Gregory E. Simon, M.D., M.P.H. (Kaiser Permanente Washington Health Research Institute, Seattle) join Dr. Dixon and Dr. Berezin to discuss the use of machine learning models to analyze electronic health records to predict antidepressant treatment response. 00:00     Introduction 02:31    Focus on practical research 04:55    Population studied 05:57    Predicting outcomes 07:20    Using diagnostic codes, not personalized notes 08:04    What three data items might be more helpful? 08:49    What key indicators are we missing in clinical care? 11:35    A billing tool, not a clinical tool 12:57    Is suicide a predictable event based on electronic health record data? 14:48    “Machine learning and artificial intelligence”  16:15    Methods 18:59     Can we do a better job clarifying what we mean by depression? 22:32    How can we use a predictive model in clinical practice? 28:20    Predictive models, probability, the weather, and communicating  Transcript Subscribe to the podcast here. Check out Editor's Choice, a set of curated collections from the rich resource of articles published in the journal. Sign up to receive notification of new Editor's Choice collections. Browse other articles on our website. Be sure to let your colleagues know about the podcast, and please rate and review it wherever you listen to it. Listen to other podcasts produced by the American Psychiatric Association. Follow the journal on Twitter. E-mail us at [email protected]