An estimated 40 million people in the United States have smartwatches or fitness trackers that can monitor their heartbeats. But some people of color may be at risk of getting inaccurate readings.

Nearly all of the largest manufacturers of wearable heart rate trackers rely on technology that could be less reliable for consumers who have darker skin, according to researchers, engineers, and other experts who spoke with STAT. Fitbit uses the potentially problematic technology in every heart rate tracker it offers, and it’s also in many Garmin and Samsung devices. Other popular trackers, like the Apple Watch, use it, too — but simultaneously track heart rates with another method.

The phenomenon has received almost no media attention, even as the market for smartwatches and fitness trackers has grown exponentially in recent years — and as both consumers and scientists have raised broader concerns about the trackers’ accuracy. There are a number of online complaints from consumers who suggest the devices can’t get a reading on darker skin. But the companies that make the devices don’t disclose the fact that they could be less accurate for some consumers.

advertisement

The potential inaccuracies have broad implications for the growing body of scientific research that relies on these wearables — as well as for the increasing number of people whose employers offer financial incentives or other benefits for using Fitbits and other trackers.

Concerns about the devices also come amid a broader reckoning over whether new technologies are as objective as they appear — and whether implicit prejudices are shaping their development.

“It really is about the existing bias in medicine that we have already,” said Kadija Ferryman, a cultural anthropologist who studies the social, cultural, and ethical implications of health information technologies at the Data & Society Research Institute. “No matter what [the] technology is, there is evidence that … inserting another tool, no matter how advanced it is, will likely continue on and continue to uphold the existing biases or exacerbate them.”

Nearly all consumer devices that track heart rate rely on optical sensors that continuously monitor the volume of your blood. In between beats, there is less blood volume at your wrist, and therefore more light that is reflected back to the sensor. But not all devices rely on the same kind of light.

Fitbits, Samsung watches, and several other brands rely on only green lights. These lights are simpler and cheaper to use than the infrared lights that power hospital-grade heart rate trackers — but, because green light has a shorter wavelength, it is more readily absorbed by melanin, a natural skin pigment that is more prevalent in people with darker skin.

In short: Skin with more melanin blocks green light, making it harder to get an accurate reading. The darker your skin is, the harder it gets.

“We could see that there [were] some differences in accuracy between, for example, different skin types and skin colors,” said Mikael Mattsson, a senior researcher at Sweden’s Karolinska Institutet and an author on one of the only papers to examine the link between heart rate tracker accuracy and race.

Fitbit spokespeople told STAT it boosted the current that powers its green light in an effort to address the issue. A spokesperson also said “Fitbit takes accuracy very seriously and our team continuously performs studies to rigorously test the accuracy of our products.” Apple, meanwhile, explained that while its devices rely on green light for continuous monitoring, the device also takes a reading with an infrared light roughly every five minutes. Garmin and Samsung didn’t return repeated requests for comment.

Researchers and scientists who spoke with STAT were careful to point out that there isn’t clear research that shows exactly how accurate consumer heart rate trackers are for people with darker skin — the issue has hardly been studied, in part because the technology changes so fast. But they also emphasized that the effect of melanin on green light absorption is well-documented — and that without more research or more public information from the manufacturers about accuracy, it is equally hard to prove there isn’t an impact.

Green vs Infrared Light
Infrared light (represented here in red) penetrates deeper into the skin than green light. Dom Smith/STAT

Vernon Ross, 49, a podcasting consultant in St. Louis, used his Fitbit Charge HR to help monitor his heart rate after he was diagnosed with a mildly elevated blood pressure in 2017.

But soon, Ross, who is black, noticed that the heart rate tracker wasn’t as accurate as he expected. It would fluctuate a lot, and his tracked heart rate didn’t always match the readings he’d take when he measured the old-fashioned way, putting his fingers on his wrist.

“Depending on what I would do, my heart rate would jump way off the charts and then fall back down,” he said. “That kind of stuff can cause quite a bit of stress if you think that you’re having, you know, a possible ‘AFib’ incident. … That could be a dangerous situation.”

Ross, who wrote to Fitbit about the issue on its website, said he has heard similar experiences from friends and hopes in the future, companies will provide data that clearly show that the technology works for people of every skin tone.

“It goes to a bit of a deeper issue, when you start talking about who’s in the room when these devices are being tested, who’s in the room when these devices are being designed, and … are the companies taking the time to make sure that the entire population is represented when they’re developing this technology,” Ross said.

The absorption effect is well-documented in research. But there’s far less research about how and whether that effect has impacted the accuracy of consumer heart rate trackers.

Just one study, a 2017 paper published in the Journal of Personalized Medicine, teased out the error rates for a host of devices that were on the market at that time. It found that while the technologies were “within acceptable error range” in most settings, device error was higher for those with a darker skin tone. The error rates for Fitbit Surge, Samsung Gear, and Basis Peak (v. 2) devices were statistically significant. Those for Apple Watch, Microsoft Band (v.1), Mio, and PulseOn were not.

Mattsson, the Swedish researcher and author on that paper, said that their study made sure to recruit a diverse body of participants — which he said was not emphasized in “many of the [early] validations from the companies themselves.”

Several researchers said the way Fitbit and Apple said they had tried to account for the issue could help. Boosting the current of the green light, as Fitbit said it has done, might help. So might Apple’s algorithm, if it balanced different readings between its green and infrared light sensors.

“The darker the skin, the stronger the current must be for you to get a good reading,” Mattsson said.

“The signal is not the only thing that matters, the algorithms really matter as well, so if you have algorithms that can account for those differences, then even though your signal is noisier, it may still be okay,” said Jessilyn Dunn, an assistant professor of biomedical engineering at Duke University.

But Dunn and Massimiliano de Zambotti, a researcher at SRI International who leads validation studies of wearable technologies in the sleep space, emphasized that without public data sets about error rates or public information about the algorithms, it is hard to tell if the companies’ efforts fully address the problem.

“We don’t know, because the pace of the technology is extremely fast. The pace of research evaluating this technology is extremely slow,” de Zambotti said. “We need to understand if we can trust their outcome or not.”

There are good reasons to use green light. Most wrist-based heart trackers use green light because it is better than infrared light in situations where the user is moving, or where there are other light sources that could also impact the sensor, the scientists told STAT. Infrared light, however, is better at penetrating more deeply into the skin, especially when the user is still — which makes it attractive in medical settings, they said. In a statement shared with STAT after this article was published, Fitbit’s chief technology officer, Eric Friedman, cited the same benefits as an explanation for why Fitbit chose to use “green light at sufficient strength” to deliver “the optimum, most consistent performance for users of all skin tones.”

The rapid pace of technological advancement makes it hard to tell exactly what color light different manufacturers are using in their latest heart rate trackers. Few make the information public, and it’s not always apparent from looking at the device. Some are even moving beyond optical sensing; Apple Watch now includes an electrical ECG monitor; a report just this month suggested the next Samsung watch will, too.

Representatives from Fitbit and Apple explained their technology to STAT for this article. A support page on Garmin’s website describes the light sensors in its optical heart rate trackers as “green.” Industry websites and videos show Samsung’s Galaxy Watch uses green lights.

STAT also spoke with two executives whose companies sell heart rate trackers that use multiple colors of light to address some of the shortcomings of green light, and both suggested that most of the popular trackers relied exclusively on green light technology.

“If you do a scan of the marketplace for measuring heart rate, most of the devices in the marketplace use green as a key component to measuring heart rate,” said Steven LeBoeuf, the co-founder and president of Valencell, which develops biometric sensors, including some that use both green and amber light. “If you were to dip your hand into a random pile of sensors, you would see that 9 times out of 10, they would have green light, and in probably most cases, they’re only green light.”

Sameer Sontakey, whose company Biostrap develops health trackers that use red and infrared light to track heart rate, explained that green lights are simply cheaper.

“It’s a commodity,” he said. “It’s easy to buy a sensor that has [green light]. There’s a lot of people that have worked with it. The barrier to entry is really low.”

“Inserting another tool, no matter how advanced it is, will likely continue on and continue to uphold the existing biases or exacerbate them.”

Kadija Ferryman, a cultural anthropologist at the Data & Society Research Institute

Accuracy issues could also mean that an increasingly robust field of research into heart rates doesn’t accurately reflect the heart rate information for certain individuals. Although numerous companies do indicate on their websites that their devices are not medically accurate, Fitbits are currently used in almost 300 clinical trials. Stanford University researchers also recently concluded the widely publicized “Apple Heart Study” of more than 400,000 participants from which they reported that “wearable technology can safely identify heart rate irregularities.”

However, the vast majority of studies performed that look into the accuracy of wrist-worn wearable devices, do not take skin tone into account when they test their subjects — a fact that Benjamin Nelson, a doctoral candidate in clinical psychology at the University of Oregon, thinks needs to change before this technology can be used for medical purposes.

“In the past, science hasn’t particularly recruited as diverse of participants as they potentially should, to be able to create generalizable results that will not only generalize to white Caucasian individuals, but will also generalize to people of different ethnic and racial minority groups,” Nelson said.

“It’s really important for future research to really collect and then control for skin tone because that has been shown in a few preliminary studies to actually influence the accuracy of these devices,” he continued.

Nelson also noted that larger sample sizes across various different demographic groups “may be able to control for different individual characteristics that will allow these devices to become more accurate in the long run.”

The potential inaccuracies could have an even bigger impact on people of color whose employers incentivize employees’ use of fitness trackers with extra vacation days, gear, or even lower health insurance premiums.

Already, big-name companies like IBM and Target offer trackers as part of wellness programs, and there are signs that employers may be beginning to make wellness programs — many of which include these trackers — mandatory. At one point, the West Virginia school system wanted to require teachers to wear them. Yale, too, was sued just this month over a program that would have charged anyone who didn’t participate in its employee wellness program a $25 a month penalty.

Twenty-one percent of large employers who offer health insurance are already collecting information from their employees’ wearable devices, a Kaiser Family Foundation survey found.

Philip Gordon, a Denver-based attorney at the law firm Littler Mendelson who focuses on workplace privacy issues, said the potential accuracy issues were yet another reason that employers should tread carefully before implementing any major wellness programs.

“It’s a good example of a potential hidden side effect of technology,” he said. “And that’s why it’s so important for employers to really understand technology and the information it’s generating […] before rolling out wearables in the workplace.”

Since he abandoned his Fitbit, Ross has switched to using an Apple Watch to track his heart rate. He thinks it’s more accurate — but he’s not sold on whether either is as good as just calculating his pulse with a timer.

“I do use it to monitor heart rate … But I also manually check my heart rate to make sure that I know exactly what my heart rate is,” he said. “It’s a good tool, it’s just not something you should depend on.”

Researchers echoed the same sentiment — that the tech is useful, but perhaps not as well-studied as it should be.

“These technologies are really being used so that we can collect heart rate out in real-world environments. Technologically, it’s one of the few ways that we can do it right now,” said Nelson, the postdoctoral candidate in Oregon. “So while there are a number of really important limitations that should be paid attention to by scientists, including things like skin tone, body mass index, and wrist circumference, it’s kind of the technology that we have right now.”

This story was updated after Apple clarified that it used infrared light in its devices for periodic heart rate monitoring, not to address the way melanin in skin absorbs green light, and to include statements from Fitbit officials. 

Leave a Comment

Please enter your name.
Please enter a comment.

  • https://www.youtube.com/watch?v=QTniLL0sQtA&t=3s

    If you want a quick and easy understanding of the technology used to measure heart rate on Apple Watch, Fitbit, Oura, Biostrap, I am a physician and have made a video on this topic. These devices use a very old technology called photoplethysmography (PPG). It is based on the technology hospital pulse oximeters use. Send me a message on here or via youtube and I will be happy to provide a clinical based answer.

  • What if you have a suntan? Too much faith is being put in these data-collecting monsters. Garbage in, garbage out, and we got a lot of garbage

    • Yes. It is the amount of melanin produced that affects how the green light from the optical sensors penetrate the skin. The more melanin, the more inaccurate readings you get.

  • I quit using my (expensive) Fitbit HR after I found conclusively that the ONLY circumstance when it was accurate was when I was resting. Exercising, fast walking, stationary biking, and lawn mowing ALL produced wildly inflated rates – compared to manually taking HR.

    Fooey.

    Stan, Binghamton, NY

  • Peggy, I have seen hand-drawn pulse waveforms from educational materials for the Chinese system of pulse diagnostics, and most of them bear little resemblance to real pulse waveforms recorded by plethysmography equipment. There is much of value to be learned from the study of real pulse waveforms, which varies with arterial wall stiffness, endothelial function (or dysfunction), etc. and it is available for study in the clinic using off-the-shelf equipment. We are still a long way from getting reliable pulse waveforms from a wearable because of motion artifact, ambient light artifact (in the case of photoplethysmograpy), acoustic artifact (in the case of MEMS microphones), etc. but progress is being made. I predict someday you will be able to buy a wearable that uses pulse waveform analysis (often called pulse contour analysis) to detect when you smoke a cigarette, drink a sugar soda, eat a high-fat meal, or do other unhealthful activities that cause endothelial dysfunction. Rapidly alerting the user (for example, by flashing a red skull-and-crossbones) may be useful in shaping the user’s behavior away from these activities, which may reduce the prevalence of type 2 diabetes, cardiovascular disease, and possibly Alzheimer’s disease. Today’s wearables can only give you heart rate, which is not nearly so informative or useful.

    • Mark,
      Thank you for sharing your enthusiasm for medical technology and sorry you took offense to my implied skepticism. We obviously inhabit different universes , though we both live on the same planet.
      Of course, Chinese Pulses was not, at all, the point of my post. The point had been the potential harm about over reading deliberate malfeasance (racism) into the now revealed limitations of Fitbit. Even if someone had that intent, we don’t know enough to pull that off. That was meant as a true, and also ironic statement. While we humans have accomplished some technological advances, some of our human our expressions, such as racism, remain crude. Let’s hope that should technology advance according to your predictions, we humans will also grow up and out of racism.
      Getting back to your objections to Chinese pulses, you said, “Most of them bear little resemblance to real pulse waveforms recorded by plethysmography equipment.”
      That’s the whole point. Medical technology measures the physical world, such as arterial wall thickness, in one moment at a time. Chinese Pulses don’t. They are about comparing energetic resonances ( or lacks) that flow through us and give our bodies life. Qi. The energy must flow smoothly for the body to function smoothly. That depends on much more than thicknesses. Like the difference between a tuned and an out-of-tune piano.

      To me, the finest medical device is our consciousness, which allows us to grasp what another is about, just from hearing and experiencing his/her voice, not seeing the voice’s wave form. Actually there are many realities we encounter that are only approximated by words, wave forms and digits. That is the beauty of Chinese Pulses. But it takes a human to take a Chinese Pulse and a very-well trained one, at that.
      But alas, someone figured out that we are not perfect, and decided the next best thing would be to make perfect instruments that would make perfect decisions so that we would have perfect lives. Perhaps create our own version of Garden of Eden – a stress-free life and retirement.
      Yes, I am a skeptic. Sounds like just one more Golden Calf after another. I would have hoped we would develop our consciousness and character to do a better job at being who we are. Flaws and all.
      No point in debating this. We will just have to ask G-d someday. 🤗
      Peggy
      Peggy Finston MD

  • Fitbit charge 3’s don’t track much of anything accurately! It’s very frustrating and I feel ripped off. They had to know how poorly this product works before they sold the first one. It counts steps with ANY movements including while driving. They should offer everyone their money back.

  • Aside from skin color, it’s possible that body fat, sweat, and quality and location of pulse (whether it is bounding or deep and weak) will likely effect results, too. One only has to attempt to learn Chinese Pulses to grasp how variable it is, even on the same person, at different times of the day and different times of the breadth cycle.
    To those who are inclined to label this “discovery” of errors as evidence of racism, your imagination is decades ahead of the actual science. Some errors are just mistakes. We like to think we know more than we do. Fortunately Fitbits are not designed to take us to the moon.
    Today, the best Personalized Medicine is not a device, but happens when a doctor has sufficient time to actually listen to patient.
    Peggy Finston MD

  • It is not accurate to say that these devices use green LED’s because they are ‘cheaper’ and easier to use. Medical grade ‘heart rate trackers’ are designed to measure blood oxygen saturation, so they use wavelengths (usually red and IR) that can accomplish that. Green light is actually more difficult to design hardware around (much larger forward voltage and less efficient), and is chosen because it results in better heart rate detection performance on ones wrist, even for darker skin tones (see section 2.2 for a summary of research on this: https://pdfs.semanticscholar.org/ca04/bcc2fdd9aed2b04e1f17c6bbb62f44dd3668.pdf). Darker skin does require much more power in order to get acceptable signal, and it’s likely the problem is more due to these wearables trying to optimize battery life in a way that degrades performance.

  • It’s not just skin tone, tattoos also can impair these sensors as was reported shortly after the original Apple Watch was introduced.

  • Thank you so much for writing this story. I have had this problem for years mainly with Garmin. I am a fitness enthusiast & like wearable devices. Almost every time Garmin releases a new product it doesn’t work on my skin tone. I call them and usually after 6+ months they release a fix. It’s maddening. I continually ask them why they don’t test their products on people of color & they never have an answer. What’s surprising is how often they use people of color in their ads & the product most likely doesn’t work on that person. Polar is another company who I’ve had the issue with. Apple has been great & I’ve never had an issue with their HR technology. Fitbit also has been ok for me as well as Oura & Whoop. It’s interesting because when I’ve spoken to white friends about this they literally have no idea this is an issue. I hope this receives more attention as this bias is only going to get bigger as stated by your article.

  • OMG, identity politics coming to a smartwatch near you. Next…the lawsuit. These companies use the best technology for the greatest number of users at the cheapest price. It is economics, not inherent bias. These are not medical devices. Unfortunately, hair removal devices also discriminate…against those with dark skin and dark hair, and those with lighter skin and white or blonde hair. How is this a story? BTW, everyone is a person of color, from pink to olive to tan to brown to black.

    • T Collins: have you not considered the well-documented fact that “economics” as we practice it has an inherent bias built into it? I do agree with Dr. Finston, though, that it is not a conscious bias, but one that just goes along with the lack of awareness that light skin is not the default, except to a select group of people who just happen to be the ones making decisions because our culture has a long history of treating light skin as superior, or more natural, or more… just more. Congratulations for noticing that we all are “colored”. Somehow, in the context you use the term, I am not reassured that your observation is meant as insight, but more likely as diversion from the problem noted here. And the fact that much of medicine is based on the response of white (male) subjects. That, my child, is the point of the article.

A roundup of STAT’s top stories of the day in science and medicine

Privacy Policy