Separating bedtime rest from activity using waist or wrist-worn accelerometers in youth

Area of Science:

• Biomedical Engineering
• Kinesiology
• Pediatric Health

Background:

• Sedentary behavior and physical activity (PA) monitoring are crucial in youth.
• Technological advances enable continuous PA monitoring using accelerometers.
• Distinguishing sleep from activity is essential for accurate accelerometer data analysis.

Purpose of the Study:

• Develop and validate a decision tree algorithm for accurate separation of bedtime rest and activity in youth.
• Determine optimal accelerometer cut-points for waist- and wrist-worn devices.
• Compare the accuracy of the developed algorithm against existing methods.

Main Methods:

• Collected minute-by-minute accelerometry data from 81 youth (10-18 years) during a 24-h stay in a whole-room indirect calorimeter.
• Utilized Receiver Operating Characteristic (ROC) curve analysis to establish accelerometer cut-points for rest and activity.
• Validated the algorithm's classification accuracy against calorimeter-measured metabolic rate and movement data.

Main Results:

• Optimal cut-points were identified: 20 counts/min (waist) and 250 counts/min (wrist) for rest; 500 counts/min (waist) and 3,000 counts/min (wrist) for activity.
• The decision tree algorithm demonstrated high accuracy in classifying bedtime rest and activity in the validation group (waist: 0.872 AUC; wrist: 0.943 AUC).
• The developed algorithm significantly outperformed commonly used automated algorithms for both waist- and wrist-worn accelerometers (p<0.001).

Conclusions:

• Validated cut-points for waist- and wrist-worn uniaxial accelerometers accurately separate bedtime rest from activity in youth.
• This method provides a reliable tool for analyzing long-term physical activity data in naturalistic settings.
• The findings support improved assessment of sedentary behavior and PA patterns in pediatric populations.