Validity of proximity sensor-based wear-time detection using the ActiGraph GT9X

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Activity Monitoring

Background:

• Accurate measurement of physical activity is crucial for health research.
• Wearable sensors like the GT9X offer a promising approach for objective activity monitoring.
• Evaluating the accuracy of wear-time detection algorithms is essential for reliable data collection.

Purpose of the Study:

• To assess the performance of proximity sensor-based wear-time detection using the GT9X device.
• To compare GT9X wear-time detection accuracy in laboratory versus free-living conditions.
• To evaluate GT9X performance against established wear-time algorithms (Troiano 2007, Choi 2012).

Main Methods:

• Fifty-two participants engaged in laboratory or free-living protocols.
• GT9X sensors were worn on the wrist and hip.
• Wear-time was validated against direct observation (lab) and temperature/self-report (free-living).

Main Results:

• In laboratory settings, wrist-worn GT9X showed 93% sensitivity and 49% specificity for wear detection.
• Free-living protocols yielded 72% sensitivity and 90% specificity (wrist) and 84% sensitivity and 92% specificity (hip).
• GT9X demonstrated lower sensitivity but higher specificity compared to Troiano 2007 and Choi 2012 algorithms.

Conclusions:

• The GT9X proximity sensor exhibits variable accuracy in wear-time detection across different settings.
• While specificity is generally high, sensitivity can be suboptimal, particularly in laboratory conditions.
• Sole reliance on GT9X proximity sensor data for wear-time detection is not advised due to observed inaccuracies.