Mayo Clinic last month released a study suggesting that an artificial intelligence (AI) algorithm could be used to better diagnose a certain heart condition called low ejection fraction—one step in the Clinic's effort to incorporate AI into cardiac care more broadly.
For the study, Mayo Clinic first developed an algorithm using a type of neural network aimed to screen for the condition, which occurs when the heart cannot contract with enough strength to pump out at least half of the blood from its chamber, MedCity News reports.
According to MedCity News, echocardiograms are typically used to diagnose this condition, they can be costly and time intensive. In comparison, the algorithm—which aims to catch patients who would otherwise slip through the cracks in routine care—relies on an electrocardiogram (EKG), a common screening device that assesses the heart's electrical signal. Patients identified by the algorithm as having the condition would then undergo the echocardiogram as well.
To test the algorithm, PCPs at 45 facilities in Minnesota and Wisconsin used it in their practices the during an eight-month trial. Over that time, more than 22,000 patients received an EKG as part of their routine care. Those patients were randomized into two groups, a control group, where they continued their routine care, and a separate group, where physicians were given access to the AI results.
Overall, the AI flagged 6% of patients as potentially having the condition, increasing the total diagnosis of low ejection fraction by 32% when compared with typical care. In other words, according to Xiaoxi Yao, the study's lead author and a health outcomes researcher in cardiovascular diseases, the algorithm yielded five new diagnoses for every 1,000 patients when compared to the usual standard of care.
According to MedCity News, Mayo—which has licensed the algorithm to Anumana, a joint venture between Mayo and nference—has submitted the algorithm for FDA review.
Broader AI initiative
The algorithm is one of several Mayo has developed for cardiac conditions. It represents one of the Clinic's initial steps to incorporate AI into cardiac care at Mayo and—via nference and Anumana—deploy the technology elsewhere, according to STAT News.
For instance, in addition to the low ejection fraction algorithm, Mayo has developed algorithms to detect pulmonary hypertension, hypertrophic cardiomyopathy, and atrial fibrillation (A-fib). For example, a recent study on historical EKG data found that the algorithm can accurately predict A-fib in roughly 80% of patients who were found to have the condition, STAT News reports. And Mayo is currently involved in a study to see whether the algorithm can better diagnose unrecognized A-fib, with the hope that it could eventually be used to diagnose the condition without need for further confirmation—enabling quicker treatment.
However, according to STAT News, even cardiologists who would like to further leverage AI into cardiac care are cautious about integrating the tools, since the information can both help by more effectively and efficiently targeting disease—or harm by potentially causing complications or spurring unnecessary care.
"As we're able to produce ever more precise information, the pressure on us to prove that we can use that knowledge to help patients is going to be ever greater," Harlan Krumholz, a cardiologist and director of the Center for Outcomes Research and Evaluation at Yale University, said. "We're still early in understanding how to harness this properly for the benefit of individuals."
And for patients with particular heart conditions such as A-fib, a diagnosis doesn't necessarily mean the patient would benefit from medical or additional medical procedures. But the use of screening automatically makes them more likely to get additional care regardless, Michael Rosenberg, an electrophysiologist at the University of Colorado.
"Screening only adds to the amount of health care people get," he explained. "That's why you have to be careful. It only goes one direction."
Still, patients who've benefited from the use of AI in Mayo's studies praise the advance. For instance, Peter Maercklein, a 73-year-old retired hospital finance executive at Mayo, was diagnosed with A-fib via Mayo's algorithm—a diagnosis that, because he is largely asymptomatic, would have been unlikely through his routine care, according to STAT News.
"For me, it worked out incredibly well," he said. "One of the best things about medicine now is they have all these remote monitors that can pick up more symptoms that they can jump on instead of waiting for something bad to happen—and then it gets a lot harder to fix and a lot more expensive" (Ross, STAT News, 4/26 [subscription required]; Reuter, MedCity News, 5/9).