Monitoring our health with smartwatches

VIDEO: Wearable devices are increasingly used for tracking health data and identifying problems. Learn from health researchers about exploring the promise and pitfalls of personal tech.


Watch the replay of the event held on May 26, 2021. (Transcript below.)

Once geared toward fitness enthusiasts, Fitbits, Apple Watches and other wearable sensors are gaining a foothold in the medical realm. Whether keeping an eye on the heart, tracking seizures, monitoring glucose levels or surveying for viral infections, the devices hold promise as a way to gather valuable information that could guide health-related decisions. But how much can they tell you? How reliable are they? And how do you bridge the gap between just wearing a sensor and taking action — from going to the ER to making longer-term lifestyle changes? Join us for a discussion with two leading experts on the promise and pitfalls of the technologies that may already be on a wrist near you.


Jessilyn Dunn, Duke University

Jessilyn Dunn is the director of the BIG IDEAs Laboratory, whose goal is to detect, treat and prevent chronic and acute diseases through digital health innovation. Her research focuses on biomedical data science and mobile health and includes multi-omics, wearable sensors, and electronic health record integration and digital biomarker discovery. She also leads the CovIdentify study to detect and monitor Covid-19 using mobile health technologies. Dunn was a visiting scholar at the US Centers for Disease Control and Prevention and at the National Cardiovascular Research Institute in Madrid, Spain.

Mitesh S. Patel, Perelman School of Medicine and the Wharton School, University of Pennsylvania

Mitesh Patel is director of the Penn Medicine Nudge Unit, the world’s first behavioral design team embedded within a health system. His research focuses on combining insights from behavioral economics with scalable technology platforms to improve health and health care. He has led more than 25 clinical trials in partnership with health systems, insurers, employers and community organizations that tested ways to design nudges, incentives and gamification to change clinician and patient behavior. This work includes digital health interventions using wearable devices and smartphones, and health system interventions using the electronic health record.


Rachel Ehrenberg, Associate Editor,  Knowable Magazine

Rachel has been covering science and technology for nearly 20 years. She has a master’s degree in evolutionary biology from the University of Michigan and a graduate certification in science communication from the University of California, Santa Cruz. In 2013-2014, she was a Knight Science Journalism Fellow at MIT.


This event is part of Reset: The Science of Crisis & Recovery, an ongoing series of live events and science journalism exploring how the world is navigating the coronavirus pandemic, its consequences and the way forward. Reset is supported by a grant from the Alfred P. Sloan Foundation. 

Knowable Magazine is a product of Annual Reviews, a nonprofit publisher dedicated to synthesizing and integrating knowledge for the progress of science and the benefit of society. Major funding for  Knowable comes from the Gordon and Betty Moore Foundation.


More from Knowable Magazine

Annual Reviews articles, authored by speakers

Other online resources, authored by speakers

Other online resources


Rachel Ehrenberg: “Hello, and welcome to ‘Monitoring our health With smartwatches.’ I’m Rachel Ehrenberg. I’m an editor at Knowable Magazine from Annual Reviews. This is the 10th conversation in our ‘Reset’ series, which focuses on the pandemic, its consequences and the way forward.

“Today, we’re going to be talking about smartwatches. You or someone you know may have one of these gadgets, a Fitbit or an Apple Watch, these little wearable computers that talk to your phone and track your movement and potentially a lot more.

“During the pandemic, when gyms were closed and there were limits on outdoor activity, a lot of us became obsessed with counting our steps, even if just within our own home. And the devices also gained attention as a way to alert us to potential health problems, including perhaps a Covid infection. But the medical community’s interest in smartwatches predates Covid. Most of us see a doctor once a year. They record our blood pressure and other vitals but these devices can take measurements several times a day, providing a really astounding amount of health-related data.

“Smartwatches have been heralded as a way to detect A-fib, an irregular heart rhythm that ups the risk of strokes and affects millions of people. They might detect infection with a bacterium that causes Lyme disease, the onset of diabetes and more. But — and there are some big buts — a lot of studies that look at what smartwatches can do are retrospective; that is, they look back on data collected before or after the event of interest. There’s less research showing that they’re actually predictive of health outcomes.

“And there are also issues with accuracy — how skin color, for example, affects readings or even things like wrist tattoos. And, because these devices fall in the wellness category, rather than the medical device category, they aren’t regulated by the Food and Drug Administration in the same way that the devices used in your doctor’s office are. It’s a little bit of a Wild West in terms of industry standards and claims.

“All that said, smartwatches are really promising and in some realms such as heart health, they do have a real track record for successfully monitoring and alerting people to problems. So, we will also talk about how can we get these devices on the wrists of people who are most at risk? How do you get them to keep using these devices and even make healthy lifestyle changes that might be prompted by data gathered by the devices?

“I’m very happy to say we have two great people here to help answer all these questions. We have Jessilyn Dunn. She’s the director of the BIG IDEAs Lab at Duke. Her research focuses on using digital health technologies to detect, treat and prevent both chronic and acute health problems. Welcome, Jessilyn, and thanks so much for being with us.”

Jessilyn Dunn: “Thanks for having me.”

Rachel Ehrenberg: “You’re welcome. We also have Mitesh Patel, a doctor who also directs the Penn Medicine Nudge Unit, which uses behavioral science and various technologies, including devices and also games, to nudge people into making good health decisions such as getting more exercise, getting vaccinated, and more. Mitesh, thank you so much for being here today.”

Mitesh Patel: “Thank you for having me on.”

Rachel Ehrenberg: “Before we begin, I want to remind everyone to add their questions to the question box, which should be in the bottom of your screen. If questions aren’t addressed during the to and fro of the conversation, I will ask them in the last part of the event.

“Jessilyn, let’s start with you. I want to kind of get at the nuts and bolts a little bit of the technology that’s actually in the devices. They can do all kinds of things. They detect movement, blood oxygen levels, heart rate. Tell us a little bit about how they work and what they detect.”

Jessilyn Dunn: “Sure. I think you hit the nail on the head with the two most common sensors that are in the majority of smartwatches. We have an accelerometer that’s used to detect steps. And we have a PPG or photoplethysmography sensor that’s used to detect heartbeats.

“What’s interesting about these is that if we think from an engineering perspective about how the devices themselves work, they’re actually not step counters or heartbeat detectors, but they’re detecting some other measurement that’s a proxy for a step or a heartbeat.

“So if we think about an accelerometer, for example, the way that it actually works is we have, from the engineering side, a mass on a spring and when there’s some movement, the mass pushes the spring and maybe we have something that conducts electricity next to that, and that changes the conductance of the electricity. We call these electromechanical systems, which means that we couple a mechanical motion with an electrical signal. So, really, when we’re talking about the measurement of a step, what that actually means is that there’s some movement of a mass on a spring that conducts electricity that then leads to a detection of a movement. And we have algorithms that sit on top of that and convert that detection of movement into what is likely to be a step.

“Most people have probably seen that, if they are to look at a video of their own steps against a step counter, it’s not always exactly correct. And if you don’t step in the same way every time, steps aren’t always detected. So it’s important to think back again to how that system actually measures what we call a step, so that we can understand when it makes some mistakes.

“In terms of the PPG sensor, the same idea goes, where we don’t have a sensor that’s actually sitting on the heart measuring a beat, but we’re measuring changes in blood volume under the skin. The way that PPG sensor works is actually it shines light under the skin and detects the light that bounces back and the change in the light between what is shined through the skin and what bounces back changes when blood volume changes. And the blood volume changes with every beat of the heart in the cardiac cycle. What we’re really measuring then is a proxy for a heartbeat, because it’s the change in blood volume.

“Again, what we’re measuring is kind of a proxy for the actual phenomenon. I think, to your question of which of these sensors works very well, I think that the best way to think about it is to think about what is the sensor actually measuring and what are sort of proxy measurements, or abstracted measurements, on top of what the sensor itself is measuring.

“For example, the sensor is not sitting on the heart measuring a heartbeat, but it is measuring a change in blood volume and that change in light should be representative of a heartbeat most of the time. Hopefully that answers your question, Rachel.

Rachel Ehrenberg: “Yeah, yeah. I think that raises a good point of when people are thinking about accuracy, each level up that you extrapolate, so you have motion that’s a step and if you’re extrapolating up to maybe stress in terms of how much someone’s moving compared to how much they have been moving before, that’s another level of this in terms of this proxy, and another level of extrapolation in terms of what the device is actually monitoring. So that may be a good thing for people to keep in mind in terms of accuracy?”

Jessilyn Dunn: “Exactly. Yeah, and so…”

Rachel Ehrenberg: “Go ahead.”

Jessilyn Dunn: “I guess, just two other things that I think I would add are that the circumstances of the measurement also matter. For example, when we talk about a measurement of a heartbeat from the PPG sensor, if the sensor is moving around a lot, if your wrists are moving a lot, that sensor is going to be less accurate than if you’re sitting still at rest.”

“It’s important to not only think about what the sensor itself is measuring, but also the circumstances surrounding that measurement. The location on the body matters as well. The wrist is harder to read than the forehead, for example. Some of the reasons for that are the tissue components that you’re measuring through, as well as how much movement there is there. There’s a lot of factors that get taken into account when we’re talking about whether or not a measurement is accurate.”

Rachel Ehrenberg: “Maybe the next wave is going to be sweatbands with sensors. We’ll all be wearing sensors on our forehead. Mitesh…”

Jessilyn Dunn: “It’s possible.”

Rachel Ehrenberg: “… you and colleagues wrote an Annual Review paper on the success of these devices for monitoring the heart. This seems to be an area where they really shine. Can you tell us a little bit about what these devices are tracking in addition to what Jessilyn just talked about, why it works well in this realm and how those data can inform medical interventions, whether sleep or ECG? What works less and more well in terms of heart health?”

Mitesh Patel: “Sure. I think there’s a lot of interest in using these types of devices to remotely and passively monitor different behaviors that could give clinicians and patients themselves indications on their heart health. The most commonly tracked behaviors, the one that we all are mostly familiar with, which is step counts for physical activity, you can actually use the cadence or how fast you’re walking to [see] if your activity is light, moderate or more vigorous. There’s good evidence that more vigorous activities is better for your heart, but that even some light activity is better than no activity.

“Our group did one of the first evaluations of the accuracy of step [counters] probably more than three or four years ago, when we took 15 participants, put them on a treadmill and kind of put a bunch of smartphones, wearable devices, and then actual pedometers on them, and kind of counted — we had a medical student count with the clicker — 500 steps, a shorter distance, and 1,500 steps a longer distance. Then they got off and we looked at all the devices to see how accurate they were.

“We found, for the most part, smartphones and wearables were fairly accurate. The smartphones were actually more precise in that they had less variability and probably because they were placed in the pocket, as mentioned these devices… As I’m sitting here and talking and moving my hands, I’m getting step counts on my wearable device that aren’t real step counts and people have different swings, and so that can affect the accuracy but overall, as these devices have evolved, physical activity measurements have been found to be fairly accurate for giving you a general sense of if you’re active or not.

“Other types of behaviors are things like sleep. Data on sleep has been less accurate. It’s, again, a calculation that’s done based on your movement. There’s been some calorie count estimations using these devices. Those are even less accurate, because again, these are calculations all really based on movement.

“The one that’s more recently gained traction and actually, with clinical trial evidence, is detecting heart rate, specifically looking at ECG and trying to detect whether or not someone has atrial fibrillation. There was a large study done by Apple in partnership with Stanford and other health systems to really enroll over 100,000 patients into the trial to see if they could detect when people had atrial fibrillation or other abnormalities. The devices were found to be fairly accurate, although they were done mostly in a younger population. There’s still questions about how these might perform in a more high-risk, older population.”

Rachel Ehrenberg: “OK, and Jessilyn, you were a coauthor on a Nature Medicine paper that came out this week that looked at clinical results and wearable data. There was very positive outcome. Can you tell us a little bit about the study design and the findings?”

Jessilyn Dunn: “Sure. Our interest with this paper was to see if there was any information in wearable device data that represents something that we would normally measure in a clinic that requires kind of an invasive test. We had a very interesting study population that we started during my time at Stanford, where we had a group of people that were coming in regularly to the clinic to give samples for sort of standard clinical testing. This was part of a larger study called iPOP that’s run through Mike Snyder’s lab. The folks in the study were also wearing smartwatches.

“We could actually look at the relationship between the smartwatch data and the measurements that were taken in the clinic. When I say measurements taken in the clinic, I mean, something like complete blood counts, so we have measurements of hemoglobin, hematocrit, red blood cell counts. We have white blood cell counts that would be representative of maybe some sort of an infection. We had, basically urine and blood analytes, all sorts of measurements. And we built machine learning models to explore whether there was any predictive capacity of the wearables to actually predict any of these analytes.

“What we found was that there actually is a relationship between a lot of the things that we measure in our blood and the measurements from the smartwatch itself. So when we’re looking toward the future of some goals of things like Apple and Fitbit to, say, maybe measure blood glucose from a non-invasive wearable, this provides evidence that that is really possible. The reason for that — it makes sense, because we know that when we’re measuring an optical signal, again, we’re not just measuring a heartbeat, we’re measuring all of the changes in the absorption of light under the skin. That can represent a lot of different things.

“In this particular case, we had four sensors: We had a PPG sensor, a temperature sensor, an electrodermal activity sensor and an accelerometer. We could create these aggregate metrics from these multimodal sensors that enabled us to get more information than we otherwise could.

“One other piece that’s quite interesting about this study was that we actually have the ability to compare measurements taken in the clinic for vital signs to measurements taken with a smartwatch. What was really clear is that there’s this time-dependent pattern of our heart rate and body temperature that is affected by our circadian rhythms. If we think about trying to track our heart rate over time with clinic visits, when we go into a clinic visit it might be at any time of the day. So if we’re comparing a morning heart rate with an evening heart rate, those actually are quite different and not necessarily because our body is changing over time, but potentially because of the circadian variation. We also showed that the wearables are really powerful in getting a really precise, consistent, resting heart rate measurement.”

Rachel Ehrenberg: “Great, that’s really interesting and that leads right into this issue of accuracy. I know we’re sort of talking generally about these devices and not specific brands. So I want to touch on kind of regulation standards in the industry. Jessilyn, there’s been concerns about the green light sensors, which I’m not sure some, if all brands, use to detect the volume of blood, they’re less accurate on darker skin because the melanin absorbs the green light.

“There’s also, we’ve spoken about fat content, hair, there was tattoo gate, tell us a little bit about things on the wrist whether part of the body like skin color, practically part of the body like a tattoo, other factors and how that affects measurements and then also in the industry and perhaps regulatory agencies, what’s being done to standardize these things, if anything?”

Jessilyn Dunn: “Right, yeah, this is a real problem that the green light, that wavelength is absorbed by melanin. Kind of like what I was saying, just before that, we’re really measuring a proxy for a heartbeat because we’re shining light through and measuring light that gets reflected back. If we have a higher content of melanin or something else that is absorptive of that light, then it’s going to change what we can receive back in the photodiode.

“This can be overcome in a variety of ways. Actually, what we showed in our recent work was that, although this is a known issue with the green light sensors, you can layer algorithms on top of the measurements that we get back in order to account for that. There are also ways to change the intensity of the light depending on the content of the melanin, so that you can actually increase the amount of light that can get absorbed back.

“Now, this brings up a new question of battery life and whether that kind of method has any effect on battery life. That remains to be seen. There are also options of using multiple different wavelengths. There are some devices that are now coming out that are using red, orange, yellow and green wavelengths together that are improving heart-rate measurements. We know that the wavelength of light — so the color of the light that we see — really matters.

“There are lots of other things that can affect the accuracy, like we’ve talked about. This is an interesting question of how we regulate the accuracy of these devices or how that accuracy is reported by the different wearable-device companies. Right now, these devices are defined as wellness devices. They’re not defined as clinically oriented and therefore they are not regulated by the FDA. That is really important for people to know that if there is a claim made by one of these wearable device companies on a wellness device, that’s specifically wellness, that that claim is not regulated by the FDA.

“Now, what gets confusing is when you have a device that performs multiple different functions, where some of those functions are FDA-regulated and others are not. One example of that is the Apple Watch atrial fibrillation detection algorithm, which has undergone FDA review and approval but that same device also measures sleep and step counts, and those actually haven’t undergone FDA approval.”

“There’s a lot going on in this space. I work closely with the Digital Medicine Society, also known as the DiMe society and together we work closely with the FDA, which has a new center for digital health excellence that’s really looking into these challenges, and how we can go from toeing this line between wellness devices and devices that are going to be used for clinical decision-making.”

Rachel Ehrenberg: “Can people, is there anywhere for the consumer that you can look up how well your particular device works on a particular measurement or is it sort of just what’s on the box when you buy the device and then Google?”

Jessilyn Dunn: “Some companies are starting to report out the accuracy of their devices. It’s rare, but it’s becoming more common that companies will put out some of the results of their internal testing. You may see on company websites white papers that give some information. It’s important to read those as if you were reading kind of the scientific literature, making sure that these were large and diverse study populations, and that the studies were done in a way that is rigorous and ensures sort of that the results that are reported are believable. In many cases, companies are doing that and I think are continuing to do this more openly to help build trust in their products. I hope that we continue to see that happening.”

“On the other hand, we also have researchers who are doing these head to head tests. Unfortunately, that’s often not supported by funding agencies. This is something that we sort of do on our own out of curiosity or because we need to in order to evaluate devices for future studies, but there is plenty of literature now coming out through the academic or sort of independent researchers on this as well.”

Rachel Ehrenberg: “OK, let’s move to actually getting people to perhaps make lifestyle changes via the information that they’re learning from this device. Mitesh, you spend a lot of time thinking about how to actually get people to change their behavior, given their personal health data. You’ve talked about issues of access, accuracy, sustainability and behavior.

“There’s a lot of people who maybe have smartwatches who aren’t the people who are most at risk of A-fib or another condition where a smartwatch prompt might be lifesaving. Can you sort of talk us through these points you’ve raised and some of your work on nudging people to take steps in terms of changing?”

Mitesh Patel: “Yeah, sure. We’ve kind of put out a framework thinking about what are the four key areas that we need to address for wearable devices to have a big improvement on population health; improving the health of Americans and so on and so forth or people in other countries.

“The first challenge, as you mentioned, is access. The people who could benefit the most from wearable devices, the ones who have uncontrolled diabetes or being consistently readmitted to the hospital are probably not the ones that are wearing wearable devices. People who wear wearable devices tend to be younger, have higher incomes, are more tech-engaged, more conscious of their health. That’s one challenge. If we really want to make an improvement in population health, we have to find mechanisms or ways for people who need these devices to get them.

“Obviously, some people could be gifted this device. That happens a lot around Christmas time but more and more, what we’re seeing, is that large employers or insurance companies are actually subsidizing the cost of these devices as part of workplace wellness programs. Many people don’t know that actually, within their own workplace wellness program, they can get a subsidized device or through their insurance program. That’s a benefit that if you’re interested in using a wearable device, we often suggest looking into.

“Many workplace wellness programs also have incentives for connecting your wearable device to your smartphone app to a platform that allows you to check in when you’re at the gym or monitor your step counts. Let’s say we get someone who needs to change their behavior access to a device, then the next step is making sure it’s accurate. We’ve talked a bit about that already. There’s a lot of challenges because many of the devices do not have clear evaluations that are published. These things are done rather sparingly. That is a challenge.

“Step counts, as we mentioned, seem to be fairly accurate and some other behaviors are less accurate, but in order for people to trust the data, they have to be able to rely on what it’s telling them. That means there needs to be some way of determining whether or not they’re accurate.

“Let’s say you have someone who needs to lose weight or control their blood sugar or whatever it might be and you got them an access to a device, they trust the data behind it. Now, you’ve got to get them to sustain using the devices. We think sustainability is a major barrier. About a third of people who get a wearable device stop using it within three to six months.

“We did a randomized trial at our hospital where we enrolled 500 people who are being discharged from the hospital. These are high risk patients with heart disease, diabetes, renal failure and other conditions. We randomly assign them to use either a wearable device on the wrist or a smartphone app. The goal was for them to be remotely monitored, all their activity patterns for up to six months. We actually found that people who were randomized to the smartphone app were 32 percent more likely to continue to use the device at the end of the six-month period. The difference was about a 16 percentage point difference from 32 percent to 48 percent.

“One of the reasons is because you’re carrying a smartphone with you wherever you go and you’re not going to stop using that just because you’re not interested, whereas with a wearable device, you might take it off and never put it on again. Some of the factors that we’ve seen, that make wearable devices more sustainable are finding ones that don’t require you to take it off. The ones that require daily charging, that’s an opportunity for you to not put it back on. Some of them could go up to a year without charging.

“The ones that are waterproof, people won’t need to take off when they go in the shower or they go for a swim, really reducing any opportunities for them to do that. Many of our studies, we tell people, even though we’re focusing on things like physical activity, we tell them, we’re also monitoring your sleep. You should wear the device to sleep, just to encourage the habit of continuing to use the device.

“Let’s say you’ve done all of this: You’ve figured out the access problem, making sure it’s accurate and you’ve figured out a way to use a device that’s more sustainable. Now, comes the biggest challenge, which is changing people’s behavior.

“This concept that you mentioned of nudge, our team runs a nudge unit at Penn Medicine and the idea here is that people are highly motivated by small changes to the environment, the choice architecture, and they’re predictably irrational, that people are more focused on kind of immediate gratification than longer-term rewards. They’re more motivated by losing something that they have than gaining an equivalent amount of it.

“We use those elements and we design either social incentives, which use kind of a game-based feature, or financial incentives to try to motivate people, like we’ve done through a workplace wellness program. One example: We did a study, a national study with employees at Deloitte’s consulting group, where we randomized people to four different arms: The control arm, where you just use your wearable for six months, or one of three game-based arms, but we flipped the game on its head.

“Usually, you have to do something like get your step count, and then you get points. Instead, we gave people points upfront that they could lose if they didn’t get their step count. Then, we baked in social incentives. You either competed with two other members in the organization, you collaborated with them, or you picked a family member or friend to get support.

“We found at the end of six months, and then through an additional three months of follow-up, the competition for this group of people work the best, maybe not surprising for a consulting firm. We’ve done similar studies in families in the Framingham Heart cohort and found collaboration and support have also worked really well, but these are really kind of just ways of sending one text message a day that gets people to build a habit. We find when we turn the text message off, that these habits are sustained.”

Rachel Ehrenberg: “Interesting, we definitely get email threads going in this office of how many steps, who has what today, going for a walk, here’s my number. So, yes, this is a lot of great information in terms of both how to trust them and maybe how to get people to use them.

“I do want to just touch on a project of yours, Jessilyn. I first heard about your research because Matthew Hutson wrote a story for Knowable on adapting smartwatches to detect Covid infections. Tell us about the CovIdentify project, how it started, where it’s at. And you’ve said that people can still submit data?”

Jessilyn Dunn: “Yup. We started CovIdentify back in March 2020. We had done quite a bit of work already in the detection of infection with wearables. The reason for that is that we know that there are physiologic changes and behavioral changes that occur when people get sick. We can adapt those principles to try to detect something like Covid.

“The idea here is that these physiological and behavioral changes aren’t super specific, right? It might be a challenge to distinguish maybe between Covid and flu, for example, but the fact that we can say, “You’re likely to have something going on in the middle of a pandemic, and you should go get tested for Covid,” gives us that extra layer of information about directing testing. That’s kind of the way that we were thinking about this was, especially at the beginning of the pandemic when there were limited tests available.

“We needed more intelligent methods of deciding who should get tested and when. If we test somebody today and then they go out and get infected tonight, then they may assume that they’re fine, but they actually might have Covid and go and spread it. For sort of the public health perspective of how can we best curb the spread of this infection, we were thinking that having intelligent testing over time and deciding who should get tested would make a lot of sense.

“And so we started this study. We enrolled about 7,000 people and we were able to see kind of the changes that we expected. There were other studies going on similarly at the same time because this is an area that made a lot of sense to a lot of folks working in wearables, who knew that we can detect infection from changes in heart rate, changes in movement, changes in sleep, and what was really cool about this is that the community of researchers in wearables were really able to show really similar phenomenon.

“We have these increases in heart rate. We have increases in variance of sleep. We have decreases in quality of sleep. We have decreases in overall physical activity that correspond with infection. And what’s even more interesting about this is that our algorithms to detect infection actually are able to detect that infection often before somebody themselves feels sick. The power of that is huge because we could actually get somebody tested before they would decide to get tested, which means that we could prevent the spread of an infectious disease before somebody might even know that they’re sick.

“That’s really what we’re trying to get at here. We’re working to evolve that technology. We were working on this for flu and rhinovirus even before the Covid pandemic had broken out. We will continue to adapt these technologies for infectious respiratory diseases and hopefully beyond.

“That being said, we do still have the study open. So if you or somebody you know had Covid and had a wearable device, we accept data from pretty much any wearable device that’s commercially available. We can go back, add that data to our current algorithm and improve that algorithm even further. Please, if you or somebody you know had Covid and had a wearable, please do consider donating your data to the study.”

Rachel Ehrenberg: “Great. I think if we don’t already, we’ll have a link to CovIdentify in the resources section on the event page on I have a lot of questions, but so do our listeners. Mitesh, you are actually changing jobs soon. You’re joining Ascension for a kind of position created for you, a nudge position. It’s one of the largest private insurance companies. Jessilyn, you’ve just made an appeal for data.

“We already have some questions coming in about privacy in terms of HIPAA applying or not applying, insurance companies, researchers. What are the issues around privacy, the sensitive data being repurposed and sold, linked to our identity? Which of you wants to go first with that one?”

Mitesh Patel: “I’m happy to go first. I think privacy is really important, as is security, particularly with sensitive patient data. I think one of the things that there’s probably room to improve upon, but is really important, is transparency around what the devices are collecting, how the data is being used, and so on and so forth. A lot of that information is buried in kind of the things that you sign, when you sign up for any app, you just kind of scroll through and click accept and people don’t really recognize that, but making it simple and straightforward in a way that you can do that.

“We’ve had some experience with in our clinical trials, where we’ve tried to simplify that and focus on kind of the main things, but then provide the more detailed information for people that want that, and generally have had good acceptance when being transparent about these things. I think when people feel like there’s an opportunity to use this data to improve their health, that’s really been beneficial for them. They’re oftentimes willing to engage; it’s this idea of not knowing or these other elements that make it more challenging and can cause people to lose trust in some of these applications.”

Rachel Ehrenberg: “Sorry, I muted there. Jessilyn, do you want to just add about in terms of academic researchers and best practices when you and other labs are collecting data? I assume there are steps for in terms of privacy. Go ahead.”

Jessilyn Dunn: “There are quite a lot of steps. With any, what we call a human subjects study, which means that humans are involved in donating data or coming in to give samples, we have to go through an institutional review board, which is sort of an ethical review of the study. There are a lot of best practices when it comes to ethics surrounding human studies research.

“One of those is privacy and security of personal data. We have different levels of what it means for data to be sensitive. When it comes to wearables, this is kind of a whole new world because we’re learning that there are a lot of things that you may be able to identify from wearables that maybe weren’t readily apparent at the outset, similar to the genomics world.

“At Duke, in particular, and previously at Stanford, there were layers of different offices that cover privacy and security and that cover information technology and ensure that we have the proper layers of encryption whenever data is moving, that we have appropriate levels of protection for who can access what level of data, and we work really hard on de-identification methods, so what it means for data to be identifiable.

“For example, if I have one year’s worth of step counts from an individual, but I have no other information about them, is it possible that I could link that information back to somebody and re-identify them to figure out who that person is? A lot of this is still in the research phase. We’re still learning about what may be possible in terms of de- and re-identification. It’s great to see that there’s research expanding in this area. It’s great to have the resources in the university and otherwise, to learn new methods for handling sensitive data.”

Rachel Ehrenberg: “Another viewer asks about sort of accuracy and focus in terms of one person is using Fitbit in connection with the Cronometer app to maintain a healthy diet and stay active. They sync with each other. He or she says, ‘I’m a little worried, I’m over-focusing on this. How correct are either of them?’ I’m not going to ask you guys to go through device by device in terms of accuracy.

“We touched a little bit on which measurements and what they’re a proxy for are likely more accurate than others. But how does just a regular person, what’s the kind of healthy mindset in terms of interacting with your wearable? How much stock to put in it, but also, using it to actually promote changes that might be really positive? Mitesh, you want to start with that one?”

Mitesh Patel: “Sure. I think there are a couple of important things that we often try to share with patients or individuals who are interested in using wearables to change their behavior. The first is that it’s important to try to set a goal and understand where you’re starting from. Using these devices to just get some sort of baseline, let’s say, it’s for step counts, how many steps you walk at baseline over a week or two, and then try to set a goal for higher than that. There’s evidence to show that people who set goals and then kind of review the data in comparison to whether or not they’re achieving their goal, are more likely to succeed at changing their behavior than those that don’t.

“The second thing we often share with people is that don’t focus too much on the specific number, whether you got 7,258 steps or 7,265 is not really important. What’s really important is the overall trend. Have you increased on average your step counts by 500, 1,000, 2,000 steps? So worrying more about whether or not you’re moving up in general and the trend is in the right direction, as opposed to this specific number.”

Rachel Ehrenberg: “Hard to hear ‘don’t focus on the specific number,’ but I’ll take it to heart. We have another question that’s come in in terms of combining devices — so using a chest strap — we’ve talked about phones versus a watch. One question, I’m curious, how many devices you guys use? Do you use them in conjunction with each other? Should people, if they do want to sort of do a deep personal dive into their own health metrics, what would be devices that might be interesting to combine or not combine? Why don’t you start, Jessilyn?”

Jessilyn Dunn: “Sure, the chest strap versus watch question is interesting because the chest strap measures something different, right? It’s an ECG. When we talk about what is the actual sensor in the device measuring, it’s actually measuring electrical signals that correspond to heartbeats. It is a different measurement than what you’re getting from the wristband. The circumstances of use kind of matter in that case because a chest strap isn’t a very comfortable thing to wear. The battery life often is not what you would get from a watch. It wouldn’t be feasible probably to say I’m going to wear a chest strap 24/7 every day, but maybe I’m more interested in what’s going on with my heart rate, or whether or not I’m having any sort of heart-rhythm issues during exercise, in which case, it makes more sense to use that chest strap during certain specific times.

“In terms of multiple devices, I think it really depends what your goal is with the tracking. A lot of people will pair things like a smart scale with a wearable. You can see if you can pair movement with weight loss, for example, and then some people will even include in there an app that is tracking diet. You can kind of get this holistic view, but again, I would emphasize what Mitesh mentioned, which is not to focus on any one data point or any one day too specifically, but really what we’re looking for here is more of changes over time. If you have multiple devices measuring the same thing and you see that those changes are occurring on multiple devices, that gives you more confidence that this trend that you’re seeing is real.”

Rachel Ehrenberg: “Do you want to add anything, Mitesh?”

Mitesh Patel: “Yeah, I think that was well said. I think one thing that we see is this question around which device, which wearable device or how many devices, which is often a barrier to people tracking their daily behaviors, because there’s just too many choices that no one is clear standing out from the other. Then, it seems like a big commitment because they’re expensive and your hope is that you’ll use them for a long time.

“We often recommend that if you’re interested in tracking your health, just get started. If you have a smartphone, that’s the best way to get started. The apps on there are free. You can start tracking your step counts right away. Actually, most smartphones track your step counts already. You may not be aware of it. You can turn the functionality off, but they’re on there. You can actually take a look at that.

“Then, as you engage with that over time, you can see whether it would be beneficial for you to know, “Hey, how’s my heart rate doing when I go on a run?” or “I’d really like to see if my weight is changing over time as my step counts are increasing.” That gives people really an opportunity to just get started and get a sense of what they want. They’ll probably learn in the first few weeks of using their device, whether another device is worth it or not.”

Rachel Ehrenberg: “Follow-up for you, Mitesh, someone has written in that they’re participating in a study at Penn using a Fitbit, they’re finding avoiding the negative feedback is more motivating than getting positive feedback. They don’t want the study to end. What can I do to maintain my progress once the study is over? I think I’ll just add this issue of kind of longevity, long-term progress. How do you talk to patients and clinicians about this not being just for the three-month duration of a study or whatever it is, but kind of actually incorporating longer-term lifestyle changes that promote better health?”

Mitesh Patel: “There are a couple of things that make it more likely that people are going to be able to stick with these changes in behavior over long term. The first is really focusing on building a habit. Thinking about are there certain times of day that you’re going to go and get your steps in or do your activity? Can you build that into some sort of routine so that when this external negative messaging or whatever it might be goes away that you’ve kind of built that habit in.

“Another way to reinforce that is to really include a social incentive. Doing this with a family member or a friend, particularly someone who lives in the home with you, is likely to help keep the activity up and weight off. We actually did a randomized trial called Lose It, where we paired people together in kind of a collaborative intervention.

“And what we found is after the study was done, the people whose partner lived with them, kept the weight off. The people whose partner didn’t live with them but they were a family member, friend, gained about half of it back. Then, the people who didn’t really know the other partner, they were kind of, they didn’t know them as well, actually gained most of the weight back. It really shows you that the social element is key and that the strength of the connection between you and that person. You think of it as like a gym buddy who’s going to keep you accountable when you’re going to the gym or trying to get your steps in, is really important towards sustaining behavior change.”

Rachel Ehrenberg: “We have another question related to the psychological consequences of wearables. One high blood pressure reading in the doctor’s office can make people very anxious. There was a New York Times article earlier this week on people sort of similar to perhaps the overuse of screening devices. The better technology gets, the more data you can gather, but that isn’t necessarily good if it prompts people to get unnecessary, very expensive tests that also can be incredibly anxiety-inducing. Can one of you talk a little bit about psychologically how you think about it and how you would encourage other people to think about it in terms of using a wearable?”

Mitesh Patel: “Sure, I can start. I think you’re right, there are a lot of opportunities for the data to make people nervous or anxious about what’s going on, particularly when you get an outlier reading or something that’s like a very high blood pressure or you gained two pounds at the end of the week after you’ve worked a lot.

“Again, some things to keep in mind: One is to look at the longer term and realize that that one issue is part of kind of a general thing, but also to create, when you’re designing a program that uses these devices, to try to create some actionable opportunities for people to act on them. If your blood pressure’s high, to be able to learn about why that might be or what you can do about it, but I think people get most nervous when they have a reading that’s unusual or something unexpected and they don’t know what to do about it, and they start getting concerned about what it might be.

“The other is to recognize that sometimes these things are not always as accurate as we think. Even in doctors’ offices, a high blood pressure reading, making sure you take that three times. This particularly happens at home. We find patients who are using self-blood pressure cuffs and they get a really high reading, making sure you’re doing it the right way, taking it again, multiple readings over the course of the day. Then, if indeed, the blood pressure is consistently high, then talking to your doctor about what you might do about that.”

Rachel Ehrenberg: “Jessilyn, do you want to add anything?”

Jessilyn Dunn: “Sure, I guess, the one thing that I might add is that there could be this self-fulfilling prophecy, where if you have a high blood pressure reading, that makes you nervous and then that actually brings your blood pressure up. It is really important like we said, to take each single measurement with a grain of salt because there is a pretty good chance that one single measurement in time is incorrect.

“Again, having these multiple measurements over time and trying not to allow yourself to get too involved with one single measurement, so that it doesn’t actually change your physiology, is really important. I guess, we try to treat it like anything else that we might take with a grain of salt to say, ‘This may or may not be true. Let me check again in a few minutes or a few hours. If I’m consistently seeing a trend, then it’s something that I’m going to start to believe and try to look into doing something about.”

Rachel Ehrenberg: “Yeah, I don’t want to discourage people from taking what could be lifesaving action if there is something real going on. Can you do it, I don’t know if you would reset the app, or is it relatively straightforward to take another reading and get a sense of whether it was an anomaly or if there may actually be something that’s going on?”

Jessilyn Dunn: Sure. “One thing that I will add, because I agree with you, I wouldn’t want somebody to see, for example, an A-fib alert on their Apple Watch and then say, ‘Oh, it’s probably nothing,’ but the way that Apple A-fib detection algorithm actually works is that it does this for you, these multiple measurements. It itself takes multiple measurements and it looks at the first measurement. And if that measurement says, ‘This is likely to be A-fib,’ it doesn’t just send you an alert to say, ‘Hey, you’re experiencing A-fib,’ it actually then does exactly what we just said, of taking multiple measurements, and it gets us some confidence level around, ‘OK, am I seeing this again and again or was this maybe just an error in my first reading?’ If it gets, I think the number is something like five measurements in a row that all indicate A-fib, then it sends you the alert.

“It’s important for us to just think in that way of how bad is this event, if it happened? For example, let’s say you take your temperature with a thermometer and the temperature reads 103. If that’s true, that’s quite bad. You want to try to quickly resolve whether your thermometer is broken or whether that’s really your temperature. But for example, if it says that your heart rate is 10 beats per minute above average, the consequences of that might not be too bad. Maybe you can take some time to figure out what’s going on. I guess, the circumstances of how bad is this event if your device is telling you that it’s going on and what do you think you need to do to get more confident that it is truly happening?”

Rachel Ehrenberg: “There are a lot of questions about individual devices and measurements. I don’t expect you guys to go through, this brand does well at this, this doesn’t. Are there resources for consumers? Is there any kind of, Jessilyn, you talked about some of the organizations you work with?”

“Mitesh, I don’t know through the Nudge Unit, there are sort of resources where people can go and say like, “OK, this device gets high marks on this measurement, less so on this one,” because they’re not all necessarily using the same technology and. Jessilyn, you mentioned before, it may be part of an app or a measurement has been given the stamp of approval by the FDA, but other ones not. I don’t want to ask people to dive into the scientific literature, is there any kind of clearing house or either through your academic institutions or through some of these nonprofits or institutions that are working toward kind of accuracy and standards in this arena? How do people look up what’s good at what?”

Mitesh Patel: “I don’t have a lot of good resources to be honest. I think it’s actually really challenging. I will say that, in general, I think most devices are fairly accurate at tracking step counts because it’s kind of like the most basic element of what they track, but when you start to create estimations based off of movement, things like sleep, calorie counts, and so on and so forth, that’s where each company or devices using their own kind of black-box algorithm and it hasn’t been as well validated. I think it’s actually quite challenging and that’s really something that we need to figure out a better way to do, to help consumers figure out which devices are accurate and which are not.”

Jessilyn Dunn: “Yes, and I would say that this is a problem, I think, on a larger scale as well when we start to think about clinical trials and the use of these devices in evaluating whether or not, for example, a drug is doing what we think it should be doing. There are resources out there for companies that are running clinical trials to evaluate devices head to head.

“One resource that I can mention is Electrolabs that does sort of this, they call it, what is like ingredients list or some sort of a nutrition facts, but it’s for a device and it tells you each of the components of the device and if there’s literature out there evaluating them how well they do. There’s another company called Kenexa that also assists in clinical trials and matching companies with devices to use for their trials. I don’t know of something like that for the everyday consumer.

“I think, it is a bit tricky to figure out what device makes the most sense for you. Thinking about what it is that you’re interested in measuring and why and how invested are you in this process might be a good way to start. I think like Mitesh said, it often makes sense to just start out with a cheap device and just to get a feel for, is this something that’s useful to you and then kind of build up from there as you see the functionalities that are exciting to you or not.”

Rachel Ehrenberg: “We have a question about Neurable and EEG brain activity measurements. I don’t know if this falls in the realm of wearables on your head, going back to maybe that sweatband that I should patent or if this is something, are smartwatches trying to also collect neurology-related data? If so, can either of you speak to where that’s at?”

Mitesh Patel: “I don’t know as much about that area. I think it’s mostly things that you wear on your head that as opposed to wear on your wrist that kind of track brainwaves. I know that there’s a lot of interest in there because they could help to not only monitor mood and other behaviors, but then also help you think about this idea of like brain training and so on and so forth. How can you focus and so on? I don’t know the evidence base as much there. I think they’re just starting to scratch the surface. That’s an area potentially in the future that there could be a lot of work in.”

Rachel Ehrenberg: “We also have a question about… Oh, sorry. Go ahead, Jessilyn.”

Jessilyn Dunn: “That’s OK. One quick thing that I would add is that when we think about neural activity, so we have the brain and then we also have peripheral nerves, there is a sensor that one of the sensors that I mentioned previously, electrodermal activity that measures what we call autonomic nervous function. It’s actually more about the peripheral nervous system.

“There’s also some information about the nervous system in the heart rhythm because your autonomic nervous system actually regulates heart rate as well. When we’re talking about neurology and the nervous system, there’s kind of the separate components that we can measure but for the EEG, for actual brain activity. As far as I know, only head patches, I think, there’s a wearable that has two patches behind the ears or a cap or your traditional EEG are the devices that I know of.”

Rachel Ehrenberg: “I think we have time for one last question in terms of the real-life clinical outcomes of people using wearables. Are they actually helping people’s health, and can either of you give an example?”

Mitesh Patel: “I think that standard programs that just give someone a wearable off the shelf, that alone is often not enough to motivate people to change their behavior in a way that changes clinical outcomes. I think the more success has been found when these devices are paired with behavior-change programs. These could be things like the gamification and incentives that we’ve come up with.

“We’ve had a lot of successful efforts in increasing people’s activity, people after having a heart attack. We just had a study published a one-year study looking at people with uncontrolled diabetes, 20 percent boost in activity that sustained through the end of that. I think there are opportunities, but I think you really need to think about all of these different aspects from sustaining behavior, changing it, and then helping people build a habit.”

Rachel Ehrenberg: “Well, unfortunately, oh, do you want to add anything, Jessilyn, real quick?”

Jessilyn Dunn: “I guess, just quickly, I would add, I think detection has also been really helpful. One example is in the seizure realm where there’s a device called Empatica that can alert people who have epilepsy to seizures before they occur, so that they can kind of prepare for them.

“As new technologies come online like this, I think we’re seeing more and more benefit to patients in the real world, but again, it takes time to validate the technologies and ensure that they do what we think they should do. I think there are plenty of examples of devices that are working well in that space and also examples of devices that are maybe not quite there yet.”

Rachel Ehrenberg: “Well, that is, I’m afraid, all we have time for. I want to thank everyone for joining this event. I especially want to thank these speakers, Jessilyn and Mitesh, for this fascinating discussion today. If this has been a good experience for you, the viewer, please consider providing a donation to Knowable Magazine. We have a campaign going. You can do this at

“This conversation will be posted on the Knowable website, where it will be free to view and share. Look for the “Reset” collection. There will also be additional resources including links to articles and projects that the speakers are involved in.

“The best way to keep up with these discussions and everything Knowable is to sign up for our newsletter, which you can do on our website as well,, or to follow us on Twitter at Knowable Mag. That’s all from me today. Thanks again very much for joining us. Thank you, both.”

Mitesh Patel: “Thank you.”

Jessilyn Dunn: “Thank you.”

This article originally appeared in Knowable Magazine, an independent journalistic endeavor from Annual Reviews. Sign up for the newsletter.

Knowable Magazine | Annual Reviews

Related posts

Kids Already Coping With Mental Disorders Spiral as Pandemic Topples Vital Support Systems


Survey: Americans Want Weight Loss Drugs Despite High Cost


How to use AI for discovery without leading science astray

Choose a language