Validation of an automated sleep detection algorithm using data from multiple accelerometer brands.

Area of Science:

• Biomedical Engineering
• Sleep Science
• Wearable Technology

Background:

• Polysomnography (PSG) is the gold standard for sleep assessment but is resource-intensive.
• Wearable accelerometers offer a more accessible alternative for sleep monitoring.
• Automated algorithms are needed to process accelerometer data for sleep analysis.

Purpose of the Study:

• To evaluate the criterion validity of an automated sleep detection algorithm using wrist-worn accelerometers against PSG.
• To assess the reliability of sleep estimates from Axivity, GENEActiv, and ActiGraph accelerometers.

Main Methods:

• 30 healthy volunteers underwent a 1-night PSG assessment while wearing three types of accelerometers on each wrist.
• Sleep parameters including sleep period time window (SPT-window), sleep duration, onset, waking time, efficiency, and wake after sleep onset (WASO) were extracted using the GGIR package.
• Agreement was assessed using 95% equivalence tests, intraclass correlation coefficients (ICCs), and limits of agreement (LoA).

Main Results:

• Accelerometer-derived sleep estimates, except for WASO, were within a 10% equivalence zone of PSG.
• Moderate reliability was found for SPT-window duration, sleep duration, and sleep onset (ICCs ≥ 0.54).
• Good reliability was observed for waking time (ICCs ≥ 0.80), but poor reliability for sleep efficiency and WASO (ICCs ≥ 0.08).
• Low mean bias was noted, but wide limits of agreement were observed for most sleep estimates.

Conclusions:

• Automated sleep detection algorithms in wrist-worn accelerometers provide comparable measures to PSG for sleep duration and onset.
• These devices are less accurate for measuring sleep efficiency and wake after sleep onset (WASO).
• Further refinement of algorithms is needed to improve the accuracy of wake detection during sleep periods.