Validation of nocturnal resting heart rate and heart rate variability in consumer wearables
Dial, Hollander, Vatne, Emerson, Edwards, Hagen · 2025 · Physiological Reports 13(16):e70527
What this work gives Aellic
13 healthy adults, 536 nights against a Polar H10 chest strap (single-lead ECG) as the reference. Resting heart rate reached Lin's concordance correlation of 0.86 (Polar Grit X Pro, which the authors themselves grade “poor”) to 0.98 (Oura Gen 4), with mean absolute percentage errors of 1.67–3.00 %. The corresponding HRV errors were three to five times larger on the same nights and devices (MAPE 5.96–16.32 %). That error ratio — not the concordance coefficient — is what supports treating nocturnal resting heart rate as the steadier of the two signals.
Correction of 4 November 2026
Aellic's recovery weighting cited “RHR CCC 0.86–0.98, considerably more robust than any PPG-derived HRV” with no source at all, and the comparison was wrong twice over. First, concordance for HRV in this very study reaches 0.99 (Oura Gen 4) and 0.94 (WHOOP 4.0) — ABOVE every resting-heart-rate value, so the claim is refuted by the same table it was drawn from. Concordance scales with between-subject spread, which is wide for RMSSD and narrow for resting heart rate; comparing the two on that statistic is not meaningful in either direction. The defensible comparison is relative error, and it is now stated that way. Second, no Apple device was tested here, yet the figure was used to justify weighting Apple's resting heart rate. Corrected again on 2026-11-04: the reference was described only as “an ECG”; it was a Polar H10 chest strap recording a single-lead ECG, not a clinical multi-lead ECG.
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.
- Recovery Recovery score weighting and age references 1 site