World Sleep 2022 | Consumer sleep wearable technology in clinical and research settings

Well, so consumer sleep technology is a broad field. It covers a big landscape from a sleep tracking device to sleep augmentation technology. From this sleep tracking, we know that there are upcoming new and new devices. You have smartphone, but also wearable classical wearable devices, smart band, ring, but also in-bed sensor. And then we see the boundary that is pushed further and further in term of tracking technology like EG system, dry EG system, headband, and so on. Even non-contact sensor that use radar and sonar technologies to measure sleep and several other sleep parameter. For the sleep augmentation side, also we see a novel introduction of devices that is able to tackle specific physiological processes, like neurostimulator devices.

We also have an upgrade of the classical white noise machine or aromatherapy. We see the robotic enter in the sleep field. I saw recently a pillow that you can hug with mechanical breathing which accompanies you to sleep by slowing down your breathing. You’re supposed to hug the pillow and then synchronize your breathing with the breathing of the pillow. And then we have several other gadget-based on the brainwave entrainment, so the space is really broad. We are interesting in sleep tracking device because mainly they offer parameter that we are used to see in our field. We are able to measure sleep and sleep aspect on a population level.

So, large data acquisition, time series data over prolonged period. And now we have the opportunity to go from a few hundred people in the lab every few couple of nights to millions of people, over a million nights – billions of data. And also this technology, especially the multi-sensor wearable sleep tracker, opens the possibility to really advance the understanding of sleep and how sleep is related to several bio-psychosocial factors. Think about that now you have app interface, you can have digital survey paired with the passive collection of this multidimensionality of data. They are able to measure sleep, but also beyond sleep, they are able to measure sleep physiology.

We know that with the photoplethysmo, they now able to measure heart rate, but also heart rate variability, and they expand more and more the feature set of the type of parameter that they can measure including now oxygen saturation. And so, look at the desaturation across the night and potentially classify apnea events. Also, they can expand even further and now we will probably see upcoming news of technology in glucose monitoring, blood pressure monitoring and alcohol sensors. So, you can really have the curve of alcohol consumption and really have high granularity data.

While there are big limitations of consumer sleep tracking technology, and some of the limitations are still unknown – you need to conceptualize that there are consumer devices, and the algorithm that they use for measuring sleep and the other physiological parameters are still unknown. So, they don’t disclose to the user wearable industry what and how they measure what they measure. And that is a limitation. Also, another limitation is that a proxy of sleep is not exactly the translation of sleep. They basically use multi-sensor technology, accelerometry, and at least photoplethysmography and features derived from photoplethysmography to guess in which sleep stage you are, to classify the sleep stage. But they model this data according to polysomnographic data. But in the field, in free living condition, there are so many factors that can modify and affect peripheral physiology and specific heart rate, heart rate variability, that automatically translate in different level of accuracy of this device to measure sleep.

An example would be drinking alcohol. If you get a couple of glasses one night, the device might have different accuracy. For an example, you have a night in which you don’t drink, or if you get caffeine, if you do physical activity, any kind of factor that modifies the feature set that this device uses to classify sleep, might change the accuracy of this device.

So today, you might have 5% misalignment, but the day after 10%, 15%, 50%, you don’t know. And that is unpredictable. There are so many factors that now modulate the way that they classify sleep. And that is a problem. Other problems are that they are commercial devices. So, on the long run, they are following different rule. They follow the market rule. So, they push new devices every time, new device model. The use case is changing, so they follow use case and rule for engagement, engaging the people, not accuracy in the way that they measure a parameter. And then they don’t guarantee that a device still produces the same amount of data every time. They might abandon some features, they might introduce new feature without even noticing. So, it’s a real challenge for us to use these devices in a longitudinal way.