Parameterizing and validating existing algorithms for identifying out-of-bed time using hip-worn accelerometer data...

Area of Science:

• Gerontology
• Biomedical Engineering
• Physical Activity Measurement

Background:

• Accurate assessment of daily physical activity and sedentary behavior is crucial for understanding health in older adults.
• Hip-worn accelerometers are widely used to objectively measure physical activity, but accurate identification of wear time (out-of-bed time) is essential for valid results.
• Existing algorithms for out-of-bed time detection require validation and optimization for specific populations, such as older women.

Purpose of the Study:

• To parameterize and validate two existing algorithms (Tracy and McVeigh) for identifying out-of-bed time using 24-hour hip-worn accelerometer data in older women.
• To compare the performance of algorithms using vertical axis (VA) and vector magnitude (VM) data.
• To assess the impact of algorithm-derived wear time on physical activity and sedentary time estimates.

Main Methods:

• 628 older women (mean age 80 years) wore accelerometers for up to 7 days, concurrently completing sleep logs.
• A visual analysis protocol served as the criterion measure for in-bed periods.
• Algorithms were trained on 314 participants to optimize thresholds for sensitivity and specificity, then validated on the remaining 314 participants.

Main Results:

• All tested algorithms showed high agreement with the criterion measure for waking wear time.
• The McVeigh algorithm using vector magnitude (McVeigh_VM) demonstrated the highest agreement, with sensitivity=0.92, specificity=0.87, and kappa=0.80.
• Physical activity and sedentary time estimates adjusted using McVeigh_VM were not statistically different from those using the criterion measure.

Conclusions:

• The McVeigh algorithm, optimized with vector magnitude data, is a suitable and accurate method for automated identification of waking wear time in older women.
• This validated algorithm can be reliably used in research when manual visual analysis is not feasible, enabling large-scale objective physical activity monitoring.
• Accurate out-of-bed time detection is critical for reliable physical activity and sedentary behavior assessment in aging populations.