Background: There is limited information on reliable and valid measures of physical activity in older people with impaired function.
Objective: This study was conducted to compare the accuracy of single-axis accelerometers in recognizing postures and transitions and step counting with the accuracy of video recordings in people with stroke (n=14), older inpatients (n=14), people with hip fracture (n=8), and a reference group of 10 adults who were healthy.
Design: This was a cross-sectional study, evaluating the concurrent validity of small body-worn accelerometers against video observations as the criterion measure.
Methods: Activity data were collected from 3 sensors (activPAL) attached to the thighs and the sternum and from registration of the same activities from video recordings. Participants performed a test protocol of in-bed, transfer, and walking activities.
Results: The sensor system was highly accurate in classifying lying, sitting, and standing positions (100%) and in recognizing transitions from lying to sitting positions and from sitting to standing positions (100%). Placement of a sensor on the nonaffected leg resulted in less underestimation of step counts than placement on the affected leg. Still, the sensor system underestimated step counts during walking, especially at slow walking speeds (≤0.47 m/s) (limits of agreement=-2.01 to 16.54, absolute percent error=40.31).
Limitations: The study was performed in a controlled setting and not during the natural performance of activities.
Conclusions: The activPAL sensor system provides valid measures of postures and transitions in older people with impaired walking ability. Step counting needs to be improved for the sensor system to be acceptable for this population, especially at slow walking speeds.