05mpul05-grifantini

How’s My Sleep?

When it comes to health and fitness, there’s an app for just about everything. Want to track how many steps you’ve taken today? There’s an app for that. Want to track the calories you’ve consumed? There’s an app for that, too. And, if you want to know if you got a good night’s sleep last night, well, now there’s also an app for that.

The U.S. Centers for Disease Control and Prevention call sleep deprivation a national public epidemic. The National Institutes for Health report that the average American adult gets less than the recommended 7–8 hours a night, while an average of 60 million Americans have chronic sleep disorders. A lack of sleep is linked to increased accidents, risk of colds and infections, obesity, heart problems, and impaired learning according to the NIH National Heart, Blood and Lung Institute. The National Highway Traffic Safety Administration attributes more than 100,000 car accidents a year to drowsy driving. And most people can confirm that a night of sleeplessness can lead to lower concentration and energy as well as increased irritability and fatigue.

In the hopes of combating these problems, companies are developing sleep-tracking apps and biometric devices aimed at providing consumers with valuable information about their sleeping habits. Aside from simple smartphone apps that purport to provide white noise or soothing sounds for better sleeping, a number of companies are developing wearable technologies, primarily based on movement-detecting accelerometers, microphones, and heart rate monitors, that seek to be a mini, portable sleep lab. These include apps that attempt to provide snore analyzers and “smart” alarms; and fitness bands and specialized bed devices that try to chart sleep stages.

“The idea is to give people a way to aggregate their three most important lifestyle data sources—sleep, activity, and nutrition—in one place to analyze and improve lifestyles and, thus, improve health and indirectly increase life expectancy,” says Petr Nálevka, Ph.D., a sleep app developer.

The problem, according to sleep specialists, is that no consumer technology on the market can currently match the level of sleep stage recording in a sleep lab.

“Sleep is at the connection of everything—disrupted sleep can be an early warning signal of serious health problems and a key problem that needs to be taken care of,” says Gari Clifford, associate professor of biomedical informatics and biomedical engineering at Emory University School of Medicine and the Georgia Institute of Technology, as well as leader of a team at the Sleep and Circadian Neuroscience Institute at Oxford University. “But just downloading any current app and expecting it to tell you your sleep status is a little fanciful. Right now we don’t have enough contextual information in current wearable consumer technologies to achieve accurate sleep information.”

While the increase of sleep tracking technologies have increased public awareness of the importance of sleep, specialists worry that technologies serving up sleep analytics or diagnoses of sleep disorders based on oversimplified metrics could do more harm than good by putting a user at ease when he or she may have otherwise sought out a sleep specialist. This could be a particularly worrisome lapse, as an onset of a sleep problem can be a red flag for a more serious mental illness, such as schizophrenia or dementia. However, a few companies are steadily making improvements in apps, wristbands, and more unique tracking items to push the field of personal sleep tracking forward.

Easy Answers Through Motion Tracking

Of the hundreds of sleep apps available for smartphones—offering everything from white noise sounds, sleep hypnosis, and dream logging—a handful offer sleep tracking and analyses. These apps (as well as the majority of sleep-tracking wristbands) typically rely on a simple tool to measure movement, the accelerometer. These devices use the accelerometer to estimate your physical movement and serve up guesses as to your overall health.

All of the sleep apps based on accelerometer readings use a simple correlation to plot restfulness: if you move more, you’re in a lighter or more fitful sleep, and, if you move less, you’re in a deeper sleep. This is based on the premise that, while in deep dreaming sleep, the brain typically inhibits physical movement so that a dreamer can’t act out their mental projections. By using this simplistic approach, many companies claim that they can analyze sleep patterns; some go further to offer a “smart alarm” that wakes users up at an estimated optimal time during the sleep cycle to reduce grogginess.

But technologies that solely rely on movement to determine sleep restfulness have an inherent problem: sleeping patterns are too complex to be represented by movement.

“Whether you’re moving or not is not going to tell you what stage of sleep you’re in,” explains Hawley Montgomery-Downs, Ph.D., associate professor of psychology, and director of the sleep research laboratory at West Virginia University, who has done analyses of market devices and found them lacking in a 2011 article published in Sleep Breath. “There’s really no evidence that currently available consumer sleep devices can differentiate between deeper and lighter sleeps. It is critically important that they be exposed to objective testing to determine whether the marketing claims are accurate.”

grifantini01-2339252

FIGURE 1 Sleep Cycle uses the accelerometer in smartphones to detect movement. The app uses these readings to attempt to graph a user’s sleep patterns. The screenshots show (a) the recommended placement of the smartphone, (b) the user’s sleep statistics, and (c) sleep quality graphs. (Image courtesy of Sleep Cycle.)

For example, someone who has insomnia or is waking up for short periods (due to sleep apnea or periodic movement disorder) may not move and, therefore, won’t be detected by a movement tracker. “People might not feel like they’re sleeping their best, but the technology tells them their sleep is fine,” says Montgomery-Downs. “None of these companies can currently replace the information that a physician can give you by doing an evaluation.”

grifantini02-2339252

FIGURE 2 The Basis wristband. (Image courtesy of Basis.)

Some apps have used continuous audio recording in addition to accelerometer readings to put together a picture of a user’s sleep. Sleep as Android, first released in 2010 and with 8 million users to date, according to the developer, offers audio recording to capture snoring, sleep talk, and potential apnea issues.

“Deep sleep tracking is only a small piece of the whole sleep tracking concept, together with sleep length and sleep deficit tracking, snoring detection, and lifestyle event tagging,” explains Nálevka, who is also the cofounder of the Prague-based Urbandroi team. The ultimate goal, according to Nálevka, is to give people a way to aggregate sleep, activity, and nutrition to help them improve overall health.

“The sleep tracking definitely isn’t error free, but this is the maximum we can now practically do in home conditions and definitely something which is worth doing,” says Nálevka, adding that the team is beginning trials to compare the app’s algorithms to polysomnography (PSG), with data from about 20 patients from a local sleep lab in Prague so far.

Sleep as Android also attempts to integrate information about lifestyle habits, such as physical activity or caffeine consumption, and relate it to sleep measures. So far, the app has integrated into Philips’ mood-changing Hue lightbulbs and the Pebble smart watch, and plans to extend the snoring detection to become a prescreen for sleep apnea.

Another popular trend among smartphone sleep tracking apps is to provide a “smart alarm,” which claims to wake up a person at an optimal, lighter-sleep period in their sleep cycle to reduce grogginess. The Sleep Cycle alarm clock, which times users’ sleep cycles to guess the best time to wake them, is one of the top ten most downloaded apps in the Apple Store’s health and fitness category. Users lay their phones on the mattress as they sleep and the app (developed by Swedish-based Maciek Drejak Labs) detects movement to wake them up within a set window of time (Figure 1). MotionX 24/7 and Sleepbot, a similar app for Android phones, purport to offer a similar service.

“The app is using an advanced algorithm to differ movements, for example, if the movements are close or far away, how strong they are, and the like,” says Sleep Cycle company representative Alexander Hallberg, adding that the iPhone is limited as a platform and that the app can’t compensate for a lack of sleep or disturbances. “We can’t accurately predict the exact optimal waking time without measuring the person’s brain waves,” he admits.

“It’s well known that tracking by accelerometers is a poor indicator of abnormal sleep patterns,” says Clifford. “All of the apps out there that rely on physical movement from the accelerometer on the phone have no verifiable scientific foundation whatsoever.” He also expresses doubt around apps that ask a user to scroll through audio data to analyze their snores and look for anomalies. “It’s already difficult for a trained expert to listen to audio and determine if you have sleep apnea or not.”

Clifford’s research group, who last year published an article in Physiological Measurements [1] reviewing sleep apps, has developed its own evidence-based algorithm that screens for sleep apnea, dubbed SleepAp. As for the future of sleep tracking, Clifford suggests that continuous visual recording of sleepers may provide a convenient way to detect sleep disruptions more accurately than the current mass-market devices (see [1]). One of his researchers achieved a 90% accuracy using video analysis, he says, but, for the work to have broad consumer appeal, users will need to be comfortable with having a camera trained on their beds.

Despite the limitations of phone-based and other low-cost sleep trackers, there is some benefit to the growing popularity of these apps, according to Montgomery-Downs. “We’ve been trying for years to convince people that sleep is a core factor for health, along with diet and exercise,” she says. She adds that the message is getting out there now, but the use of sleep tracking technologies needs to be tempered with an awareness of their limitations, and consumers should look for empirical studies supporting products’ claims. But most of the apps keep their sleep algorithms private and proprietary, so consumers don’t have easy ways to evaluate a product’s accuracy.

Specialized Tracking Through Wristbands

In addition to smartphone apps, specialized fitness wristbands, such as products from Fitbit, Jawbone, Garmin, Philips, Nike, Misfit, and others, similarly track users’ movements to call up the number of hours slept and the duration of restful versus fitful sleep. These devices are advantageous in that they are unobtrusive, making it more likely for users to regularly wear and activate them for consistent sleep tracking. However, those that solely use accelerometers will run into the same questions of accuracy as the smartphone apps.

FIGURE 3 The Basis tracker monitors sleep in addition to fitness and other health parameters with a heart rate monitor and accelerometers. (Image courtesy of Basis.)

FIGURE 3 The Basis tracker monitors sleep in addition to fitness and other health parameters with a heart rate monitor and accelerometers. (Image courtesy of Basis.)

“There’s a huge interest in tracking sleep because, when sleep breaks down, everything breaks down,” asserts Christopher Winter, M.D., medical director of Martha Jefferson Hospital’s Sleep Medicine Center in Charlottesville, Virginia, who works with athletes in the National Basketball Association and Major League Baseball to improve their sleep. “But if you just take movement and try to recreate a sleep stage diagram, it’s going to be very off.”

Winter carried out his own informal sleep experiment, strapping several of the popular trackers to his wrist to record his sleep and compared them against an independent sleep technician’s PSG [2]. None of the devices he tried were as accurate; however, he found that one wristband that monitored heart rate in addition to movement exceeded the others. “The Basis wristband was better not only in terms of what it did but also in how it did it,” Winter says, citing its streamlined user interface (Figure 2). Winter has, since his analysis of consumer sleep devices, become a spokesperson for Basis ­(Figure 3), which was acquired by Intel in March 2014.

FIGURE 4 The Jawbone wristband aims to integrate sleep measurements, fitness, and activities such as caffeine intake to present a picture of the user’s health. (Image courtesy of Jawbone.)

FIGURE 4 The Jawbone wristband aims to integrate sleep measurements, fitness, and activities such as caffeine intake to present a picture of the user’s health. (Image courtesy of Jawbone.)

The San Francisco-based company worked with independent researchers at the University of California, San Francisco, the San Francisco Veterans Affair Medical Center, and the Northern California Institute of Research and Education to publish a recent study [3], showing that, in a small sample of 12 patients tracked for one to two nights, the Basis band was fairly close to detecting sleep cycles compared to PSG readings, says Tejash Unadkat, director of product management at Basis Science.

“We use heart rate and the accelerometer in combination to get what is pretty close to medical-level sleep staging in normal subjects, based on preliminary results,” says Unadkat. “At a very high level, heart rate varies at different sleep stages and we found it is a very legitimate proxy for sleep staging.” He adds that the company may look toward getting U.S. Food and Drug Administration approval in the future as a potential additional tool for sleep labs, says Unadkat.

If adding a heart rate monitor as a metric makes such a difference in accuracy, why don’t more companies incorporate heart rate detection into their sleep tracking technology?

Unadkat says that integrating a heart rate monitor into a ­simple-to-use, unobtrusive form factor like a wristband isn’t easy. Heart rate monitors are typically worn around the chest, which can be uncomfortable to wear at night and a barrier to everyday use.

More and more fitness watches have integrated heart rate monitoring, such as the Mio Alpha and Scosche bands. Just this year, a host of tech giants have started to enter the consumer space: Samsung offers a curved heart-tracking Gearfit wrist display; LG’s G watch and Motorola’s Moto 360 will feature Google’s Android Wear system with sleep tracking; and Apple will likely launch a fitness watch this year rumored to offer sleep tracking functionality.

Beyond Sleep Tracking

The company that produced Zeo, an unusual consumer sleep tracking device consisting of a headband that tracked brain activity to accurately record sleep cycles, shut down in 2013. Its chief executive officer cited a lack of a strong business model, not a technology barrier, for the closure. Although wearing a headband was likely a barrier to consumer adaptation, the shutdown of what may have been one of the most accurate personal sleep trackers on the market points to another challenge for these technologies: once they provide users with precise sleep data, what are the users supposed to do with it?

Similarly, specialized beds and mattress pads that track heart rate, movement, and other metrics without requiring the user to wear anything—such as the Sleep Number x12 bed powered by BamLabs, the Withings Aura system, or the Bluetooth-enabled Beddit bed sensor—must also answer the same question.

“Sleep trackers need to have a practical message for its users. Ones that try to give useful feedback by claiming to wake you up during ‘light’ sleep and track your sleep stages throughout the night are based on a flawed premise,” asserts Jordan Gaines, a sleep expert and Ph.D. candidate who has written extensively on the topic for Nature Education, Psychology Today, NBC News Health, Scientific American, The Washington Post, and others. “Giving the average person a graph that estimates their sleep cycles or counting the number of times they move just isn’t necessary.”

Gaines, whose research has shown that mental health problems are associated with persistent insomnia, points to immersive technology as beginning to provide a practical context. The Jawbone’s UP [24] platform (Figure 4), which states that it can learn a user’s activity over a few weeks, detects trends in daytime activities (such as caffeine consumption) and how one sleeps and makes recommendations, is one such system. Although it may not yet be as accurate as PSG in terms of sleep cycle tracking, Gaines claims that the bridging of information is a step in the right direction to help users contemplate and improve their sleep quality.

Despite his skepticism of the current commercial devices, Clifford is optimistic about the future of sleep-tracking technologies. “If tracking technologies are used in the correct context, with enough support—either through trained humans, or good decision support algorithms which provide useful feedback to alter behavior—then they have great potential,” he says.

For personal sleep-tracking devices to be truly helpful to consumers, they will need to not only meet the gold standard of PSG testing but also customize to individuals’ sleeping habits through smarter algorithms, provide contextual information relating sleep to other health parameters, and know when to suggest users contact a sleep technician in case of a sleep disorder. As personal sleep trackers grow more sophisticated in both hardware and software, they may ultimately help users understand—and improve—their sleep and overall health.

References

  1. J. Behar, A. Roebuck, J. S. Domingos, E. Gederi and G. D. Clifford. (2013). A review of current sleep screening applications for smartphones,” Physiol Meas., vol. 34, no. 7.
  2. C. Winter. (2014, Feb. 26). Personal sleep monitors: Do they work? Huffington Post.
  3. S. Patel, T. Ahmed, J. Lee, L. Ruoff, and T. Unadkat. Validation of Basis Science advanced sleep analysis, Basis Science, San Francisco, CA.

  • Colin Hay

    As a chronic insomniac I’ve been using a fitbit for the past 5 months to track my sleep as I make lifestyle changes, work on sleep hygiene and get used to using a cpap machine. The motion sensing isn’t very accurate from a numerical standpoint but there seems to be a benefit comparing data from month to month. For example in June before I had many changes, I had nights that were registering very low sleep and I did feel awful after those nights (under 2 hours). The past few months, I haven’t had a night under 2 hours. Still not where I want it to be but the changes have made improvements both in what the fitbit says and how I feel in general.

    For awhile I was concerned about finding a “better” device but I’ve realized that it doesn’t exist right now. And as discussed in the article, even if the device was more accurate, what exactly would the data be telling me to do to fix it?

    I’ve found that using the devices for motivation is almost more important than the exact numbers that it provides. I also dump the data to a spreadsheet and add in caffeine consumption, exercise, etc to try to make some longer term correlations.