As consumer adoption of wearable technology has soared in the past few years, there’s been a lot of excitement around how these platforms can be used to manage and improve health. For good reason — wearables provide the unprecedented ability to track an assortment of metrics 24/7, without the need for user input. As a computing platform, they are probably second only to smartphones in terms of convenience. But setting aside the hype, how much evidence do we actually have on wearables and their utility in healthcare? And what are some future directions for investigation?
We’re going to focus on the Apple Watch as a case study, since it is leading the way both in terms of market share (~38% of wearables market in 2019) and healthcare impact (through FDA clearances). To date, 135 papers have been published on the Apple Watch for various healthcare purposes. The first publications came out in 2015, shortly after the product’s launch, and the number skyrocketed in 2019, following the groundbreaking FDA clearances.
Current literature is centered around exercise and heart rate monitoring
The large majority of studies published to date on the Apple Watch fall into one of two categories:
Studies evaluating the ability of the Apple Watch to track exercise / energy expenditure / steps. Most of these have either been validation studies, evaluating the Apple Watch’s exercise metrics versus gold standard measures (example), or comparison studies, evaluating the Apple Watch alongside other commercial wearables like Fitbit, Garmin, and Samsung (example).
Studies evaluating the ability of the Apple Watch to record an ECG / monitor heart rate variability / detect atrial fibrillation. Most of these have also been validation studies (example) or comparison studies (example).
It’s not a surprise that the bulk of existing literature on the Apple Watch focuses on exercise and heart rate monitoring — the former is one of its most important selling points and the latter was the first “clinical” feature Apple decided to add.
It was, however, disappointing to see the paucity of research that has been done outside of these two areas. Researchers have conducted exploratory studies on a multitude of other health applications for the Apple Watch, but the current level of evidence is weak. If you think about research in any field as progressing through stages of maturity, the literature around exercise and heart rate monitoring has evolved to a stage where meta-analyses and systematic reviews can be published, while other Apple Watch features are still in early stages of investigation. For all the excitement around wearables and their potential impact on healthcare outcomes, the evidence is lagging behind.
Future studies should validate new metrics and evaluate existing ones as surrogates
Nevertheless, the exploratory studies in other areas do provide a glimpse of where research should go from here. The first direction is in expanding the number of metrics the Apple Watch can track and validating those in healthy populations. With the launch of the Apple Watch Series 6 this year, Apple added a blood oxygen sensor to the product’s health capabilities, allowing patients to both proactively and passively monitor their blood oxygen levels over time. I would expect that this new feature will spawn a wave of studies evaluating the accuracy of the blood oxygen monitor and comparing it to other existing devices.
Thinking more broadly, researchers are starting to use validated Apple Watch metrics not just as individual measures but as surrogate endpoints of sorts to monitor other health events. One timely example of this is a recent Mount Sinai study looking at whether heart rate variability (HRV) measurements from the Apple Watch could allow for early detection of COVID-19. Out of the 279 participants enrolled, there was a significant difference in the circadian patterns of participants diagnosed with COVID-19 (n=13) compared to healthy participants. Importantly, these differences were observable 7 days before initial diagnosis, suggesting that HRV measurements from the Apple Watch could have predictive capability for COVID-19.
Another example of this is a study published last year that used data captured by the Apple Watch to track sleep and distinguish between different stages of sleep. The 39 enrolled participants were given an Apple Watch to wear during the 2-week follow up period, and also came in for a laboratory sleep session where they were monitored with a polysomnogram (PSG), the gold standard for sleep measurement. The investigators designed a machine learning model which took data from the Apple Watch as inputs and was trained against the PSG gold standard data. Compared to the PSG, the machine learning model combined with the Apple Watch data distinguished time asleep with an accuracy of 93% and sleep stages with an accuracy of 72%.
A third example of this is a study looking at whether HRV measurements from the Apple Watch can be used to monitor stress. Twenty participants were enrolled and put through two stages, one which simulated relaxation and one which simulated stress, while wearing an Apple Watch. Researchers found that there was a significant change in several HRV parameters captured by the Apple Watch between the relaxed and stressed stages, suggesting the device could be used for passive stress monitoring.
Objectives should shift from monitoring to improving health
A second direction for future studies is to shift from using the Apple Watch for monitoring health to improving actual outcomes. This involves focusing on specific patient populations as opposed to healthy individuals, and leveraging data that can be captured from the Apple Watch to influence patient trajectories. One example of this is in diabetes, where the company One Drop has studied the use of an Apple Watch application to capture physical activity and allow patients to record blood glucose measurements, medication doses, and food consumption for diabetes management. Using the wearable to track food consumption was correlated with greater improvement, but this finding was limited by the fact that the study was not a randomized controlled trial.
A more ambitious application of the Apple Watch can be found in the Binge Eating Genetics Initiative, which plans to enroll 1000 participants to evaluate the impact of genetics, gut microbiota, and behavioral factors in bulimia and binge-eating disorder. The Apple Watch will be used both to capture active data, on binge-eating, nutrition, mood, and thoughts, and passive data, on physical activity and heart rate. The study is ongoing, but the investigators hope that incorporating Apple Watch data will enable deeper longitudinal phenotyping and advance the goal of personalizing treatment. Regardless of the results, the study should serve as a model for how wearables can be used more broadly to improve outcomes for other indications.
Scope should expand to real-world investigations
A third direction is to widen the scope of research from clinical experiments to real-world settings. As the Apple Watch’s health features are adopted by end users, data will be generated on how useful they are in practice — their effectiveness rather than efficacy. One instance of this is the pushback against the Apple Watch’s atrial fibrillation detection feature, which some cardiologists believe will do more harm than good due to the high number of false positives. The only way to resolve this debate is to study it, by leveraging things like claims data to evaluate the actual false positive rate and risk benefit equation in the real world.
Another important real-world question to study is the health economics of wearables. Several payers (like Aetna, United) have recently announced programs to provide their members with free or discounted Apple Watches, hoping that it will promote a healthier lifestyle and reduce healthcare costs down the road. There are, however, few publicly available studies evaluating whether this is actually the case.
One of these studies published last year by RAND Europe compared two plans offered by a South African insurance group — one which gave members the opportunity to earn rewards by being physically active, and another which allowed members to purchase an Apple Watch at a heavily discounted price, with the monthly payment amounts based on exercise levels. The results showed that the Apple Watch plan was associated with a 34% increase in the number of active days per month. The increase persisted for at least 24 months, supporting the hypothesis that wearables may have a measurable impact on long-term health outcomes and costs. The findings also suggest that loss-framed incentives (i.e., if you exercise more, you pay less) may be more effective than gain-framed incentives (i.e., if you exercise more, you earn rewards) in changing outcomes.
For now, the jury is still out on how much impact wearables will truly have in healthcare. Like many things in digital health, they have been heralded as a game changer for patients. Those hopes will have to be reconciled with a harsher truth — that healthcare is driven not by novelty but by evidence.