Title
Salvaging Forbidden Treasure in Medical Data: Utilizing Single-Record Data to Improve Rare Event Prediction via Integrative Learning.
Abstract
The vast repositories of Electronic Health Records (EHR) and medical claims data hold untapped potential for improving the prediction of rare but critical events, such as suicide attempts. However, conventional predictive modeling setups often exclude the single-record patients -- those with a single documented medical encounter, which include patients who attempted suicide but without any prior historical information. Addressing this gap, we innovate an integrative learning framework to improve rare event prediction by leveraging the forbidden yet valuable information within the single-record data. Our approach employs a multivariate supervised learning component to learn the latent variables that connect target/primary outcomes (e.g., suicide) and surrogate outcomes (e.g., concurrent health conditions) to historical information. It simultaneously employs another unsupervised learning component to utilize the single-record data, allowing latent variables to be shared and self-learned, even with no or insufficient historical information. As such, our approach offers a general strategy for information integration crucial to modeling various rare conditions. We consider neural network-based extensions and discuss some theoretical aspects of the general problem. With hospital inpatient data from Connecticut, we demonstrate that single-record suicide attempts carry valuable information, and utilizing them with our proposed approach can substantially improve suicide risk prediction.
Bio
Kun Chen is a Professor in the Department of Statistics at the University of Connecticut (UConn) and a Research Fellow at the Center for Population Health, UConn Health Center. He is a Fellow of the American Statistical Association (ASA) and an Elected Member of the International Statistical Institute (ISI). His research focuses on large-scale multivariate statistical learning, statistical machine learning, and healthcare analytics. Dr. Chen received his B.Econ. in Finance and Dual B.S. in Computer Science from the University of Science & Technology of China in 2003 and his Ph.D. in Statistics from the University of Iowa in 2011.
