Personalized feedback – Client characteristics as moderators of the effect of type of feedback on treatment outcome: A machine learning approach.

Shachaf Tal & Prof. Sigal Zilcha-Mano

Studies suggest that methods of monitoring and providing feedback to clinicians about client mental health status over the course of psychotherapy may improve treatment outcomes, at least for some clients (Lambert, 2015). Several methods for providing feedback about client mental health status have been developed, with promising results reported in the literature about their utility. High dimensional relationships are hypothesized. Thus, machine learning and other data-driven techniques will be used to predict outcome for a given patient, based on pre-treatment characteristics. Predicted outcome will be used to identify an ideal feedback condition for each patient.
This is a secondary analysis of a study in which 547 clients were randomized to one of five feedback conditions: (a) a control group in which therapists did not receive any feedback; (b) therapists received raw weekly feedback on clients’ psychological dysfunction by being given access to the outcome questionnaire (OQ) answered by clients; (c) therapists received weekly raw feedback about clients’ alliance perception by being given access to the Working Alliance Inventory (WAI) answered by clients; (d) therapists received raw weekly feedback about clients’ OQ and WAI; and (e) therapists received weekly feedback containing Lambert’s OQ progress feedback report. Elastic net regularization was used in variable selection, and gradient boosting algorithm was used to predict outcome.
Although the findings are observational and will need further validation before they can be used to guide feedback selection, specific client characteristics may guide clinical decision regarding the type of feedback that may result in the most favourable results for specific sub-groups of clients.

Citation:
Coming Soon