Associations of Common Depression Treatment Metrics with Patient-centered Outcomes

Background: Various metrics predicated on Patient Health Questionnaire-9 (PHQ-9) scores denote depression "response" or "remission" over time, but few have been empirically validated. We compare the associations of depression response and remission metrics with concomitant clinical improvement in patient-centered outcomes (PCOs). Subjects: Secondary analysis of PHQ-9 and PCO data from the treatment arm (n=906) of the Improving Mood - Promoting Access to Collaborative Treatment (IMPACT) trial. Research Design: We conducted univariate correlations between 9 depression treatment metrics and 4 PCOs. For each PCO, we specified a multivariate linear fixed-effects regression model with penalized LASSO (least absolute shrinkage and selection operator) variable selection that included parameters for each incremental absolute PHQ-9 decrease between 0 and 16 points. Model predictive properties were assessed using a split sample analysis. Results: There was a notable variation in depression improvement rates across metrics. Each metric was significantly associated with PCOs in univariate analyses. In the multivariate models, the cumulative likelihood of PCO improvement was most improved by absolute PHQ-9 score decreases of 7-9 and 14-16 points. The multivariate models showed greater area under the curve (0.671-0.804) in out-of-sample predictions of PCO changes than the univariate models (0.529-0.649). Conclusions: Choice of depression response metric impacts observed response and remission rates, though PCOs tend to improve with depression improvement regardless of metric choice. Absolute incremental PHQ-9 score decreases are broadly associated with an increased likelihood of favorable PCO scores. Our findings support a novel PHQ-9 metric defined by an absolute score change of 8 points or greater.