What does it take to add new digital biomarkers to health wearables?

Consumer wearables, like the Apple Watch, Fitbit Tracker, Whoop band, and Oura ring, are commonly used digital health wellness tools. They are part of a lucrative market, with an estimated value of $37.37 billion in 2024 for digital fitness smartwatches and smart bands. As of March 2024, Statista reported that 41% of survey respondents owned a wearable device. These wrist and ring devices aren’t there just to look cool, their sensors collect quite a bit of data from users.  

Fitbit began as a digital pedometer, syncing step count with mobile apps that stored the data and showed trends in activity. They reported the walking data back to the user with messages like “Congrats! You’ve walked 10,000 steps!” Fitbit hoped that with the positive reinforcement of meeting goals users would continue to log exercise for those praise endorphins. But eventually, step tracking wasn’t enough to attract new buyers. Tech never sleeps – it has to continue to iterate and increase product value to consumers to keep the company alive.  

Increasing product value often equals making new feature decisions. Over time, consumer wearables have become more complex, adding additional sensors that measure temperature, pulse, sleep, and heart rate. Companies both spit this data back at users with pretty charts and also develop proprietary algorithms like an opaque ‘readiness score’ to theoretically inform how hard to hit it at the gym. This isn’t regulated clinically relevant information. It’s firmly wellness data. To go beyond personal wellness data that encourages users to “Be the expert in you” per Oura’s marketing language, companies have to cross into the world of digital biomarkers.  

For consumer wearable companies to add digital biomarkers that are clinically validated they have to conduct research – and the quality of this research varies widely. The path of least resistance is to utilize the large datasets that the companies already have access to – that of their user’s data. Oura Labs specifically speaks to how “Analyzing member feedback and aggregated data is an integral part of our science and product development, both for improving existing features and building new ones.”   

Member data equals convenience samples for research. These convenience samples often lead to incidental findings that make for great marketing fodder, like that there are gender differences in sleep, stress, and activity levels. Researchers on the team look for correlations between bits of information that they haven’t seen before. For example, in 2022 Whoop announced that they had found changes in heart rate variability that could be indicative of preterm birth.  

As a researcher and women’s health enthusiast, I was excited when I saw the news about the Whoop preterm birth study, but when I dug into the published literature I concluded that a variety of things about the study were flawed. The study participants were already using the Whoop band, which would likely place them in a health-conscious population. The research was only conducted with singleton pregnancies, and didn’t collect race, ethnicity, or socioeconomic status information about participants, nor note any comorbidities like gestational diabetes or high blood pressure that could also affect birth timing.  

In theory, tech companies could “productize” convenience sample research findings quickly. I connected with Dr. Brinnae Bent, Executive in Residence at Duke University Pratt School of Engineering who shared, “There is currently no oversight into how large a study would need to be or how representative the sample must be of the intended users. In theory, you could pilot an algorithm on a few office mates for a week and ship the algorithm to production to your thousands (or millions) of users.” 

Brandon Mathis, a senior software developer, shared that from an engineering standpoint, most backend developers could build out these features based upon statistical models handed over from research teams. In the case of the Whoop study, if HRV inverted for X number of weeks prior to estimated delivery date, a user could receive an alert that they may be at risk for preterm labor. However, Mathis emphasized that for these algorithms to be meaningful they should continually be trained with additional data over time and tied to real reported health outcomes by patients.  

I choose to believe that the people working to develop smarter wearable health devices want to do the right thing. They want to continue to provide value, either for people to improve their own wellness management at home or to improve our understanding of human health with digital biomarkers. In order for companies to do both of these things they have to invest in quality research – both from a device validation standpoint and a clinical usefulness standpoint. The Digital Medicine Society (DiMe) collaborated with researchers, clinicians, and industry partners to develop the V3 Framework which outlines best practices to verify sensor performance and validate the algorithms so that it “acceptably identifies, measures, or predicts a meaningful clinical, biological, functional state or experience, in the stated context of use.” This is a cornerstone resource for device companies. DiMe also publishes case studies with examples of utilizing the V3 Framework and its updated version, V3+, here as a public resource. 

Ultimately, wearable technology companies are going to examine their user data for incidental findings in hopes of uncovering new correlations between sensor data and health outcomes, and should continue to do so. Science has to start somewhere. But to provide true value to the healthcare community these companies should not stop at convenience sample data before shipping features to the general population. They need to invest in larger research studies that, if done right, can add clinical and company value.  

Katie D. McMillan, MPH is the CEO of Well Made Health, LLC, a business strategy consulting firm for health technology companies. She is also a curious researcher and writer focusing on digital health evidence, healthcare innovation, and women’s health. Katie can be reached at katie@wellmadehealth.com or LinkedIn.