Table 2.

Epoch-by-epoch agreement: classification of 4 sleep stages.

Variable		Overall					SNUBHa		CLCb
		Accuracy	Weighted F1	Cohen κ	Macro F1	Macro F1		Macro F1
Airable
	SleepRoutine (n=67)c	0.7106d	0.7166d	0.5565d	0.6863d	0.7188d		0.6551d
	SleepScore (n=38)	0.4329	0.4472	0.2065	0.4049	0.4408		0.3094
	Pillow (n=74)	0.2830	0.2906	0.0741	0.2588	0.2604		0.2564
Nearable
	Withings Sleep Tracking Mat (n=75)	0.4921	0.5007	0.2455	0.4496	0.4205		0.4837
	Google Nest Hub 2 (n=33)	0.4121	0.4089	0.0644	0.3009	0.2676		0.3299
	Amazon Halo Rise (n=28)	0.6634	0.6706	0.4807	0.6242	0.6231		0.6031
Wearable
	Google Pixel Watch (n=30)	0.6355	0.6143	0.4044	0.5669	0.5381		0.5925
	Galaxy Watch 5 (n=22)	0.6494	0.6499	0.4177	0.5761	0.6261		0.5651
	Fitbit Sense 2 (n=26)	0.6464	0.6296	0.4185	0.5814	0.5130		0.6268
	Apple Watch 8 (n=26)	0.5640	0.5731	0.2976	0.4910	0.5436		0.4203
	Oura Ring 3 (n=53)	0.5427	0.5518	0.3492	0.5186	0.5187		0.5211

^aSNUBH: Seoul National University Bundang Hospital.

^bCLC: Clionic Life Center.

^cThe number in parenthesis indicates the number of participants tested with each device.

^dTop-performing consumer sleep tracker.