Feasibility of artificial intelligence-based measurement in psychotherapy practice: Patients' and clinicians' perspectives

Abstract

Background

Tracking session-by-session patient-reported outcomes (e.g. alliance and clinical symptoms) has been shown to improve treatment outcomes. However, self-report measures are cumbersome to collect, and completion rates are inconsistent. Proof-of-concept machine learning research applications using psychotherapy data sets suggest that it may be possible to generate fully automated predictions of patient-reported alliance and symptom ratings based on behavioural markers extracted from video recordings of psychotherapy sessions. For these artificial intelligence (AI)-based measurements to be feasible, patients and clinicians must be comfortable with video recording their sessions and must be open to deploying such automated AI-based models in their psychotherapy practice.

Methods

We conducted two online survey studies between December 2022 and March 2023. We asked 954 patients and 248 clinicians about the use and usefulness of (1) self-report measures for routine outcome monitoring, (2) video recording therapy sessions and (3) utilising AI-based prediction models in their treatments.

Results

Patients and clinicians found the use of self-report measures useful but burdensome. While both patients and clinicians reported interest and willingness to embrace AI-based technology for measurement-based care, patients reported significantly more willingness to record their sessions, and more positive views on the use and usefulness of AI-based measurement feedback for clinical outcomes, compared with clinicians.

Conclusion

Clinicians should be provided with more practice and training in the use of AI-based tools to aid their clinical work before such AI-based measurement tools may be successfully implemented into clinical practice.