MATCH: client-therapist matching with machine learning.
Journal:
Psychotherapy research : journal of the Society for Psychotherapy Research
Published Date:
Dec 15, 2025
Abstract
OBJECTIVE: Therapist assignment can have a profound impact on the success of counseling. Accordingly, we present a framework that combines archival data and computer science algorithms to match clients to therapists so that client outcomes are, on average, better than assignment to the next available therapist while respecting each therapist's availability. METHODS: We estimated the association between therapists and clients' outcomes using high-dimensional data from CCAPS-62 and OQ-45 surveys from 2014 to 2019, obtained at a clinic of a large private university (exactly 10,100 courses of therapy). These predictions informed constrained client-therapist matches. We evaluated the method using simulations and statistical tests. RESULTS: Our model's predictions identified therapist-client pairs that are associated with an average improvement of client outcomes-0.145-3.65 OQ-45 points higher (d = 0.11) than usual, with 95% confidence, evaluated using out-of-sample data. Accounting for constraints on the availability of therapists (largely unconsidered in prior literature), our method could aid in improving outcomes with minimal impact on client wait times and few administrative changes necessary. CONCLUSIONS: Improving client outcomes given therapist constraints through data-driven client-therapist matching might still retain a sizable portion of the benefit of data-driven matching.
Authors
Keywords
No keywords available for this article.
