Staying ALERT: AI puts spotlight on heart patients with untreated valve disease
Transcatheter aortic valve replacement (TAVR) is a safe, effective treatment option for severe aortic stenosis (AS). Time is of the essence, because patients are much less likely to live if severe AS goes untreated.1
Cardiologists have always wanted a better way to identify heart patients in need of care early in their course of disease. Artificial intelligence (AI) may be the tool physicians have been waiting for.
An ongoing challenge
Wayne Batchelor, MD, a veteran interventional cardiologist and president of the Medicine Service Line at Inova, says far too many patients who would benefit from heart valve replacement are still missing out on the care they need because they don’t receive a timely diagnosis. Even as techniques and technologies evolve and improve, he says, this remains a major problem—and one that all physicians must band together to solve.
“The good news is this is a treatable condition if the patient is detected in time,” Batchelor explains. “And we’ve developed fantastic surgical and transcatheter therapies to treat these valves. But not being treated puts the patient at risk. The stakes are high for patients, physicians, healthcare facilities and our communities at large.”
Determined to do something about this trend, Batchelor and a group of like-minded colleagues partnered with Tempus, a Chicago-based digital health company, and Medtronic, one of the largest medtech companies in the world. The researchers were impressed by Tempus’ AI expertise and believed its algorithms could be a game-changer for health systems looking to identify more patients who may benefit from care.
Could AI help find patients at risk of falling through the cracks? Could it be implemented into daily workflows without being a distraction? Those are just some of the questions Batchelor and his group wanted to answer.
Why are so many heart patients left untreated?
Step one was gaining a better understanding of just how many patients have untreated AS. In all, Batchelor et al. found that fewer than 50% of all patients who meet American College of Cardiology/American Heart Association criteria for aortic valve replacement actually receive the treatment they need.2,3
They also confirmed this issue consistently impacts some patient groups more than others.1 “There are still a lot of significant inequities when it comes to accessing heart therapies,” he says. “Women are less likely to get referred than men, Black and Hispanic patients are less likely to get referred than white patients and older patients are less likely to get referred than younger patients. And we need to all recognize those issues and try to learn from them or things are never going to get better.”
“These data suggest that a digital health strategy in combination with physician decision-making could result in significant improvement in care and may provide a reliable safety net for clinicians to make sure that they're not having patients drop through the cracks.”
- Wayne Batchelor, MD, President of the Medicine Service Line at Inova
Heart patients in more rural areas also are at a disadvantage when it comes to access to care. In small towns where the nearest hospital is a 30-minute drive away, for example, the healthcare ecosystem to identify and diagnose patients with untreated AS may just not be in place.
“In some cases, it’s just a matter of geography more than anything else,” Batchelor says. “If you don’t live in a region that has good access to contemporary valve therapies, you’re much less likely to get referred.”
It goes beyond inequities. Another major challenge is that patients seen by a primary care provider vs. a cardiologist are much less likely to be referred for an aortic valve replacement. Even among cardiologists, who know just how fatal untreated AS can be, referral patterns are often inconsistent.
Batchelor notes that there are certain clinical challenges as well. Defining and diagnosing severe AS can be complex, for instance, and patients who meet some criteria for a referral based on their echocardiogram results may not meet other criteria for referral. Another challenge is paradoxical low-flow, low-gradient (LFLG) AS, a category of disease that can be quite deceptive when clinicians review echocardiogram results.
“The aortic valve is severely narrowed in patients with LFLG AS, but because the heart is weak, the pressure gradient is deceptively low,” Batchelor explains. “The low gradient can fool physicians into thinking that maybe the valve isn’t severely narrowed—but it is. Those patients require care, but they just are not getting it.”
Putting AI to the test
The group’s plan was simple: use advanced AI from Tempus to scan echocardiogram results and then alert clinicians through the electronic health record (EHR) when a patient may benefit from being referred to a multidisciplinary heart team. The idea bloomed into ALERT, the largest randomized controlled trial (RCT) of its kind. ALERT recruited more than 750 clinicians across five U.S. health systems, randomizing them 1:1 to either receive or not receive AI-enabled alerts.
The AI software at the center of this study was designed to be both EHR- and vendor-agnostic. In addition, clinicians were able to help determine exactly what the alerts would say — and ensure they weren’t too pushy or demanding.
“With these alerts, we are just prompting the clinician to think about referring that patient,” says Batchelor, who led the ALERT trial. “We aren’t mandating anything; it’s just a suggestion. And the AI is agnostic to race, ethnicity, sex, age and any other patient factors, so the hope is that it will not be prone to some of the systemic biases we see so often in healthcare.”
Overall, the AI-enabled alerts were associated with multiple benefits. When clinicians received the alerts, patients were more likely to be evaluated by a multidisciplinary heart team (22.7% vs. 17.9%) and more likely to undergo a valve intervention (13.4% vs. 9.6%) than when clinicians did not receive the alerts.1
“That’s a 40% increase in valvular procedures, and they were all appropriate,” Batchelor says. “And there was no difference in the impact of the alert by age, sex, race, social deprivation, inpatient or outpatient setting, and whether the hospital was rural or urban. These data suggest that a digital health strategy in combination with physician decision-making could result in significant improvement in care and provide a reliable safety net for clinicians to make sure that they're not having patients drop through the cracks.”
It’s worth noting that results may vary by site since hospitals used different workflows and approaches to the alerts. The alerts rely on clinicians to act (patients aren’t automatically referred), and the program requires specific technology and resources that may not be available everywhere. In addition, outcomes were measured over 90 days, so longer-term impact is unknown.
Providing value to physicians and hospital leaders alike
While some physicians remain uncertain about AI, ALERT participants were excited to have the extra help. In fact, once the study was finished, they wanted to keep using the alerts going forward.
“The cardiologists were all pretty enthusiastic about keeping the AI,” Batchelor says. “Once you have something like that and you know it’s providing value, you don’t want to just give it away and go back to how things were.”
From a workflow perspective, the alerts also made a major impact. When heart patients undergo an echocardiogram, the goal is to get them evaluated, diagnosed and in the cath lab if necessary within 90 days. AI-enabled alerts helped care teams reach that goal, putting a spotlight on patients who may otherwise experience a delay in care.
Looking ahead
As time goes on, Batchelor expects that these alerts may become the standard of care when it comes to a variety of conditions — for cardiologists as well as specialists from a wide range of healthcare specialties.
“AI-enabled clinician notifications are going to be the way of the future,” he says. “The way things are now, clinicians can’t effectively manage the massive volume of data coming at them without a little help. We’re going to have to rely on AI and other tools to help clinicians make better decisions and deal with this deluge of data we are all being exposed to on a daily basis. This is a perfect example of how AI can be leveraged to work with clinicians instead of replacing them.”
Dr. Wayne Batchelor did not receive compensation for this article. He has been a paid consultant to Medtronic.
References:
1. Batchelor W, Lindman B, Coylewright M, et al. Automated Alerts to Improve Timely Evaluation and Treatment of Valvular Heart Disease: The ALERT Trial, JACC (2026). Jun, 87 (23) 3335–3346.
2. Brennan J.M., Lowenstern A., Sheridan P., et al. Association between patient survival and clinician variability in treatment rates for aortic valve stenosis. J Am Heart Assoc. 2021 Aug 17;10(16):e020490.
3. Li Shawn X., Patel Nilay K., Flannery Laura D., et al. Trends in utilization of aortic valve replacement for severe aortic stenosis. J Am Coll Cardiol. 2022 Mar 8;79(9):864-877.
Michael has more than 19 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.