Selecting the optimal treatment for a depressed individual: Clinical judgment or statistical prediction?

Suzanne C. van Bronswijk*, Lotte H.J.M. Lemmens, Marcus J.H. Huibers, Frenk P.M.L. Peeters

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

BACKGROUND: Optimizing treatment selection is a way to enhance treatment success in major depressive disorder (MDD). In clinical practice, treatment selection heavily depends on clinical judgment. However, research has consistently shown that statistical prediction is as accurate - or more accurate - than predictions based on clinical judgment. In the context of new technological developments, the current aim was to compare the accuracy of clinical judgment versus statistical predictions in selecting cognitive therapy (CT) or interpersonal psychotherapy (IPT) for MDD.

METHODS: Data came from a randomized trial comparing CT (n=76) with IPT (n=75) for MDD. Prior to randomization, therapists' recommendations were formulated during multidisciplinary staff meetings. Statistical predictions were based on Personalized Advantage Index models. Primary outcomes were post-treatment and 17-month follow-up depression severity. Secondary outcome was treatment dropout.

RESULTS: Individuals receiving treatment according to their statistical prediction were less depressed at post-treatment and follow-up compared to those receiving their predicted non-indicated treatment. This difference was not found for recommended versus non-recommended treatments based on clinical judgment. Moreover, for individuals with an IPT recommendation by therapists, higher post-treatment and follow-up depression severity was found for those that actually received IPT compared to those that received CT. Recommendations based on statistical prediction and clinical judgment were not associated with differences in treatment dropout.

LIMITATIONS: Information on the clinical reasoning behind therapist recommendations was not collected, and statistical predictions were not externally validated.

CONCLUSIONS: Statistical prediction outperforms clinical judgment in treatment selection for MDD and has the potential to personalize treatment strategies.

Original languageEnglish
Number of pages9
JournalJournal of Affective Disorders
Volume279
DOIs
Publication statusE-pub ahead of print - 7 Oct 2020

Cite this