Early identification of psychotherapeutic change: AI-derived predictors from routine data

Area of Science:

• Psychology
• Artificial Intelligence
• Data Science

Background:

• Measurement-based care (MBC) is effective for enhancing therapy outcomes.
• Advanced statistical methods can optimize feedback within MBC programs.
• AI algorithms can improve the predictive performance of MBC.

Purpose of the Study:

• To identify key predictors of therapy outcomes using AI.
• To analyze data from a large, diverse client sample in university counseling centers.
• To test the efficacy of AI-driven approaches in optimizing MBC.

Main Methods:

• Utilized ant colony optimization to select a content-valid item subset from the Behavioral Health Measure.
• Employed logistic Lasso regression to identify influential item-level predictors of treatment success.
• Analyzed data from 9,591 diverse clients, focusing on early treatment sessions (3-5).

Main Results:

• Five key items emerged as strong predictors of early treatment outcomes: self-esteem, general anxiety, substance use, cognitive attention, and work/school life purpose.
• AI-derived methods successfully identified a subset of items predictive of clinically significant change.
• These predictors were identified within the first three to five therapy sessions.

Conclusions:

• MBC platforms can be enhanced by focusing on specific data patterns for progress monitoring.
• Early identification of key indicators can significantly aid therapists and clients in tracking treatment progress.
• AI-driven analysis offers a powerful tool for refining MBC and improving therapeutic outcomes.