Cross-trial prediction of depression remission using problem-solving therapy: A machine learning approach

Area of Science:

• Computational psychiatry
• Clinical informatics
• Machine learning in healthcare

Background:

• Psychotherapy is a standard depression treatment, but prognosis prediction relies on subjective clinical judgment.
• Provider variability and trial-and-error approaches can impact treatment effectiveness.

Purpose of the Study:

• To develop machine learning (ML) algorithms for predicting depression remission.
• To identify patients likely to respond to 6-month problem-solving therapy (PST).

Main Methods:

• ML models were trained and validated on data from 2 randomized trials (ENGAGE-2 and RAINBOW).
• Predictor variables included baseline characteristics and intervention engagement.
• Depression remission was defined as a Depression Symptom Checklist (SCL-20) score < 0.5 at 6 months.

Main Results:

• The best ML model accurately predicted depression remission above chance levels in both internal and external validation cohorts.
• Key predictors identified included female sex, lower sleep disturbance, and reduced negative problem orientation.
• Models achieved 72.6% accuracy at baseline and 72.3% at 2 months in internal validation.

Conclusions:

• ML models utilizing clinical and patient-reported data can effectively predict depression remission in PST.
• This enables prospective identification of treatment responders.
• Personalized early treatment optimization strategies can be developed based on these predictions.