Raw and Count Data Comparability of Hip-Worn ActiGraph GT3X+ and Link Accelerometers

Area of Science:

• Physical Activity Measurement
• Biomedical Engineering
• Wearable Technology

Background:

• Ensuring comparability between accelerometer models is crucial for accurate inter- and intrastudy comparisons in physical activity research.
• ActiGraph accelerometers are widely used, but differences between models like the GT3X+ and GT9X Link require investigation.

Purpose of the Study:

• To compare the comparability of raw and count data between hip-worn ActiGraph GT3X+ and GT9X Link accelerometers.
• To evaluate the performance of different data processing methods (raw vs. count data) for physical activity intensity classification.

Main Methods:

• Adults (n=26) wore ActiGraph GT3X+ and Link accelerometers during a 80-min protocol with varied activities.
• Raw accelerometer data (60 Hz) and derived activity counts were analyzed in 30-s epochs.
• Machine learning (Montoye 2015) and regression models (Freedson 1998, 2011) were used to predict energy expenditure (METs) and classify activity intensity.

Main Results:

• Count data demonstrated higher correlations (e.g., VM: 1.00 ± 0.00) between accelerometers compared to raw data (e.g., VM: 0.80 ± 0.28).
• Freedson count-based models showed significantly higher percent agreement (95.1%-95.5%) for activity intensity classification than the Montoye raw data model (61.5%).

Conclusions:

• Accelerometer count data exhibit superior comparability between the ActiGraph GT3X+ and Link models.
• Improvements in raw data processing, such as data filtering or enhanced models, are necessary to achieve better comparability between these accelerometer devices.