Comparison of postural classification from thigh-worn Actigraph GT3X+ and ActivPal3 accelerometers under laboratory and free-living conditions

Purpose: This study compared sitting, standing, and stepping classifications from thigh-worn Actigraph and ActivPAL monitors under laboratory and free-living conditions.

Methods: Adults wore both monitors on the right thigh while performing 6 sitting, 2 standing, 8 stepping, and 1 cycling activity under laboratory observation (n=21), and during three days free-living (n=18). Percent of time correctly classified into postures was calculated under laboratory conditions. Between-monitor percent agreement and weighted kappa were calculated under free-living conditions.

Results: n the laboratory, both monitors correctly classified 100% of time in standing activities and >95% of time in four sitting activities. Both monitors demonstrated substantial misclassification of laboratory stool sitting (Actigraph 14%, ActivPAL 95%), and ActivPAL (14%) misclassified sitting with legs outstretched more often than Actigraph (0%). Actigraph misclassified more time spent descending stairs (14%) and ascending stairs (8%) than ActivPAL which correctly classified >95% of time in all stepping activities. The ActivPAL classified cycling as stepping 93% of time, but the Actigraph classified cycling as stepping <1% of time. During free-living wear, Actigraph and ActivPAL data matched second-by-second had high observed agreement (86%) and substantial agreement when accounting for chance (weighted kappa=0.77). The levels of agreement are due primarily to the relatively high volume of time recorded as sitting (Actigraph 64%, ActivPAL 62%), but there were differences in time recorded as standing (Actigraph 21%, ActivPAL 27%), and stepping (Actigraph 15%, ActivPAL 11%).

Conclusions: Differences in data processing algorithms based on thigh angle may have resulted in disagreement in posture classification between thigh-worn Actigraph and ActivPAL.