Cross-trial prediction of treatment outcome in depression: a machine learning approach

To view the full text, please login as a subscribed user or purchase a subscription. Click here to view the full text on ScienceDirect. Figure 1 Analysis pipeline In the STAR*D cohort, label level 1 treatment completers according to whether they reached remission or not (1). Set up ten repeats of ten-fold cross-validation (2). Identify predictors that are most predictive of treatment outcome with a data-driven selection method (elastic net regularisation; 3). Use top 25 predictive features to train a machine-learning algorithm to predict treatment outcomes for citalopram (4). Examine model performance in three treatment groups of an independent clinical trial cohort (COMED; 5). QIDS=quick inventory of depressive symptomatology. Figure 2 Cross-trial prediction of antidepressant treatment outcomes Arrows indicate where a model was trained (arrow origin), and tested (arrow head).…


Link to Full Article: Cross-trial prediction of treatment outcome in depression: a machine learning approach