Finding the Goldilocks Zone for Toddler Accelerometry: How Many Days Are Needed for a Reliable Estimate of Physical...

Area of Science:

• Pediatrics
• Biomedical Engineering
• Kinesiology

Background:

• Accurate measurement of sedentary time and physical activity (PA) in toddlers is crucial for understanding early development.
• Machine learning (ML) models offer advanced capabilities for analyzing accelerometer data.

Purpose of the Study:

• To determine the optimal duration and daily wear time of accelerometers for reliable estimation of sedentary time and PA in toddlers using ML.
• To establish guidelines for accelerometer wear to maximize data quality and participant retention.

Main Methods:

• One hundred and nine toddlers wore hip-worn accelerometers for 7 days.
• A validated ML model assessed time in sedentary time, light PA, moderate to vigorous PA, and total PA.
• Intraclass coefficients and the Spearman-Brown prophecy equation were used to evaluate reliability across various wear time combinations (3-12 hours/day, 1-10 days).

Main Results:

• Reliability estimates ranged from 0.32 to 0.98, increasing with more hours per day and more days of wear.
• Four days of 6 hours of daily wear achieved a reliability threshold of 0.7.
• This wear protocol retained 94% of participants.

Conclusions:

• A recommendation of 6 hours of accelerometer wear per day for at least 4 days provides a balance between acceptable reliability and participant retention.
• These findings support the use of this protocol for ML-based analysis of toddler sedentary time and PA.
• This approach can facilitate further research into early childhood physical behavior using advanced ML techniques.