RAPID: Prediction of Cardiac Dysfunction in COVID-19 Patients Using Machine Learning

点击翻译按钮获取中文摘要

Recent reports demonstrate the critical influence of COVID-19 on the cardiovascular system, with up to 20% of COVID-19 patients suffering acute cardiac injury. Approaches to identify COVID-19 patients at risk for cardiac dysfunction have not yet been developed, and no alerting clinical parameters are available to address the impending decline of cardiac function and mortality. The goal of this project is to develop a machine learning approach to identify COVID-19 patients at risk for cardiac dysfunction and sudden cardiac death. Utilizing such an approach will provide early warning and enable the delivery of early goal-directed therapy, reducing mortality and optimizing allocation of resources. The machine learning classifier is to be distributed to any interested healthcare institution, to augment their ability to successfully treat patients. This project also provides fundamental new scientific knowledge: how COVID-19-related cardiac injury could result in cardiac dysfunction and sudden cardiac death. Such knowledge is of paramount importance in the fight against COVID-19 and the post-disease adverse effects on human health. Features that will serve as input into the machine learning classifier will be extracted from both time series (ECG, cardiac-specific laboratory values, continuously-obtained vital signs) and imaging data (CT, echocardiography). Data will be collected from patients admitted to Johns Hopkins Hospital and Johns Hopkins Health System; other hospitals in the Chesapeake area; and potetially hospitals in NYC, with a confirmed diagnosis of COVID-19 based on nucleic acid or polymerase chain reaction testing. We will develop a time-varying risk score that will determine the posterior probability of hemodynamically-significant cardiac disease outcome within 24 hours of certain time points. For new patients, the model will be used to perform a baseline prediction which will be updated in a Bayesian fashion each time new data becomes available.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.