Early Improvement Predicts Treatment Response in Depression: An Ecological Momentary Assessment Study in an Inpatient and Day Clinic Setting

Predicting treatment response through early improvement can reduce patients' time in ineffective treatments before considering alternatives. However, for psychological interventions, there is no consensus on what time window and improvement rate early in the treatment is the most informative for distinguishing treatment responders from nonresponders. This study investigated these aspects in an inpatient and day clinic setting among severe depressed patients who perceived intensive psychological treatment and compared Weekly Questionnaire Assessments (WQA) and Ecological Momentary Assessment (EMA) regarding their power to predict treatment response through early improvement. Fifty-two depressed patients were randomly assigned to one of three intensive 7-week psychological interventions (two individual and two group sessions per week) applied in an inpatient or day clinic setting. Early improvement was assessed three times daily using EMA and weekly using questionnaires (BDI-II). Linear Regression Models and Receiver Operating Characteristic Analyses were conducted to predict treatment response (BDI-II improvement from pre- to postintervention ≥50%) in patients who received a full course of treatment. Moreover, ratios of true negative/false negative predictions were calculated to explore the predictive value of different early improvement definitions: 10%, 20%, 30%, or 40% improvement after 1, 2, 3, or 4 treatment weeks. Both EMA and WQA significantly predicted treatment response after 3 weeks with AUC values of 73% (EMA) and 77% (WQA). A WQA-assessed 10% improvement after 4 weeks yielded the highest ratio of true negative/false negative predictions, with a true negative rate of 22% and a false negative rate of 0%. 10% improvement in depressive symptoms assessed with WQA after 3 to 4 weeks of treatment was the best predictor in our study. Further research is needed to validate the results. This trial design is registered with osf.io/9fuhn.