Optimizing Outcomes in Psychotherapy for Anxiety Disorders Using Smartphone-Based and Passive Sensing Features:...

Area of Science:

• Psychiatry and Mental Health
• Digital Health
• Machine Learning in Healthcare

Background:

• Cognitive behavioral therapy (CBT) is effective for anxiety disorders, but only half of patients benefit.
• Predictive algorithms using digital and passive sensing data could improve patient selection for CBT.

Purpose of the Study:

• To identify predictive features for transdiagnostic CBT response in anxiety disorders.
• To investigate mechanisms underlying treatment response to CBT.

Main Methods:

• A randomized controlled trial comparing CBT to a waitlist (3:1 ratio, N=150).
• Multimodal data collection including self-reports, experimental tasks, biological samples, ecological momentary assessments, and passive sensing.
• Machine learning models (e.g., XGBoost, PRN, KNN) used for predictive modeling.

Main Results:

• The trial is completed, and results are pending dissemination.
• Data analysis involved feature selection and machine learning for response prediction.
• Mechanistic hypotheses regarding self-efficacy and daily symptoms were investigated.

Conclusions:

• The study aims to enhance CBT by accurately predicting treatment response.
• Understanding and augmenting treatment mechanisms will enable personalized anxiety disorder treatment.