In a validation study conducted by the esteemed sleep research team from the National University of Singapore, Oura’s latest sleep staging algorithm reached accuracies of 76-78% for 4-stage classification (light, deep, REM, and wake) compared to gold-standard polysomnography.
This study sets itself apart by collecting a large number of nights in a realistic at-home setting. Most sleep studies involve laboratory settings, rely on a limited dataset of ~20-30 nights, and only collect one night per participant. This study collected 157 nights from 58 participants and collected up to 3 nights of data per individual.
Check out the full paper in Nature and Science of Sleep or read the key findings below:
- Oura was 76-78% accurate for 4-stage classification (light, deep, REM, and wake)
- Oura was 92-93% accurate for 2-stage classification (sleep, wake)
- Oura’s latest sleep staging algorithm outperforms older versions, especially for younger adults
Oura’s new sleep staging algorithm was refined using machine learning on as many data points from the ring as possible across 3 larger datasets from around the globe. This ensured diversity in age and race, and consequently improved performance in the present validation study.