Wrist-specific accelerometry methods for estimating free-living physical activity.

Area of Science:

• Biomedical Engineering
• Exercise Physiology
• Wearable Technology

Background:

• Accurate physical activity measurement is crucial for health research.
• Wrist-worn accelerometers are increasingly used but their accuracy compared to hip-worn devices needs validation.
• Existing predictive models for wrist accelerometers vary, necessitating a comparison of their performance.

Purpose of the Study:

• To compare physical activity estimates from nine wrist-specific accelerometry models against a hip-specific reference method.
• To evaluate the accuracy of different analytical approaches for wrist-based physical activity monitoring.

Main Methods:

• A prospective cohort study involving 110 participants wearing accelerometers on the wrist and hip for one week.
• Utilized three-axis accelerometer data to calculate physical activity estimates using three linear and six artificial neural network wrist-specific models.
• Compared wrist-derived estimates against a hip-specific reference method at both epoch (≤60s) and weekly levels.

Main Results:

• All nine wrist-specific models yielded different total weekly moderate-to-vigorous physical activity (MVPA) values compared to the hip reference.
• At the epoch level, the Hildebrand et al. (2014) model showed the strongest correlation (r=0.69) and highest agreement (94.1%) for MVPA.
• Significant differences (p<0.001) were observed in the determination of sedentary behavior, light physical activity, and MVPA based on the analysis method.

Conclusions:

• Inconsistent analysis methods can lead to incomparable physical activity results.
• While a linear wrist model approximated hip-based weekly MVPA, modest epoch-level agreement highlights the need for improved analytical techniques.
• Further research is required to enhance the accuracy of physical activity estimates from wrist-worn accelerometers.