This AI can predict heart attacks a year in advance
In the United States, cardiovascular disease (CVD) is a major health problem accounting for nearly 40 percent of all deaths each year, according to the US National Library of Medicine. Now, scientists from University of Utah Health have confirmed that artificial intelligence (AI) can better predict cardiovascular disease, including risk factors, onset, and course.
Fortunately, several risk factors for heart disease—such as tobacco use, hypertension, obesity, elevated low-density lipoprotein cholesterol (LDL-C), and hypercoagulable states can be modified. Much like detecting cancer early, therapeutic lifestyle changes and drug treatment can be highly effective at reducing a patient’s risk of heart attack and stroke if risk factors can be identified in patients early. But AI technology could improve this process, with the potential of saving lives before adverse events occur.
The Centers for Disease Control and Prevention (CDC) lists heart disease as the #1 leading cause of death in the US, followed by cancer and Covid. Cardiovascular disease is a grave concern in the field of medicine. Here in Utah, University of Utah Health researchers have been working closely with physicians at Intermountain Primary Children’s Hospital to develop computational tools that accurately measure the combined effects of existing medical conditions on a patient’s heart and blood vessels.
While the initial research is limited to cardiovascular disease, it’s only the beginning. Researchers see the vast potential of AI technology and how it can essentially help identify and pinpoint risk factors in a broad range of medical diagnoses.
“We can turn to AI to help refine the risk for virtually every medical diagnosis, [including] the risk of cancer, the risk of thyroid surgery, the risk of diabetes—any medical term you can imagine,” says Martin Tristani-Firouzi, the study’s corresponding author, a pediatric cardiologist at U of U Health and Intermountain Primary Children’s Hospital, and scientist at the Nora Eccles Harrison Cardiovascular Research and Training Institute.
The current methods for calculating various risk factors on cardiovascular diseases—such as medical history and demographics—are subjective and imprecise, says Mark Yandell, senior author of the study, professor of human genetics, and co-founder of Backdrop Health. Since these methods fall short, they fail to identify those interactions that can profoundly affect the health of a person’s heart and blood vessels.
Instead, the researchers focused on measuring comorbidities and how they influence patient health. Yandell, Tristani-Firouzi, and their colleagues from Intermountain Primary Children’s Hospital and U of U Health sorted through more than 1.6 million anonymous electronic health records (EHRs) utilizing AI. These EHRs included detailed information about patients, including lab tests, diagnoses, medication prescribed, and medical procedures, which helped researchers identify which comorbidities were most likely to aggravate cardiovascular disease.
“The important thing is that we can now calculate any outcome given multiple combinations of prior events in the patient’s medical record,” Yandell says. “This allows us to refine a patient’s risk for a medical diagnosis and understand how prior events influence future ones.”
The researchers found that patients with a previous diagnosis of cardiomyopathy, a disease of the heart muscle, had an 86 times higher risk of needing a heart transplant than those without cardiomyopathy. Individuals with viral myocarditis had about a 60 times higher chance of needing a heart transplant. Transplant risk for those who used the drug milrinone (used to treat heart failure) rose by 175 times—the strongest predictor of a heart transplant.
In certain cases, the combined risk was significantly higher. When individuals took milrinone and had cardiomyopathy, for instance, their risk of needing a heart transplant jumped to 405 times higher than individuals with healthier hearts.
“This novel technology demonstrates that we can estimate the risk of medical complications with precision and even determine better medicines for individual patients,” Josh Bonkowsky, director of the Primary Children’s Center for Personalized Medicine, told U of U Health.