Comparing objective measures of activity in the women health initiative study

Purpose: The study aims were: 1) to propose an improved version of Activity Intensity (AI) to summarize raw tri-axial accelerometry data; 2) to compare AI and Activity Count (AC) with regard to distinguishing between different activities and predicting energy expenditure.

Methods: 200 postmenopausal women performed 9 lifestyle activities in the laboratory, each wearing an accelerometer (Actigraph GT3X+) on the hip. 30 Hz tri-axial accelerometry data, as well as oxygen consumption during each activity were recorded. AI and AC was computed using the raw data, and Metabolic Equivalent of Task (MET) was computed using oxygen consumption. AI and AC were then compared on their performance of distinguishing between different activities and predicting MET.

Results: AI was showed to have a similar scale as AC does, but with a higher consistency. AI was also found to be more sensitive to sedentary and light activities, and therefore able to better distinguish between them. AI was showed to be better associated with MET and to have greater ability to differentiate activities with different levels.

Conclusion: The improved AI proposed here provides a novel and valid way to summarize densely sampled acceleration into activity level in any chosen epoch. AI not only provides consistent values which helps better distinguish between different activities, but also better associates with energy expenditure.