Machine learning better predicts bleeding risk during coronary procedures
Machine learning techniques can better predict bleeding risk for patients undergoing percutaneous coronary intervention (PCI) than traditional methods, report Yale researchers.
The research team analyzed data from the American College of Cardiology’s (ACC) National Cardiovascular Data Registry (NCDR) from 2009 to 2015 using machine learning, a branch of artificial intelligence capable of performing tasks by inferring patterns in data. The database includes more than 3 million procedures conducted at hospitals across the United States. The team found that machine learning analytics improved the prediction of bleeding risk after PCI (often used to open up blood vessels narrowed by plaque build-up), which could better inform decisions by patients and doctors.
“We are discovering that machine learning may enable us to improve our ability to predict risk better than our traditional approaches,” said Dr. Harlan Krumholz, Yale cardiologist and director of the Yale New Haven Hospital Center for Outcomes Research and Evaluation (CORE). “Importantly, the key is in how the information about the patients is processed even before the analysis begins. In the future these techniques will enable us to personalize estimates to a much greater extent.”
“An exciting aspect of this study is the teamwork between experts across a spectrum of disciplines,” said Bobak Mortazavi, adjunct assistant professor at Yale and assistant professor at Texas A&M. “We found a deeper understanding of the problem and a better solution by working together and created a step toward tools to better personalizing each patient’s care.”
The team included clinicians, clinical scientists, and data scientists. This study is one of the first to employ machine learning to the massive registries of the ACC. CORE is a partner with ACC in the Institute for Cardiovascular Computational Health and this project is one of the first products of that collaboration.