They're tracking movements. Apple has a doc on how it works here[0]
>Apple Watch tracks movement with triaxial accelerometer signals, which capture coarse motion of the
body as well as fine movements including motion associated with breathing. Apple developed an algorithm
that uses the accelerometer time series data to classify Breathing Disturbances that occur during sleep
tracking, which are temporary interruptions in the breathing pattern.
> A total of 1499 participants were enrolled, with 1448 completing the study. The sensitivity was 66.3%
(95% CI: 62.2% to 70.3%), and the specificity was 98.5% (95% CI: 98.0% to 99.0%), demonstrating that
the feature meets the design objectives to confidently identify sleep apnea while minimizing false positives.
>It’s important to note that the specificity was 100% (95% CI: 99.7 to 100%) for the
normal category, indicating that all participants with a “positive” algorithm result had at least mild sleep
apnea. Also, sensitivity was higher in the severe category at 89.1% (95% CI: 83.7% to 93.2%), indicating
that the large majority of severe cases were identified.
>Apple Watch tracks movement with triaxial accelerometer signals, which capture coarse motion of the body as well as fine movements including motion associated with breathing. Apple developed an algorithm that uses the accelerometer time series data to classify Breathing Disturbances that occur during sleep tracking, which are temporary interruptions in the breathing pattern.
> A total of 1499 participants were enrolled, with 1448 completing the study. The sensitivity was 66.3% (95% CI: 62.2% to 70.3%), and the specificity was 98.5% (95% CI: 98.0% to 99.0%), demonstrating that the feature meets the design objectives to confidently identify sleep apnea while minimizing false positives.
>It’s important to note that the specificity was 100% (95% CI: 99.7 to 100%) for the normal category, indicating that all participants with a “positive” algorithm result had at least mild sleep apnea. Also, sensitivity was higher in the severe category at 89.1% (95% CI: 83.7% to 93.2%), indicating that the large majority of severe cases were identified.
0:https://www.apple.com/health/pdf/sleep-apnea/Sleep_Apnea_Not...