Detecting prolonged sitting bouts with the ActiGraph GT3X.

Area of Science:

• Biomedical Engineering
• Physical Activity Epidemiology
• Wearable Technology

Background:

• The ActiGraph is widely used for physical activity monitoring but has limitations in accurately classifying posture and sedentary behavior.
• Accurate measurement of sedentary behavior, particularly prolonged sitting, is crucial for understanding its health implications.

Purpose of the Study:

• To develop and validate a novel ActiGraph posture classification algorithm for detecting prolonged sitting bouts.
• To compare the performance of the new algorithm against proprietary ActiGraph cut-points and the activPAL device.

Main Methods:

• A waist-worn ActiGraph (30 Hz) and activPAL were simultaneously worn by 38 office workers for a median of 9 days.
• Automated feature selection and a machine learning algorithm were employed for minute-based posture classification.
• Bland-Altman statistics were used to compare the new algorithm and ActiGraph cut-points (100 & 150 cpm, with/without LFE) against the activPAL criterion.

Main Results:

• The developed ActiGraph algorithm demonstrated high accuracy in predicting time spent in prolonged sitting bouts (bias ≤ 7 min/d).
• Among proprietary ActiGraph methods, only 150 cpm without LFE showed non-significant differences for prolonged sitting (bias ≤ 18 min/d).
• The 100 cpm with LFE cut-point accurately predicted total sitting time (bias ≤ 7 min/d), but neither cut-point is ideal for detailed bout analysis.

Conclusions:

• The new ActiGraph posture classification algorithm offers a significant improvement for measuring prolonged sitting bouts.
• For prolonged sitting measurement using ActiGraph cut-points, 150 cpm without LFE is recommended; for total sitting time, 100 cpm with LFE is suggested.
• The developed algorithm is recommended for research investigating the health effects of prolonged sitting measured by ActiGraph.