Online Program

Understanding and Predicting Acute Cardiac Events Using Electronic Health Records (302709)

Keywords: Electronic Health Records, Risk Prediction, Machine Learning, Risk Assessment

Electronic health records (EHRs) create a unique opportunity for medical researchers to understand and predict the risk of acute clinical events. Individuals undergoing hemodialysis (HD) are at increased risk of cardiac events such as myocardial infarction, stroke, and sudden cardiac arrest. EHRs capture detailed and changing information about patients' health status. One of the challenges in working with large data sets is defining the optimal way to analyze the data. In this talk, I will illustrate how different questions require different cuts of the data and different methodologies to best answer the question. Specifically, we will examine (1) a propensity-matched data set to assess the "causal" impact of a common dialysis drug, (2) functional spline regression methods to identify biomarkers for cardiac events, and (3) machine learning methods to derive a predictor and assess how far out an event can be forecast. By leveraging the large data size (22 million sessions across 100,000 people), we create tailored data sets to best address each question. The talk highlights both the potential and challenges for acute events with a dense set of data.