Medical Insights, Now Available on Your Wrist

Wearable devices like Fitbits, Apple and Garmin watches, and Oura Rings can provide users with a fairly comprehensive view of their own health––they provide data about resting heart rates and measure sleep cycles, record exercise segments and suggest modifications or new activities, and some can even use temperature readings to track menstrual cycles.

The ubiquity of these tools has made it easier than ever to capture long-term data about a person’s health. But for this information to be helpful, it needs to go somewhere beyond the wearer’s wrist.

That’s why transforming abstract biometric data into useful health insights is the mission of Jessilyn Dunn and the BIG IDEAs Lab (Biomedical Informatics Group: Integrating Data Engineering and Analytics) at Duke University.

“Our hope is that these wearable devices can fill a gap in the health care world,” said Dunn, an assistant professor of biomedical engineering and biostatistics & bioinformatics.  “Coming into a clinic can be a big burden for a lot of people, and the vitals and data they measure during the appointment only show a specific point in time––but health is continuous.”

Since starting at Duke in 2019, Dunn and her lab have been steadily working to identify digital biomarkers from wearable devices that can help recreate an overall picture of someone’s health. Their goal is to use these markers to identify health events like acute infections from diseases like COVID-19 and the flu, to more long-term problems like heart disease and diabetes.

“For a lot of these diseases, there is health data that shows that something was happening subtly for a longer period of time, and it got so bad that it led to something like a heart attack,” said Dunn. “We want to use the data from these tools and pair it with clinical data to indicate when someone may be at risk before these events happen.”

Identifying Markers of Chronic Disease

One of the major projects Dunn and her team are working on involves diabetes. The Centers for Disease Control and Prevention estimate that nearly 40 million people in the US have diabetes, but almost a quarter of that population doesn’t know it.

“One in three Americans is prediabetic, and that’s detected when a person goes into a clinic for a blood glucose test,” said Dunn. “Ideally, that’s representative of the norm for that person, but if that’s not the case, then the diagnosis and subsequent therapies would be missed. We wanted to come up with a more continuous metric that would give more accurate information about someone’s likelihood of becoming prediabetic and hopefully prevent the progression to type 2 diabetes.”

For this study, the lab collected data from 102 participants who wore a continuous glucose monitor for two weeks; 34 people had normal glycemic levels, 34 were prediabetic, and the remaining 34 had type 2 diabetes but weren’t yet receiving treatment. These different populations enabled the team to see the spectrum of disease progression. As the study progresses, they plan to compare this data to electronic health records and wearable and smartphone data to assess what biometric changes could indicate someone needs to come into a clinic to be tested.

“This could involve a combination of measuring the number of steps a person takes and their circadian change in heart rate,” says Dunn. “The combination of the change in multiple factors could actually be predictive of someone’s insulin sensitivity, as it gives us an idea of their heart’s function and their physical activity level.”

Allowing patients to gather these measurements via wearable devices also limits the need for invasive testing, as patients are much more likely to wear a smartwatch than they are to have regular blood draws.

Can Your Watch Catch COVID?

Another key project is “Infection Watch,” which explores how wearable device data can help detect early COVID infections. Working with the Center for the Biomedical Advanced Research and Development Authority, Dunn and her team collect smartwatch data from patients around the time they had a positive COVID test.

This work is an evolution of CovIdentify, a 2020 BIG IDEAs program that collected data from smartphones, smartwatches and health surveys to identify biomarkers that could indicate COVID infection. Participants were asked about social distancing and if they had common symptoms like nasal congestion, runny nose, cough, sore throat, headache, fever, chills, and COVID-specific symptoms like shortness of breath, nausea, and a loss of sense of taste and smell. Study responses were compared with biometric data, like sleep schedules, blood oxygen levels, activity levels and heart rate. Although the goal was to identify potential early COVID infections and suggest testing, the team also hoped the broad data collection would improve their ability to differentiate COVID from other illnesses, like allergies or colds.

This work also facilitated an ongoing collaboration with Dr. Chris Woods, the executive director of the Hubert-Yeargan Center for Global Health at Duke, to identify biomarkers for other infectious diseases like RSV and different strains of the flu.

“Wearables are a non-invasive and accessible tool that could help us control the spread of harmful diseases,” said Dunn. “These devices could help arm health care professionals with the information they need to intervene early and deliver the right treatment to the right person at the right time.”

Fixing the Blind Spots in Biomedical Data

But the findings from these studies are only as accurate as the smartwatches themselves. For example, many wearables were advertised as being clinically validated, even when there was no set standard for these devices. This meant that some companies could say that their heart rate measurements were accurate, even if they were only tested across a limited range of environments and body types.

Although Dunn and the lab have worked with the Digital Medicine Society (DiMe) to develop a framework to better evaluate and standardize the clinical usefulness of these tools, they know that problems still persist. Two of these problems involve who training data is collected from and evolving changes to the sensor technologies themselves.

Data for training such algorithms is expensive to collect and analyze, so researchers often work with data from easy-to-access sources. Unfortunately, this data usually comes from populations that are healthy, wealthy and white. This limitation often leads to bias, where the algorithms do not work equally well for all people under all circumstances. This can be a significant problem when studying diseases like COVID, which had a disproportionate effect on underserved communities, minorities and people with chronic illnesses.

Alternatively, the constant evolution and improvement of sensor technologies has also led to a complication known as drift, which occurs when algorithms are trained using data from sensors that collect data differently over time. For example, blood pressure could be measured by either the auscultatory method (i.e., a traditional upper arm cuff) or using optical signals similar to pulse oximetry technology.

“The question becomes, ‘Is this data actually comparable? If you develop an algorithm based on data collected from one technology, is it going to work as well for the data coming from a different technology?’” said Dunn.  “These are questions that we need to know so we can set guidelines about when to retrain a model or when more data is needed to improve the model.”

Dunn is optimistic that her BIG IDEAs Lab can take on these challenges, especially with the support of grants like the NSF CAREER Award and the partnerships she’s formed both inside and outside of Duke.

“Health happens 24 hours a day, seven days a week,” she said. “These tools give us an opportunity to provide health care tools to people who may not have easy access to standard healthcare in the first place, and we want to do everything we can to ensure that we’re successful.”