Thomas McCoy on Computing Transdiagnostic Psychiatric Phenotypes From the Health Record (TIPS 2018)

Human communication is at the core of all clinical interactions, especially in psychiatry and psychology, since changes in social function and use of language are often among the first noticeable clues of an emerging psychiatric illness or impending episode. Systems that capture and analyze naturally occurring speech or written language could therefore have transformative potential to aid in low-burden mental health surveillance strategies to support individuals most at risk with both prediction and optimal prevention strategies. This session brings together experts in both computational aspects of natural language processing (NLP), and their deployment in a range of psychiatric illnesses and treatment contexts, including mining electronic medical records for risk stratification, analyzing text-based encounters with a crisis coach to optimize online therapeutic encounters, and predicting individual-level prognosis from open and directed samples of speech and writing.

These remarks were part of the 2018 Technology in Psychiatry Summit, an event sponsored by the McLean Institute for Technology in Psychiatry, which occurred November 1-2, 2018 at Harvard Medical School, Boston, Massachusetts. Part of panel discussion, Harnessing Natural Language for Prediction and Prevention.

Themes/keywords: prediction and prevention using technology; natural language processing; linguistic predictors of mental health; mining electronic health records; translation of tools or methodologies between languages or cultural contexts.

Thomas McCoy, MD, is the director of research at the MGH Center for Quantitative Health and an assistant professor of Medicine and Psychiatry at Harvard Medical School. He attended Dartmouth College and Cornell Medical School before completing the MGH/McLean psychiatry residency training program, an informatics fellowship in the Center for Experimental Drugs and Diagnostics, and serving as chief resident. His research focuses on development of computed phenotypes in, and applicable to, secondary use of healthcare data generated as part of routing care. He has applied computed phenotypes to both prediction and risk stratification, and to genomics research.

Please visit to learn more about the McLean Institute for Technology in Psychiatry.