Performance of the Verily Study Watch for measuring sleep compared to polysomnography

Area of Science:

• Wearable technology
• Sleep science
• Biomedical engineering

Background:

• Polysomnography (PSG) is the gold standard for sleep assessment but is resource-intensive.
• Wearable devices offer a potential alternative for convenient, long-term sleep monitoring.
• Accurate sleep detection by wearables is crucial for clinical and research applications.

Purpose of the Study:

• To evaluate the performance of the Verily Study Watch (VSW) against PSG for key sleep measures.
• To assess the VSW's accuracy in sleep vs. wake classification and sleep stage detection.
• To quantify the bias and agreement of VSW-derived sleep parameters with PSG.

Main Methods:

• Data collected from 41 adults during a single overnight laboratory PSG.
• Epoch-by-epoch comparison of VSW and PSG for sleep/wake and sleep stage classification.
• Analysis of total sleep time (TST), wake after sleep onset (WASO), sleep onset latency (SOL), sleep efficiency (SE), and number of awakenings (NAWK).
• Performance metrics included sensitivity, specificity, Cohen's kappa, and Bland-Altman analyses.

Main Results:

• High sensitivity (0.97) and specificity (0.70) for sleep vs. wake classification.
• Cohen's kappa of 0.64 for 4-class sleep stage detection, indicating moderate agreement.
• Proportional bias observed in most VSW sleep measures, with notable mean biases for TST (14.0 min) and WASO (-13.1 min).
• Low bias for sleep onset latency and number of awakenings.

Conclusions:

• The Verily Study Watch demonstrates good performance for sleep vs. wake detection and moderate accuracy for sleep staging.
• While some sleep parameters exhibit bias, the VSW is a viable tool for tracking overnight sleep in free-living conditions.
• Further research may refine algorithms to minimize bias in specific sleep metrics for enhanced accuracy.