Is My Sleep Tracker Tracking my Sleep?

The purpose of this research study is to collect health and physiological data using commercially available wristband fitness tracker devices (FitBit and Garmin devices) to help determine their accuracy and reliability at measuring percent of night spent in REM sleep, oxygen desaturation, and apnea hypopnea index compared with currently available methods of in-laboratory polysomnogram and home sleep testing.

To date, there is an incomplete picture of the reliability of wearable device trackers to depict sleep quantity and sleep staging information. Prior studies have compared various iterations of wearable sleep trackers to the so-called gold standard of PSG, often but not universally in health populations. A number of important observations have been made to date. 1. Wearable device trackers overstate total sleep time (TST) and understate wake after sleep onset (WASO), thereby overestimating sleep efficiency. When comparing a dichotomous of sleep/wake categorization, wearable sleep tracker overestimates of sleep time result in high sensitivity for categorizing a given sleep period (a 30 second or one minute "epoch") as "sleep," but as a consequence, there is an attendant drop in specificity, as true "wake" on PSG is more frequently mislabeled as "sleep" on the wearable device. 2. Second, wearable sleep trackers, due to technical limitations of inability to correctly categorizing N1 versus N2 sleep, collapse those stages into a combined category of "Light Sleep." 3. Third, the raw data, with heart rate data and heart rate variability data which feed into the proprietary wearable device algorithm to assign sleep stage, are not directly available to researchers. Moreover, the wearable device derived data on sleep staging extracted from the device are often provided in one-minute windows (not the 30-second epoch or window used in PSG scoring). Therefore, the so-called "epoch by epoch" comparisons of exported data from the wearable device, compared to the PSG gold standard, have inherent limitations. 4. Nonetheless, even with those limitations acknowledged, important correlations between wearable device-derived sleep time, light sleep and REM staging have been established, using so-called epoch by epoch analysis, which however have varied according to the device chosen and population studied. For the current study, the investigators do not plan to examine an epoch by epoch assessment of sleep staging as a primary analysis, in part due to its inherent limitations consisting of: (a) lack of raw data from device; (b) difficulty matching up epochs due to differences in timing of the so-called "window" of time observed (30 seconds versus one minute); (c) differences in sleep time recording, thus resulting in different denominators of sleep time; (d) poor test-retest or interrater variability for PSG scoring itself, even among expert academic centers performing epoch by epoch analyses of the very same PSG. Instead, the investigators plan to focus on a more clinically accessible and, for the consumer, more relevant question: how well does the amount (or the percentage) of REM sleep and total sleep time estimated by the wearable sleep tracker correlate with a simultaneous sleep study? Secondary analyses will also assess sleep/wake and additional sleep stage comparisons, and assessments of respiratory parameters of oxygen desaturation, and a comparison of wrist tracker device and PSG sleep compared to Level 3 home sleep test derived recording time, in a population of subjects being evaluated for sleep apnea and other sleep disorders. Summary assessments of the sleep variables for the night will be compared to assess the accuracy of the wearable devices and Level 3 home sleep test to polysomnogram. Through the study, the investigators hope to contribute to building a body of evidence assessing the level of accuracy of the latest generation of consumer wearable sleep tracking devices. The investigators plan to use two devices, the FBI3 and the GVS5 fitness activity trackers, for the study, as these devices are among the most recent versions available, are widely used, are highly affordable (models under $150), and provide ease of measurement (as no continuous Bluetooth smartphone connection is needed to collect data).