Background: Most predictor analyses search for single predictors or rely on data from randomized controlled trials. We aimed at detecting a set of clinical baseline variables for prediction of response and remission in 1014 naturalistically treated inpatients with major depressive episode treated for 53.62 ± 47.5 days.
Methods: A three-staged procedure was implemented. First, univariate tests were used for finding associations with baseline variables. Second, logistic regression and third-CART analyses were used to determine predictors of response to inpatient treatment.
Results: Presence of suicidality, a higher initial HAMD-21 total score, an episode length <24 months, fewer previous hospitalizations, and absence of any ICD-10F4 comorbidity predicted response in 2 different statistical models. Remission was predicted by lower HAMD-21 baseline score, episode length <24 months and fewer previous hospitalizations in both models.
Limitation: Results were assessed by a post-hoc analysis, based on prospectively collected data. No controlled study design.
Conclusion: Contrary to current beliefs, baseline suicidality might be associated with higher chances for response. In addition, baseline severity might impact outcome depending on which criterion (remission or response) used.
Copyright © 2011. Published by Elsevier B.V.