A Validation of Six Wearable Devices for Estimating Sleep, Heart Rate and Heart Rate Variability in Healthy Adults
Miller, Sargent, Roach · 2022 · Sensors 22(16):6317
What this work gives Aellic
n = 53 across 6 devices against polysomnography: agreement on multi-class sleep stage only 50–65 %, κ 0.20–0.52 (Apple 0.20 up to Somfit 0.52). Sleep sensitivity 90–98 %. This is why the stage distribution now weighs only 0.10 in Aellic instead of 0.25.
Correction of 25 July 2026
The registry filed this paper under “chinoy2022” — Chinoy is not an author at all. The upper bound of κ 0.65 noted earlier was also wrong; it comes from robbins2024. And “sensitivity ≥95 %” holds for only two of the six devices.
The entry on the register was wrong. It was not replaced but kept, with what holds instead standing next to it.
Where this sits in the engine
1 declared site: constants and thresholds in the code that point at this entry.
- Sleep Sleep stages and bedtime 1 site