New Methods in Predictive Analytics Using Big Data to Develop Personalized Treatment Models for Anxiety and Mood Disorders Ronald C. Kessler, Ph.D. McNeil Family Professor of Health Care Policy Harvard Medical School Presentation at Academy Health in the session “Personalized Medicine Ain’t Just Genomics: Using Predictive Analytics to Deliver Patient- Centered Care and Population Health” June 28, 2016 1
Previous Research on Predictors of Differential Anxiety-Depression Treatment Response • Many significant predictors • Few of these predictors are available in EMRs • Existing research is based on small clinical trial samples • Comparisons typically involve active comparison of 2 treatment options (e.g., psychotherapy vs. medication) 2
Challenges in Model-Building • Feature selection • Web-based self-report • EMR data • The requirement for a large sample • Double-robust observational data vs. controlled trials • Non-random selection into type of treatment • Non-random loss to follow-up • Heterogeneity of treatment response • Development of a distinct model for each type of treatment • Cross-validated individual-level outcome comparison across treatments • Estimation • Ensemble machine learning 3
Challenges in Implementation • Demonstrating effectiveness • Pragmatic trials • Level of entry • Imposing on busy clinical schedules • Best practices quality assurance • Clinical decision support • Primary care & collaborative care support • Specialty care support 4
Contact Ronald C. Kessler, Ph.D. kessler@hcp.med.harvard.edu Background Material Kessler, R.C., van Loo, H.M., Wardenaar, K.J., Bossarte, R.M., Brenner, L.A., Ebert, D.D., de Jonge, P., Nierenberg, A.A., Rosellini, A.J., Sampson, N.A., Schoevers, R.A., Wilcox, M.A., Zaslavsky, A.M. (2016). Using patient self-reports to study heterogeneity of treatment effects in major depressive disorder. Epidemiology and Psychiatric Sciences , [Epub ahead of print], 1-15. PMID: 26810628 5
Recommend
More recommend
