Physical activity tracking using mobile devices: Can a heterogeneous sample of smart phones accurately quantify steps?

Abstract

Physical activity monitoring and adherence to exercise programmes can be objectively reported through wearable or mobile devices. However, a large variety of mobile devices are available to potential researchers. Therefore, the purpose of this study is to determine the ability of a random sample of mobile device to accurately quantify steps. A healthy sample of 39 students volunteered to take part in the study (male: 20; female: 19, combined stature 170 ± 17.5cm and mass 67 ± 23.7kg). All participants walked on a treadmill at three speeds (2.5, 5.0, 7.5 km/h) for no more than five minutes. Their personal devices (16 iOS and 23 Android smartphones) were used to track the number of steps. Concurrently, steps were recorded by a video camera. The percentage error between the actual number of steps and those reported by the devices were calculated. No significant differences were present between the devices (p=0.191) or walking speeds (p=0.641). A Spearman’s correlation was present between the actual steps and those captured by the devices (r=0.80, p<0.0001). Limits of agreement were established using a Bland-Altman test and ranged between -35.7% and 87.8% error for the various walking speeds. Measurement biases for the three walking speeds ranged between -6.1% and 11.3%. The majority of devices showed a similar level of accuracy and agreement between step counts but tended to underestimate the number of steps. Future research should use a single device model, operating software and application in order to clarify the present findings.

Keywords: Wearable devices, physical activity, remote monitoring, smartphone, mobile application