Estimating Sedentary Time from a Hip- and Wrist-Worn Accelerometer.

Area of Science:

• Physical Activity Epidemiology
• Biomedical Engineering
• Wearable Technology Validation

Background:

• Sedentary behavior (SB) is linked to adverse health outcomes.
• Accurate measurement of SB in free-living conditions is crucial for research.
• ActiGraph accelerometers (AG) are commonly used to estimate SB, but their validity varies by placement and method.

Purpose of the Study:

• To validate existing ActiGraph GT3X+ accelerometer (AG) methods for estimating sedentary behavior (SB) under free-living conditions.
• To compare the accuracy of hip-worn versus wrist-worn AG placements for SB estimation.
• To identify the most reliable methods for quantifying SB using AG data.

Main Methods:

• Forty-eight young adults wore AGs on the hip and wrist during four 1-hour free-living sessions.
• Direct observation video served as the criterion measure for SB (sitting/lying posture, <1.5 METs).
• Thirteen different AG data processing methods were evaluated, including count-based, raw acceleration, and machine learning models.

Main Results:

• Hip-worn AG methods CPM100, CPM150, Soj1x, and Soj3x accurately estimated SB.
• Wrist-worn methods, including Sedentary Sphere and ENMO44.8, showed significant over or underestimation of SB.
• Mean absolute percent error was generally lower for hip-based methods compared to wrist-based methods.

Conclusions:

• Accurate group-level SB estimation is achievable with hip-worn AGs using count-based (CPM100, CPM150) or machine learning (Soj1x, Soj3x) approaches.
• Wrist-worn AG methods are not recommended for accurate or precise SB estimation.
• Future research should focus on developing open-source calibration datasets to improve SB measurement.