The ubiquity of smartphones and wearables makes it easy to count steps, spends, and snoozes in the name of optimizing health and happiness. But how reliable is consumer sleep tracking?
The challenge for mainstream sleep trackers is to achieve practical insights into sleep quality without being able to measure sleep directly. A direct scientific assessment of sleep is known as polysomnography, and it measures several important variables in order to profile sleep accurately. These often include recordings of eye movements, muscle activity, breathing, and brain activity. This comprehensive analysis, which is usually done in a sleep lab, requires multiple devices and can be cumbersome and expensive.
Consumer-focused sleep trackers need to take shortcuts. This means that most of them rely on correlations between particular body movements and sleep.
Generally speaking, the body acts differently during different phases of sleep: For example, REM sleep involves less muscle activity and movement than non-REM sleep, and people naturally move more when they are awake than when they are asleep. Therefore, devices that measure body movements, either using a microphone to record the sound of movement or an accelerometer to detect changes in device position, could potentially infer whether a person is sleeping or not.
But how strong is the relationship between body movement and sleep quality? Many people who struggle to fall asleep actually remain relatively still while lying in bed, so how would a device know that they are awake? The simple answer is that the device can’t know for sure, and there will inevitably be errors in sleep data. Any healthy skeptic will want to know how serious those errors are; are they significant enough to warrant throwing out the sleep tracker altogether?
The first and most obvious test is to compare popular sleep tracking apps against scientifically validated methods for monitoring sleep. In a 2015 study on 65 adolescents, researchers put the data from a wrist-worn sleep tracker side-by-side with the data from polysomnography. The polysomnography used brain activity readings, muscle recordings, and eye movements, while the commercially available sleep tracker used accelerometer (movement) data.
The only sleep measure that was relatively equal across the two methods was how long it took people to fall asleep after going to bed. Compared to polysomnography, the consumer sleep tracker overestimated total time asleep and underestimated total time awake. However, the sleep tracker was only missing the mark by approximately 10 minutes on average.
The readings were less coherent when it came to identifying different stages of sleep. While the sleep tracker defined “light sleep” and “sound sleep”, these categories didn’t show a consistent relationship with the polysomnography data. For example, in contrast to what you would expect, light sleep defined by the sleep tracker correlated with deep rather than light sleep according to polysomnography.
A review paper in 2018 combined the results from eleven different experiments on sleep tracking apps in order to reach a consensus about their efficacy. The conclusion was consistent with the study above: most smartphone applications did a good job of telling apart wake from sleep during the night, with a tendency to overestimate sleep. But inferences about different sleep stages — distinguishing light sleep from deep sleep — performed poorly and were not validated by the results of polysomnography.
The only product that did a somewhat better job used additional data from an external sensor that recorded heart and breathing rate in addition to the typical body movement data from an accelerometer. It’s not yet clear how much heart rate and breathing rate readings can improve the detection of sleep stages for consumer products. But some evidence does highlight a practical role for heart rate data, which is encouraging news for wearable sleep trackers that include these recordings.
addition to sleep trackers’ imperfect science, it’s also hard to gauge a particular model’s reliability. Businesses tend to avoid sharing proprietary material like sensor accuracy or data quality. So even when the overall evidence suggests that accelerometer and heart rate technology is feasible for basic sleep tracking, it’s possible that individual products don’t meet minimum requirements to reliably collect the data. Also important is that the evidence itself is limited: Most existing studies that compare sleep trackers to polysomnography use volunteers with healthy rather than unhealthy sleeping patterns.
Sleep trackers are likely to work more effectively for some people than others. Unfortunately, they may be least effective for people who have the worst sleep quality — exactly the kind of people who most need the technology. When devices fail to reliably detect wake times, the users who have the most frequent sleep disruptions will naturally get the least accurate readings because they wake up most often.
Defective insights can inspire defective correction strategies. A statement released by the American Academy of Sleep Medicine in 2018 argued that consumer sleep tech cannot yet be used in the diagnosis or treatment of sleep problems. For now, it should be limited to providing additional data points during a more comprehensive evaluation with a health professional.
Unfortunately, they may be least effective for people who have the worst sleep quality — exactly the kind of people who most need the technology
In June 2019, leading sleep researchers around the world came together at the Annual Meeting of the Associated Professional Sleep Societies to preliminarily showcase some of their emerging findings in sleep tech. Some of the work replicates previously established facts, including the oft-cited concern that movement-based sleep trackers struggle with detecting wake times. But a few projects also tackled the question of boosting sleep quality. Researchers who used cognitive behavioral therapy to treat insomnia found that they could utilize remote sleep tracking with an app in order to reduce the burden of in-person visits when analyzing progress. In another small pilot study, sleep-related coaching in combination with a wearable tracker was helpful in reducing sleep disturbances.
So sleep tracking apps could be a practical tool within a broader therapeutic strategy. And while it’s important to wait until the data is fully published in peer-reviewed journals before putting too much faith in the results, it’s encouraging to know that experiments are now studying the relevance of consumer tech in sleep problems.
ooking to the future, the potential of consumer sleep trackers is clear. Their widespread and regular usage means that researchers can easily access real-life datasets on sleep quality from large populations, which can help clinicians to better understand why so many people struggle with sleep and what types of adjustments achieve the strongest improvements.
For current at-home users, sleep trackers may improve awareness of bad sleeping habits, which in turn can motivate people to improve their sleep hygiene. Sleep is clearly important for human health, so products that encourage attention toward sleep quality are likely to do some good.
But for those who already take their sleeping habits seriously, either by adhering to their own disciplined bedtime schedule or by using alternative methods such as sleep diaries, it’s not yet clear whether consumer sleep trackers have anything important to add.
The technology will no doubt improve, and the science may eventually offer a confident thumbs up for the more revolutionary elements of popular sleep trackers. In the meantime, users who hope to improve their sleep quality should be cautious in how they interpret the numbers, and everyone should be wary of what their overnight digital companions say about light versus deep sleep.
All Rights Reserved for Erman Misirlisoy