- A study of heart rate sensors on six wearable device models found accuracy fell by 30% on average when subjects were walking compared to when seated at rest. Skin tone variations did not significantly affect heart rate measurement accuracy, according to a group of researchers at Duke University's Department of Bioinformatics and Biostatistics.
- Fifty-three participants were studied in four activity states: resting, walking, paced deep breathing, and typing. Heart rates were measured against an electrocardiogram (EKG), the clinical gold standard.
- The results, published Monday in NPJ Digital Medicine, also revealed differences in accuracy by device manufacturer across the activities.
Wearables are now seen frequently, with 16% of Americans reporting owning a smartwatch by one recent count. Apple, Samsung and Fitbit generated 88% of all unit sales, according to the 2019 report from market research firm The NPD Group.
With growth in the use of wearables in healthcare outpacing predictions, the Duke researchers set out to answer questions about the much debated subject of data accuracy.
Heart rate sensors were tested for six of the most popular devices on the market: the consumer-targeted devices Apple Watch, Fitbit, Garmin and XiaoMi Miband and research-grade devices Empatica E4 and Biovotion Everion. The study found that different wearables are similarly accurate at resting and prolonged elevated heart rate, but differences exist between devices in assessing changes in activity.
Consumer-grade wearables were found to be more accurate than research-grade wearables at rest. Among the consumer wearables tested at rest and during physical activity, the Xiaomi Miband 3 had the highest error reading and the Apple Watch 4 had the lowest. The rhythmic movement of walking showed significantly higher errors in all devices except the Apple Watch 4.
The study also investigated concerns about whether fitness trackers are able to provide accurate heart rate readings for consumers with darker skin. A STAT News article, published while the Duke study was underway, described a problem with wearable sensors’ reliance on green light that is absorbed by melanin, making it harder to get an accurate result. That article cited a 2017 paper suggesting a link.
However, the Duke researchers found that skin tone did not significantly affect the accuracy of heart rate measurements for the wearable devices they tested. Equal numbers of people with skin tones from across the Fitzpatrick skin tone scale were represented in the study. Still, researcher Brinnae Bent said on Twitter the finding does not undermine those concerns.
We found that there was no difference in accuracy across skin tones. Although we did not find statistically significant differences in wearable HR measurement accuracy across skin tones, it doesn't invalidate past concerns with technology equity. (4/8) pic.twitter.com/aZBPumjWdd— Brinnae Bent (@RunsData) February 10, 2020
Apple itself is looking to leverage study data to demonstrate the usefulness of its devices. The company’s landmark Apple Heart Study released in November looked at whether a heart rate pulse sensor could identify atrial fibrillation, a type of irregular heart beat. That study, involving more than 400,000 volunteers, showed mixed results, generating a low rate of false positives but failed to notify some people of the condition who subsequently were diagnosed with AFib.