Identifying bedrest using 24-h waist or wrist accelerometry in adults

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Sleep Science

Background:

• Accurate measurement of bedrest and wake periods is crucial for health research.
• Existing algorithms for accelerometry data often require adaptation for different populations.
• Automated detection of sleep and activity patterns can improve data collection efficiency.

Purpose of the Study:

• To adapt and refine a youth-specific decision tree (DT) algorithm for identifying bedrest in adults.
• To evaluate the accuracy of the adapted DT algorithm using accelerometry data from waist-worn and wrist-worn devices.
• To provide an openly accessible tool for automated bedrest detection in adult populations.

Main Methods:

• Healthy adults (n=141) wore accelerometers (waist and/or wrist) within an indirect calorimeter.
• Minute-by-minute accelerometry data were analyzed using a DT algorithm to identify bedrest and wake periods.
• Algorithm performance was assessed for sensitivity, specificity, and accuracy against objective room calorimeter data.

Main Results:

• The adapted DT algorithm achieved high accuracy in identifying bedrest for both waist (0.755) and wrist (0.859) worn accelerometers.
• Optimal algorithm parameters were determined for block length, threshold, and bedrest triggers for each device location.
• The DT algorithm demonstrated superior accuracy compared to the Cole-Kripke algorithm for wrist-worn devices.

Conclusions:

• The adapted decision tree algorithm accurately identifies bedrest in adults using accelerometry data from wrist or waist placement.
• This automated DT algorithm provides a reliable method for bedrest/sleep detection in both youth and adults.
• The "PhysActBedRest" R package offers open access to this validated algorithm for researchers.