲۰train-test split. Machine learning models underwent hyperparameter optimization via ۱۰-fold cross-validation and were thoroughly evaluated using accuracy, precision, recall, F۱-score, and ROC-AUC metrics, with SHAP (Shapley Additive Explanations) values utilized to determine explicit feature interpretability. Findings: Results indicated that ۶۲.۴%of the participants (N=۵۱۴) demonstrated clinically significant improvement following therapeutic intervention. Therapy responsiveness exhibited significant positive correlations with meta-emotion facets (acceptability: r=〖۰.۴۱〗^(**); controllability: r=〖۰.۵۳〗^(**)) and cognitive flexibility domains (alternatives: r=〖۰.۶۱〗^(**); control: r=〖۰.۴۸〗^(**)). Among the evaluated machine learning classifiers, the XGBoost model achieved the highest predictive performance on the test set (n=۱۰۳), yielding an overall accuracy of ۸۵.۴%and an ROC-AUC of ۰.۹۱. Furthermore, SHAP value analysis explicitly identified the Alternatives facet of Cognitive Flexibility (Mean Abs SHAP: ۱.۲۴) and the Controllability facet of Meta-Emotion (Mean Abs SHAP: ۰.۹۸) as the most highly significant positive predictors of successful therapeutic outcomes. Conclusion: Assessing cognitive flexibility and meta-emotion utilizing advanced algorithmic modeling provides a highly accurate framework for predicting therapy responsiveness, thereby directly facilitating the crucial transition toward proactive, personalized mental health care.

کلیدواژه ها:

Therapy Responsiveness ، Meta-Emotion ، Cognitive Flexibility ، Machine Learning