Using AI To Improve Cardiovascular Health

Yale University

In 2007, S. Raquel Ramos was working as a registered nurse in a cardiac catheterization lab in the U.S. Midwest. There, she assisted doctors with minimally invasive tests and advanced procedures to diagnose and treat cardiovascular disease.

But there was one experience she'll never forget. A young patient came in and told her it felt like someone was sitting on his chest. To be safe, the patient underwent an angiogram to visualize his heart arteries for any blockages.

"We were thinking, 'Oh, this guy is so young,' as it was not common for individuals his age to complain of tightness in the chest or to be getting an angiogram," said Ramos, who is now an associate professor at the Yale School of Nursing (YSN).

The angiogram revealed triple-vessel coronary artery disease where all three of the major coronary arteries are blocked. The patient had to go for immediate open-heart surgery.

"What was most striking about this is that he was only 36 years old," Ramos said. "He was the youngest person we had ever seen on the table. I've never forgotten that."

Today Ramos focuses her research program on cardiovascular health - and how to improve outcomes for patients like the one she provided care for nearly two decades ago. That includes the recent research development of CARDIO, a large language model designed to deliver cardiovascular health education in the clinical setting.

CARDIO is not a general-purpose language model. It was specifically fine-tuned on evidence-based cardiovascular guidelines and curated clinical educational materials to prioritize accuracy, safety, and clinician-aligned recommendations.

Filling in the gaps

Cardiovascular disease (CVD) is the leading cause of death in the United States, with one person dying of it every 34 seconds. Chronic conditions contributing to CVD include diabetes, high blood pressure, high cholesterol, obesity, and more. In 2023, roughly 76% of U.S. adults had at least one chronic condition. The majority of these chronic illnesses, however, are preventable. Taking up practical lifestyle and behavioral modifications - like walking daily, quitting smoking, or drinking more water and less sugar sweetened beverages - can help minimize risk.

For patients, CARDIO can provide information on heart disease and diabetes prevention in an accessible, evidence-based way. And with large language models like ChatGPT exploding in popularity, Ramos saw an opportunity to meet patients where they already are.

"People are sicker than ever, and the health care system isn't always able to meet the demands of patients, especially when they're so complex," said Ramos, who also has a joint appointment at the Yale School of Public Health's Department of Social and Behavioral Sciences. "But addressing risks through a tool like CARDIO can reduce multiple disease burdens and close large education gaps created by limited clinician time and high patient volumes."

"AI is exploding," she added. "It's something that you just can't stay away from. It's everywhere."

CARDIO is intended to complement, not replace, clinicians, Ramos said. To date, they have completed technical validation through a suite of engineering and expert checks confirming CARDIO's accuracy, guideline adherence, safety (including hallucination and harm testing), and robustness. The team is getting ready to launch the patient pilot test.

For the pilot, clinicians will meet with patients for typical appointments, during which they'll discuss their cardiovascular risk factors. Patients will then interact with CARDIO at the conclusion of these visits to go over some supplementary information.

"For example, if the clinician is talking about how a patient needs to take medications for their blood pressure, CARDIO could continue that conversation and explain, 'This is the reason why, and these are the steps you can take to make sure that you reduce your chances of heart disease risk,'" Ramos said.

Patients can use CARDIO by either selecting an icon or by typing in a question (similar to how ChatGPT works). The icons will link to suggested questions about various topics based on the American Heart Association's "Life's Essentials," a framework that focuses on enhancing heart health through lifestyle and behavior modifications. For example, if a patient selects the icon about physical activity, they might get advice on taking daily walks.

To develop CARDIO, Ramos convened a multidisciplinary team of experts in nursing, computer science, medicine, and public health - drawing on their skills and years of expertise across Yale University.

"We've all come together to solve this really pressing issue of how we can enhance patient teaching and health outcomes in people at risk for chronic conditions," she said.

/University Release. This material from the originating organization/author(s) might be of the point-in-time nature, and edited for clarity, style and length. Mirage.News does not take institutional positions or sides, and all views, positions, and conclusions expressed herein are solely those of the author(s).View in full here.