UHN tests whether smartwatches can predict heart-failure outcomes

Researchers at Toronto’s University Health Network are midway through a groundbreaking study that is evaluating whether heart-failure patients can use Apple Watches instead of a cardiopulmonary exercise test (CPET) – specialized treadmills or bicycles based in medical facilities – to determine whether their condition is deteriorating.

Additionally, researchers are applying AI to the data they’re collecting to predict whether patients are getting better or worse.

The innovative watches can measure heart rate and blood oxygen levels – which is essentially what the exercise machines are doing when patients go to a clinic and are put on the treadmill with various leads attached to their bodies.

However, the Apple Watches have some major benefits. First, they can measure heart rate and blood oxygen levels continuously over a long period of time. Second, for many patients, it’s much more convenient to wear the watch than to go to a clinic.

“CPET is the gold standard, but it’s under-utilized,” said Chris McIntosh, PhD, a scientist at UHN’s Peter Munk Cardiac Centre who is a co-investigator in the study. He also heads his own lab researching applications of AI in healthcare.

“If you live in Toronto, you have a lot of access to CPET,” he asserted. “It involves a lot of expensive equipment and specialized staff. But if you live in northern Canada or rural areas, it’s difficult to get to. People aren’t going to get on a plane or take a long drive to go for a CPET test.”

Using a smart medical wearable, like the Apple Watch, could mean that the playing field is leveled for patients living in remote areas, as their heart health could be monitored continuously.

For its part, Apple donated the equipment needed to conduct the test – 200 Apple Watches, version 6, along with 200 iPhones. The lab at UHN created an app that automatically transfers data from the watch worn by the patient to his or her iPhone; that data is then sent on to the lab.

Measurements are automatically performed every six or seven minutes, and the data is analyzed using AI. Dr. McIntosh noted that he and his lab have created their own large-language model to apply to the study and decipher the results.

The study has produced a massive quantity of data, as heart rates, blood oxygen readings and ECGs are all picked up by the watches many times each day. That has led to the collection of “terabytes of data”, he said.

Dr. McIntosh observed that Deep Learning techniques have lately become extremely effective and they’ve been very useful when it comes to interpreting the results and making predictions.

At the start of the study, a CPET test was conducted on each of the patients to establish a baseline. During the test, the patient is encouraged to reach their maximum exercise level, to determine their maximum level of blood oxygen. The test is repeated after three months, the duration of the trial for each patient.

During the course of the three-month trial, patients are asked to perform various exercises, to test their abilities.

The project began in 2020, and Apple CEO Tim Cook was so excited about it that he tweeted to his network at the time. The study is led by Dr. Heather Ross, who is also the head of cardiology at UHN and a professor of medicine at the University of Toronto.

The team now has preliminary results and has submitted a paper to a leading journal; it is currently being reviewed.

Dr. McIntosh said he was unable to discuss the actual results before publication.

He noted that the 200 patients who were evaluated were in various stages of heart disease. He explained that HF patients are often in poor condition, and what is easy for a healthy person to do is often difficult for a patient with heart failure. “Even going up the stairs, dressing themselves, eating, can actually push them to exertion,” he said.

However, the goal is that by monitoring the patient over time, using the Apple Watch, researchers and clinicians will be able to more readily detect whether the patient is deteriorating.

“We want to understand the relationship between this free-living environment and the progression of disease,” said Dr. McIntosh.

Also factored into the study is whether the patients have needed to go to hospital during the time of the study – which is also an indication of poor health.

While some physicians and other experts have expressed skepticism about the accuracy of iWatches and other consumer-grade medical wearables, Dr. McIntosh said the length of the study acts as a corrective.

“We’re not collecting data just once,” he observed. “It’s collected over time,” and the results prove to be more accurate over the long term. If the study is deemed to be a success, Dr. McIntosh and his colleagues would like to see it expanded. It’s now mostly focused on patients from the greater Toronto area, and McIntosh would like to see it extended across the province and into other provinces.

Moreover, if the testing is successful, it would improve the care of HF patients. No matter where they live – whether it’s in an urban centre or the remote North – they could be continuously monitored. The data generated by smart wearables can be analyzed, as well, by AI-powered computers that are able to detect meaningful changes and send alerts to doctors and nurses.

“We can test people remotely, and we can test them more often than ever before,” said McIntosh.