Performance comparison of different interpretative algorithms utilized to derive sleep parameters from wrist...

Area of Science:

• Sleep Science
• Biomedical Engineering
• Wearable Technology

Background:

• Wrist actigraphy is a common tool for sleep monitoring.
• Interpretative algorithms (IAs) are crucial for analyzing actigraphy data.
• Comparing IA performance is essential for accurate sleep assessment.

Purpose of the Study:

• To evaluate and compare the performance of four popular interpretative algorithms (Cole-Kripke, Rescored Cole-Kripke, Sadeh, and UCSD) for deriving sleep parameters from wrist actigraphy.
• To determine the most suitable IA for assessing sleep in healthy adults and for detecting changes in sleep over time.

Main Methods:

• A comparative study involving 40 healthy adults.
• Simultaneous sleep assessment using Motionlogger® Micro Watch Actigraphy (MMWA) and Zmachine® Insight+ electroencephalography (EEG).
• MMWA data analyzed by four IAs; EEG data scored by its proprietary IA as a reference standard.

Main Results:

• All four MMWA algorithms demonstrated high sensitivity (~94-98%) but moderate specificity (~42-54%) for sleep detection compared to EEG.
• All algorithms underestimated Sleep Onset Latency (SOL).
• Sadeh IA showed the least bias in estimating Wake After Sleep Onset (WASO), Total Sleep Time (TST), and Sleep Efficiency (SE).
• Sadeh and Rescored Cole-Kripke IAs had the highest agreement with EEG (Cohen's Kappa ~51%).
• UCSD IA had the lowest agreement (Cohen's Kappa ~47%) but the smallest minimum detectable change.

Conclusions:

• The Sadeh IA is recommended for deriving general sleep parameters in healthy adults.
• The UCSD IA is most suitable for monitoring changes in sleep parameters over time or in response to interventions.
• Algorithm choice depends on the specific research or clinical objective.