Performance Evaluation

We evaluated the ability of the Oura Ring to detect age-related differences in PPG waveform, in comparison to a clinical-grade fingertip pulse oximeter. We demonstrate that PPG signals from consumer wearable can reasonably estimate vascular age, facilitating scalable monitoring of vascular health in the field.

We evaluated distal skin temperature increase as a marker of sleep onset and found that the rise in peri-sleep DST shows a consistent temporal profile for both naps and nocturnal sleep, which was similar in different environments, and across age and sex.

We evaluated the accuracy of Oura ring-derived HR and HRV measures against ECG, and found that accurate HRV measures can be obtained from Oura’s PPG-derived signals with a stringent validity proportion threshold of around 80% for each 5 min segment and aggregating over time windows of at least 30 min.

Wearable sleep trackers can rival or exceed the performance of ‘research actigraphs’ costing up to 3-5x more in assessing sleep timing, duration, and regularity. This finding affects the type of sleep tracker research groups purchase for long term ambulatory measurement of sleep.

On average, their accuracy falls with user age and with persons with less efficient sleep (either longer sleep latency or greater wake after sleep onset). This finding is of practical importance to persons who take sleep tracking seriously.

Not all ‘consumer wearables’ are comparably accurate. The low-cost OEM devices made by putting together sensors and off the shelf algorithms without evidence for testing or refinement, perform poorly.

One of the sources of inaccuracy in sleep measurement relates to misclassification of periods of quiet wakefulness as sleep. The team devised a protocol to test this in a realistic manner that should be incorporated into future testing protocols. We showed how this protocol can be used to differentiate better quality and lower quality trackers. This testing protocol can / will be used by future researchers seeking to evaluate in wearable sleep trackers in realistic settings.

We showed how the PPG waveform signal, changes according to sleep stage, reflecting greater parasympathetic influence at work during N3 compared to light sleep and REM. The waveform recorded in N3 is also associated with measures that correspond to relatively lower arterial stiffness. This provides additional evidence of the benefit of N3 duration.

For characterizing sleep regularity, we used large scale data from HPB’s HiSG study to show that observation periods longer than 1 week (up to 18 days) are necessary to achieve ~80% confidence in assessing sleep regularity.