Predicting treatment success and failure using routine outcome data: The role of therapist effects in dynamic predictive modelling

Background: The goal of this research was to identify how early in psychotherapy a reliable prediction regarding the probability of premature termination, treatment failure or treatment success can be made, and which variables are most predictive. This research breaks new ground by including data on each clinician's historical outcomes (therapist effects) to explore to what degree predictions and decision rules might differ with therapist effectiveness. Materials & Methods: Data was drawn from a large online database of clients regularly completing self-report outcome questionnaires. The sample included clinicians (n = 1020) treating clients (n = 68,690) with an intake score in the clinical range during a treatment episode of between 2 and 10 sessions. Variables available at the first session and later sessions were tested for predictive validity for status at subsequent sessions using general linear regression models and logistic regression models. Results: The results suggest that therapy outcomes can be predicted as early as the second session with a good degree of accuracy, 65%. The analyses revealed that predictions made as early as the second session remain valid even for clients taking up to 10 sessions to complete treatment. Adding therapist effects to the predictive model provides a much more nuanced understanding of the results. By session two, therapists in the top quartile demonstrate effect sizes comparable to the effect sizes of bottom quartile therapists at session ten. Conclusion: The probability of client success, failure or early termination varies significantly depending on the clinician. Implications for training and supervision are discussed.