FitBits Aren't Doing Their Job 50 Percent of the Time, Say Researchers

"As researchers, we need to be more cautious when using this type of data." 

Fitbit step counts

Smartwatch betrayal cuts deep. There are few things more frustrating than returning from a five mile run, only to have your watch rudely claim you only covered four. A new review published in JIMR Health reveals that some FitBits are guilty of that treachery. Fortunately, the team adds, there are likely improvements in the works.

Led by Lynne Feehan, Ph.D., a post-doctoral fellow in the Department of Physical Therapy at the University of British Columbia, this study pooled together data from 67 studies on FitBit accuracy, conducted both in the lab and in the real world, to determine what role FitBits should play in making healthcare-based decisions. With new algorithms looking to use wearables to diagnose medical issues, Feehan wondered whether FitBit’s technology was sophisticated enough for these lofty ambitions.

She isn’t alone. Consumers are also demanding accuracy from wearables: In an industry-based survey released in 2016, 29 percent of people who abandoned their wearable did so because they believed it wasn’t accurate enough.

“If you want to live a healthier lifestyle, there is a lot you can learn about your daily activity using these tools,” Feehan tells Inverse. “However, as researchers we need to be more cautious when using this type of data, because it could be used to inform public policy and healthcare decisions.” Such changes are already well within the realm of possibility: In 2017, researchers at the University of Illinois at Chicago introduced the possibility that FitBit data may be used to deny people health insurance.

JIMR calories burned Fitbit
Errors in Fitbit energy expenditure (calories) readings determined in a lab setting. The acceptable margin of error in a lab setting is three  percent, but it's ten percent outside the lab

On this front, Feehan’s results show that FitBit doesn’t make the medical grade. Measurements from FitBit’s wrist-based monitors didn’t fall within levels of “acceptable accuracy” (that is, within ten percent of the true amount of steps taken in a non-lab setting) fifty percent of the time. She and her colleagues also found that speed of movement tended to throw these wrist-based trackers off their game. In lab settings, wrist-based FitBits tended to underestimate distance traveled during brisk walking or jogging by as much as 15 percent and overestimated the amount of calories burned during normal walking speeds.

Fitness fanatics can take solace in the caveat that the FitBits actually used in these tests are older models — the studies used in this review were primarily published in the past two years — so these findings may not apply to models debuted recently, like the FitBit’s Charge 3. Feehan is optimistic that, as technology improves, FitBit will take steps to up its game as well.

“It is very likely that FitBit accuracy will improve with technological advances in the motion sensors in the device themselves,” Feehan says. “Evolving software upgrades that integrate advanced learning algorithms could help identify more complex patterns of motion.”