Sleep Wearables Generate Stable Estimates

Data about sleep patterns derived from wearables correlates with data from self reports, according to a new study from the Sleep Research Society. However, there is not a clear connection.

Consumer-grade wearables show promise in generating useful sleep quality data, but a lack of standards around polysomnography limits the accuracy and usefulness of this data. To analyze the quality of sleep data in one particular device, the researchers compared sleep data from a Withings brand wearable to standard sleep self reports.

The researchers used data from a previous observational study that focused on sleep and nocturnal urination. The study was conducted in a multi-national population of people who purchased a Withings watch. Participants (6230 in total) were at least 18 years old (mean age = 47.4 [SD = 13.9] years; 59.3% male).

The researchers analyzed questions about habitual sleep duration and total number of awakenings. Participants used the watch for a mean length of 214 nights and completed a questionnaire.

Mean sleep duration from the wearables is almost 40 minutes higher than indicated in the self reports and the median is nearly 25 minutes higher than indicated in the self-reports. The mean number of awakenings was comparable between the two.

The length of use of the wearables didn’t influence habitual sleep duration correlation significantly. They became more correlated as the duration of use grew longer (r = .453 for 5-50 nights; r = .588 for 51-100 nights; r = .597 for 101-150 nights; r = .624 for 151-200 nights; r = .612 for 201-250 nights; r = .654 for 251-300 nights; r = .635 for 301-367 nights).

Overall the researchers confirmed that wearables tend to record longer sleep duration. They also found a large correlation between self-reported typical sleep duration and sleep duration as assessed with the wearable watch.

While the researchers found that wearing the watch for longer periods of time (6 months or longer) increased data accuracy, they also hypothesized that the wearables data may influence the perceptions of the users, causing them to edit their reports to more closely match what the watch says.

Data was limited to that of a single company and its hardware and software. The company did not provide source code. Also, “scant” evidence suggests sleep data from Withings meets the gold standard of PSG-based validation. The researchers also assume strong socioeconomic and demographic biases. Finally, the response rate was about 5%.

A number of the study authors reported industry ties. Please see the original document for full disclosure.

Reference

Bliwise DL, Chapple C, Maislisch L, Roitmann E, Burtea T. A multitrait, multimethod matrix approach for a consumer-grade wrist-worn watch measuring sleep duration and continuity. Sleep, zsaa141.