Developing Transdiagnostic Internalizing Disorder Prognostic Indices for Outpatient Cognitive Behavioral Therapy.

Area of Science:

• Psychiatry and Psychology
• Computational Psychiatry
• Machine Learning in Healthcare

Background:

• Heterogeneity in treatment response to Cognitive Behavioral Therapy (CBT) is a significant challenge.
• Predictive models are needed to guide personalized treatment planning for patients with internalizing psychopathology.

Purpose of the Study:

• To develop and validate machine learning-based prognostic models for predicting treatment outcomes in CBT.
• To identify key pretreatment predictors of treatment response for internalizing disorders.

Main Methods:

• Utilized supervised machine learning (super learning) on a dataset of 1,210 outpatients with internalizing psychopathology.
• Employed a comprehensive pretreatment assessment with 324 predictors.
• Evaluated models for principal diagnosis improvement, principal diagnosis remission, and transdiagnostic remission at 12-month follow-up.

Main Results:

• Models for principal diagnosis remission (AUROC=0.71-0.73) and transdiagnostic remission (AUROC=0.71-0.73) demonstrated the best performance.
• Patients in the top predicted probability tertile for principal diagnosis remission were more than twice as likely to achieve it (77.3% vs 35.0%).
• Key predictors included principal diagnosis severity, social anxiety, hopelessness, temperament, and global impairment.

Conclusions:

• Machine learning models show promise in predicting CBT outcomes for internalizing psychopathology.
• While calibration needs improvement, these models can inform treatment planning for more effective and efficient care.
• Further research is warranted to enhance model performance and clinical utility.