Predictive Modeling and its Applications in Healthcare
Now more than ever, electronic health records (EHRs) are generated in large quantities and diverse content. This explosion of information has naturally enabled powerful data analyses to potentially improve healthcare. This talk will review my ongoing research projects and production systems focusing on predictive modeling from structured and unstructured EHR data. We have developed Bayesian networks and utilized machine learning methods in conjunction with natural language processing to predict 30-day hospital readmissions, detect diseases from emergency department reports, and classify the severity of psychiatric reports. Recently, we began developing an infant mortality prediction model based on EHR data and other available information. I will also demonstrate one of our currently deployed production systems, the System for Hospital Adaptive Readmission Prediction and Management (SHARP). This is integrated into the EHR system in place at the Children's Hospital of Pittsburgh of UPMC and demonstrates how the research in my lab can translate directly into practice.