Using machine learning algorithms to predict the effects of change processes in psychotherapy: Toward process-level...

Area of Science:

• Psychology
• Psychotherapy Research
• Machine Learning in Healthcare

Background:

• Personalized medicine is increasingly important in psychotherapy.
• Identifying which therapeutic change processes are most relevant for individual patients can optimize treatment.
• Mastery and clarification are two key psychotherapeutic change processes.

Purpose of the Study:

• To develop and validate machine learning algorithms for predicting the individual relevance of mastery and clarification processes in psychotherapy.
• To assess the feasibility of using these algorithms to guide treatment recommendations.

Main Methods:

• A naturalistic outpatient sample (n=608) receiving integrative treatment was studied over the first 10 sessions.
• Within-patient effects of therapist-perceived mastery and clarification on subsequent patient-evaluated outcomes were estimated.
• Machine learning algorithms were trained on a subsample (n=407) to predict process relevance using baseline characteristics and tested on a holdout subsample (n=201).

Main Results:

• Significant within-patient effects of mastery and clarification on outcomes were observed.
• Predictive algorithms showed small-to-medium correlations for mastery (r=.18) and clarification (r=.16) in the holdout sample.
• Algorithms identified patients for whom mastery (14%) or clarification (18%) were indicated, with mastery focus correlating with better outcomes in the indicated group (r=.33).

Conclusions:

• Individual predictions of psychotherapeutic process relevance are feasible using machine learning.
• These findings support the potential utility of algorithms for therapist feedback and personalized treatment recommendations.
• Replication with prospective experimental designs is needed to confirm these results.