Program: Data Science Master's
Location: Not Specified (onsite)
Student: Lee Kostick
In treating advanced heart failure patients, healthcare providers must weigh the immediate clinical benefits and risks of implanting a ventricular assist device (VAD) with the long-term quality-of-life of the patient. Heart failure has been shown to be a prominent risk factor for stroke, and while implanting a VAD will help mitigate this risk, it does impact the likelihood and timeliness of receiving a heart transplant. Improvements in medical device technology and care management have led to better health outcomes in patients with a VAD. This has led to revisions in the heart transplant waiting list criteria, with patients with an implanted VAD receiving lower priority. Current approaches for predicting stroke in heart failure patients have shown modest results. This paper first proposes a methodology for identifying a cohort of advanced heart failure patients using electronic health record (EHR) data. From this cohort we then considered an Elastic Net regression and a Random Forest model to predict the risk of stroke or death from the onset of advanced heart failure. The primary objective was to lay a theoretical foundation for a model that would improve clinical decision making by quantifying the risk of adverse events in advanced heart failure patients.