Wearable sleep monitor ring parallels research-grade actigraphy

The Oura ring, a wearable sleep tracking device, can detect sleep and wake states in adolescents as well as research-grade actigraphy, according to a new Singapore study. On the other hand, there is still room for improvement in terms of sleep staging.

“In our evaluation of the Oura ring, over multiple nights of sleep in each participant, we found the device to give comparable assessment of sleep timing and duration with respect to polysomnography (PSG) as the Actiwatch 2 research actigraph at the latter’s default settings,” the researchers said. “Both devices significantly underestimated adolescent sleep in this sample across all three sleep opportunities.”

The study included 53 adolescents (aged 15–19 years; 28 girls) who were monitored overnight using both the Oura ring and the Actiwatch 2 devices; measurements were collected multiple times, over three levels of sleep opportunities. The SOMNOtouch device was used for PSG and was set as the standard.

After excluding data points due to technical issues, 27 PSG, 27 Oura ring, and 13 Actiwatch recordings qualified for analysis. Compared to PSG, the Oura ring consistently and significantly underestimated total sleep time (TST) by 32.8–47.3 minutes, while overestimating wake after sleep onset (WASO) by 30.7–46.3 minutes (p<0.001 for all). [Nat Sci Sleep 2021;13:177-190]

In addition, the Oura ring significantly underestimated rapid eye movement (REM) and light (stages N1 and N2) sleep, and overestimated time in deep sleep (stage N3).

Nevertheless, epoch-by-epoch analysis revealed excellent Oura accuracy in terms of sleep-wake cycles, with specificity and sensitivity values of 0.88 and 0.89, respectively, across all times in bed (TIBs) scenarios.

Repeated-measures analysis of variance (ANOVA) showed a significant main effect of device on TST and WASO measurements across all TIB conditions. Subsequent pair-wise analyses showed that Oura significantly, but only slightly, underestimated TST more than the default Actiwatch setting (M10), with average differences ranging from 6.9–13.4 minutes (p<0.05).

In terms of WASO, Oura performed comparably as Actiwatch in most TIB conditions. In the 9-hour TIB scenario, Actiwatch proved to be slightly superior, overestimating WASO by 8.4 minutes less (p=0.006).

On the other hand, the Actiwatch setting that has a higher wake threshold performed significantly better at measuring TST and WASO compared to the M10 default actigraphy setting and to the Oura ring.

“This similar underestimation by the Oura ring and the Actiwatch 2 at default M10 setting may be due to both devices being optimized for adults, and thus warrants further validation in an adult population,” the researchers said.

“The possibility for improving the accuracy of sleep detection for different populations, such as different age groups, by customizing sensor sensitivity is exemplified by the better correspondence of Actiwatch 2 output and PSG with lowered sensitivity to motion,” they added.