2.3.2. Free-Living Study

In both the I.Family study and the LIV2013 study the ActiGraph GT3X+ (ActiGraph, Pensacola, FL, USA) was used to collect acceleration data and was worn over the right hip in an elastic belt around the waist. The participants were instructed to wear the accelerometer for seven days and to remove it during sleep and during water-based activities. The accelerometer was set to record acceleration data at 30 Hz sampling rate with an acceleration range of ±6 g and idle sleep mode were enabled. The raw triaxial data was extracted according to available file specifications [24].

Raw triaxial data was processed using our new algorithm to produce a measure of PA intensity expressed in mg including a band-pass filter with the cut-point range of 0.29–10 Hz [17,18]. The epoch-length was set to 3 s to capture children's activities as well as intermittent activities in adults [18]. Valid days with at least 12 h were included in the analyses. Samples in which the sensor status was idle according to the idle sleep-mode were considered non-wear time. More subjects in the adult group wore the accelerometer during sleep compared to the children; therefore, the acceleration recorded between 23:00–06:00 was not included in the analysis. Our previous research has confirmed that processing of GT3X+ data and AX3 data generate similar output [16].

Accelerometer mg cut-points for LPA, MPA, VPA and VVPA from the VO2net-based calibration and the standard MET-based calibration were applied. The participants were divided into four age-categories: Children (<13 years, n = 321 (50% female), adolescents (13–16 years, n = 96 (50% females)), younger adults (<50 years, n = 366 (75% female)) and older adults (≥50 years, n = 369 (68% female)). The time distribution in the intensity categories was compared between the age-categories with the VO2net-based calibration versus the MET-based calibration. As the VO2net-based calibration would imply lower Speedabs and mechanical work in children and adolescents compared to the MET-based calibration, and consequently lower accelerometer mg cut-points, we assume more free-living PA in these age groups applying the VO2net-based accelerometer mg cut-points.

All data processing and analyses in the two sub-studies were performed in MATLAB 2018b (MathWorks, Natick, MA, USA).
