Predicting ECT response with machine learning

Abstract

In this issue of JAMA Psychiatry, Redlich et al 1 assess the predictive potential of baseline (before electroconvulsive therapy [ECT]) structural neuroimaging with a set of machine-learning– based approaches. In the pre-ECT evaluation and consent process, the ECT health care professional must balance the anticipated benefits and risks for an individual patient. Ideally, the ECT health care professional would be able to accurately determine a patient’s likelihood of response based on demographics, symptom severity, phenomenology, and treatment history. A number of indices were created as early as 1950 based on clinical variables that included melancholia, family history, symptom severity, age, sex, and the presence of psychosis and personality disorders. Unfortunately, a recent metaanalysis of factors associated with ECT 2 has shown that these clinical factors are imperfect. Only longer duration of symptoms and antidepressant treatment resistance are reliably associated with ECT nonresponse.