Better Than Doctors: AI Predicts Heart Attacks

Using data from the medical records of more than 300,000 patients, researchers created an algorithm that can accurately predict 7.6 percent more heart attacks than doctors using standard methods.  Doctors have gotten very good at determining which of their patients are at high risk for a heart attack. But new research suggests that artificial intelligence may be even better — and that widespread use of AI could save many of the lives now lost to heart disease, the leading killer in the U.S.

In preliminary tests, AI computer programs developed at the University of Nottingham in England were significantly more accurate at predicting which patients were at high risk. The programs, which employed an algorithm based on data from the medical records of more than 300,000 patients, accurately predicted 7.6 percent more heart attacks than doctors using the standard method, which involves careful consideration of patients’ age, blood pressure, cholesterol levels, and other potential risk factors.

The algorithm also prevented the accidental tagging of patients as “high-risk.” That means these patients could safely forgo the medications often prescribed by doctors — and thus avoid the severe muscle problems and other serious side effects these cholesterol-lowering drugs sometimes cause.

What makes the algorithm so accurate? It analyzes more types of patient data over a longer time period. In addition to the risk factors commonly used to determine cardiac risk, it takes into account the interactions of certain medications that are now known to be associated with a heightened risk for heart disease. With this new level of sensitivity, the algorithm can predict heart attacks up to ten years in the future.

A paper describing the findings, published in the journal PLOS One, has created a bit of a buzz among doctors.

“I can’t stress enough how important it is and how much I really hope that doctors start to embrace the use of artificial intelligence to assist us in care of patients,” Dr. Elsie Ross, a Stanford University vascular surgeon who was not involved in the research, told Science magazine.

Dr. Stephen Weng, a research fellow at the University of Nottingham and leader of the team of physicians and computer engineers who developed the algorithm, told NBC News MACH in an email that it would be easy for doctors to use the algorithm in their practices. Once that happens, he said, doctors could quickly generate a list of at-risk patients — and arrange appropriate drug and behavioral interventions with enough time to stave off heart trouble.

But Weng doesn’t think computer programs will replace medical professionals anytime soon.

“You’ll always need doctors and nurses,” he said. “An AI algorithm won’t be able to tell you that the patient was nervous because they had a big job interview in the afternoon and was caught in a traffic jam on the way to the appointment and thus their blood pressure was high on the day. What AI does allow is for doctors to become more efficient at their jobs.”