Wear OS Weekly
My weekly column focuses on the state of Wear OS, from new developments and updates to the latest apps and features we want to highlight.
Smartwatches have tracked sleep for years, but as we see more sleep coaching, energy scores, sleep apnea detection, and general AI analysis, it’s fair to question how accurate a Pixel Watch’s data is compared to, say, a polysomnogram or actigraphy watch.
So I sat down with Dr. Logan Schneider — an associate professor at Stanford University and Google’s clinical lead for sleep health — and had a frank discussion about the usefulness, limits, and future of smartwatch sleep tracking.
Google frequently enlists practicing doctors to work with its engineers on health and wellness features. Earlier this month, I previously interviewed Google’s clinical lead of heart health, Dr. Shreibati, about Loss of Pulse and how smartwatches will eventually deliver “personalized” diagnoses for heart conditions.
In this case, Dr. Schneider, who worked on projects like the animal-themed sleep profiles on Fitbits and Soli-powered sleep tracking on the Nest Hub, had plenty of insights about how clinically useful these kinds of consumer-tech devices can be compared to medical-grade tech.
The science behind smartwatch sleep tracking
When it comes to studying sleep patterns on an individual or population level, polysomnograms are “impractically expensive, time-consuming, and resource intensive,” says Dr. Schneider.
A smartwatch is physically incapable of tracking brain waves (EEGs), eye movement, and some full-body motions that a clinical sleep study can detect. However, it does capture enough useful data to extrapolate the essentials without getting health insurance involved.
That’s why Schneider saw the advent of smartwatches as an opportunity to “understand sleep in a novel way, looking at population health, looking at real world changes in sleep over time,” and began reaching out to tech companies like Google. Smartwatches fill a niche that polysomnograms can’t.
You can only get so good until you’re literally starting to look at the brain waves.
I asked Dr. Schneider to give a clinical run-down of how sleep tracking works. He explained how your “body’s physiology and its state of motion versus immobility recapitulates what’s going on in the brain” and how they can determine your “rest and digest” or “fight or flight” states as a reflection of your brain states.
Heart rate is an obvious tool for sleep stage approximation, but he put even more emphasis on movement tracking as a vital tool. The inactivity of deep sleep, twitchy paralysis of REM, and more frequent movement in light sleep can all be estimated by your watch’s accelerometer.
False-positive movements — like when your partner shifts the bed while you’re in deep sleep — can be one of the big factors when your watch gets sleep estimates wrong. That’s why Schneider remains proud of their work taking the Nest Hub’s Soli radar and turning it into a “full body actigraph and cardiopulmonary physiology sensor that could stage sleep quite accurately” because it gets a wider view of your body.
Ultimately, Schneider believes, smartwatches “can only get so good until you’re literally starting to look at the brain waves.”
The challenges of making wrist-based sleep data more accurate
I asked Dr. Schneider what, if anything, could make smartwatch sleep data better without EEG tech. He argued there probably isn’t some “gold signal” that makes things perfect, but that isn’t strictly necessary.
The main key for accuracy, he explained, is having “redundant signals.” Neurologists looking for a brain issue don’t go straight for the MRI or scalpel; they might test your muscular reflexes to catch a “dysfunction” first. Since there are “several different types of redundant signals of the autonomic nervous system,” smartwatch brands can look at those secondary sources and get more reliable results.
We already see that with factors like heart rate variability (HRV), breathing rate, and skin temperature, as secondary information for whether your body is “normal” or if something is trending wrong.
There’s some general event horizon where you’re never going to get on the wrist a measure of what’s going on in the brain.
One potentially new smartwatch metric Schneider brought up was “galvanic skin response.” Essentially, if you look at your skin’s sweat creation and salt content, you can determine your body’s sympathetic nervous system (SNS) response and see, for instance, when your sleep state is unstable.
The Pixel Watch 3 and Fitbit Sense 2 have a cEDA sensor that looks at sweat levels on your skin, primarily for stress detection. That seems like a perfect fit for using sweat levels to improve sleep accuracy, but I didn’t think to ask about this during the interview. Maybe we’ll see this in the future!
But ultimately, it comes back to Google’s PPG. He credited Google’s engineers for “optimizing the signal acquisition and quality,” as well as the multi-path optical sensor that ensures reliable, continuous data that approaches what you’d get from a “gold standard polysomnogram.”
For what it’s worth, testers like the Qualified Scientist say the Pixel Watch 3 is highly reliable for sleep accuracy. But Dr. Schneider still says that “there’s some general event horizon where you’re never going to get on the wrist a measure of what’s going on in the brain.”
Smartwatch sleep tracking has untapped clinical potential
I asked Dr. Schneider about the recent trend of smartwatches and smart rings starting to detect sleep apnea and if there are other potential sleep conditions (aka parasomnias) that he believes smartwatches could detect.
He very enthusiastically responded that “it could be a great opportunity for tech like this to actually pick up on” conditions like Restless Leg Syndrome and sleepwalking that can be difficult to catch in limited clinical contexts but that a smartwatch will detect with enough patience.
Smartwatches will need to be able to differentiate between REM behavior disorders (e.g., sleep paralysis), non-REM parasomnias ranging from sleepwalking to sleep talking and even sleep eating disorders, and seizure disorders.
Dr. Schneider seems very optimistic that watches could “pick up on interesting signals” in the future, “depending upon what set of sensors you have.”
This would have a direct clinical application, too. After someone is treated for a sleep disorder, it’s hard to determine if you’re “truly” cured or just have “fewer episodes that you’re aware of.” A smartwatch could passively solve that mystery and give patients (and doctors) peace of mind.
Why sleep tracking matters to you and me
I have a complicated relationship with sleep tracking. I wrote last year about how I hated wearing a smartwatch to bed, and I’ve found that seeing bad sleep data can often compound my stress. So I asked Dr. Schneider why people can’t trust their own bodies to feel how tired they are.
He responded with an anecdote about a man who said he was sleeping great while not believing his wife, who said he was “dying” at night and falling asleep at the wheel.
“People are terrible estimators. And why is that? Our brain habituates to our current state of being,” he explained. You can function with a “severe amount of sleep impairment that you’re not even fully cognizant of,” and sometimes you need help recognizing that.
He believes it’s worth the discomfort of a smartwatch to have this self-knowledge. And since he has to “convince people to wear a Darth Vader mask every single night” for sleep apnea, he thinks it’s worth getting accustomed to some form factor of wearable tech to help you “figure out how you’re feeling.”
But if, like me, a smartwatch isn’t the right fit for you, then Dr. Schneider thinks you really should consider that Nest Hub radar. To which I’d say, if Google hurries up and makes a 3rd-gen Nest Hub with Gemini, I’ll consider it!