With the increasing popularity of consumer activity trackers, which provide users with information about physical activity measures such as steps taken and energy expenditure along with easily monitored physical activity tools and goals, provides a potentially powerful tool for use in populations with cardiovascular disease (CVD). These devices are used to track free-living physical activity behavior as an outcome after different treatments and to encourage individuals to be more active as a preventive or management tool for CVD. However, before they are implemented broadly into the clinical health field, more research is needed to evaluate their performance accuracy in the free-living environment.
The researchers compared different activity monitor results during treadmill and simulated free-living activities to manually counted steps using the StepWatch (SW). Hickey and colleagues recruited fifteen participants to perform laboratory-based treadmill exercise at varying speeds and simulated free-living activities, such as cleaning a room, while wearing different activity monitoring devices (including activPAL, Omron HJ720-ITC, Yamax Digi- Walker SW-200, two ActiGraph GT3Xs in “low-frequency extension” and in “normal-frequency” mode, an ActiGraph 7164, and an SW). Participants also wore monitors for a day in their free-living environment.
Based on current findings, pedometers and accelerometers are accurate in detecting steps at normal walking speeds. However, at very slow treadmill walking and during running, the errors in step estimates are large and variable among monitors. An important finding of this study is that with the exception of the SW pedometer, pedometers and accelerometers were less accurate in detecting steps from non-rhythmic and multidirectional movements such as those that take place during activities of daily living. This has implications for studies using monitor-based step estimates to characterize physical activity and relating this physical activity metric to health. Substantial underestimation of daily steps may occur if participants spend a higher proportion of their day in activities of daily living.
This study highlights the need to verify step-counting accuracy of activity monitors with activities that include different movement types and directions. It is important to understand the origin and magniture of errors in step-counting during free-living conditions so that prediction algorithms may be refined to improve step estimates.