Biostatistics Monthly Seminar.

Dr. Kun Chen Professor of Statistics at the University of Connecticut and a Research Fellow at UConn Health’s Center for Population Health. presents Title: Advancing Rare-Event Modeling for Suicide Risk with Electronic Health Records Abstract: Electronic health records (EHRs) and claims data hold untapped potential for studying rare but critical events such as suicide, yet progress is hampered by many challenges, including uncertain event identification, extreme sparsity in diagnosis-derived features, and the routine exclusion of patients with few encounters. We discuss practical solutions to these problems, with emphasis on outcome enrichment and salvaging “single-record” patients. Conventional pipelines discard individuals with only one documented visit, but, in our data, those diagnosed with a suicide attempt at their sole encounter comprise a striking 70–80% of all attempt cases. We introduce a hybrid, integrative learning framework; it employs a supervised component to connect concurrent outcomes to historical information and simultaneously employs an unsupervised component to utilize the single-record data through sharing the latent subspace. In a suicide risk study of adolescents and young adults, we demonstrate that our framework substantially improves model performance. Bio: Kun Chen is Professor of Statistics at the University of Connecticut and a Research Fellow at UConn Health’s Center for Population Health. He is a Fellow of the American Statistical Association and an Elected Member of the International Statistical Institute. His research spans large-scale statistical learning, machine learning, and health data science, with recent emphasis on data-driven suicide prevention using electronic health records and health information exchanges. He has authored over 100 peer-reviewed papers, graduated 18 PhD students, and his work has been supported by the Simons Foundation, NSF, NIH, and USDA. Dr. Chen helped establish the New England Statistical Society in 2017 and serves on steering committees for the New England Rare Disease Statistics Workshop and the Pharma Data Science Conference. He is a Board Member of the International Chinese Statistical Association and an Associate Editor for several journals including the Annals of Applied Statistics. Meeting ID: 961 3147 3264, Passcode: 334135