Introduction: To avoid counting movement artifacts as steps, some activity monitors require several consecutive seconds of stepping before steps are retroactively counted towards the total step accumulation.
Purpose: To study how algorithms in different activity monitors affect the number of steps recorded during brief, intermittent walking bouts when the walking interval or rest interval is manipulated.
Methods: Participants (n=21; 19-57 years) wore StepWatches on the ankle, the Fitbit Zip, New Lifestyles Digi-Walker SW-200, Omron HJ-322U, and ActiGraph GT3X on the hip, and the Fitbit Charge, Garmin Vivofit 2, Polar A360, Withings Pulse Ox and ActiGraph GT3X on the wrist. Part A: 5 trials where participants walked 4, 6, 8, 10, or 12 steps followed by 10 s standing rest. Part B: 6 trials where participants walked 4 steps followed by 1, 2, 4, 6, 8, or 10 s standing rest. Cadence was 100 steps/min and each trial lasted 2 min. Hand-counted (HC) steps served as the criterion. The data for each device were reported as percent of actual steps.
Results: In Part A, the StepWatch, Digi-Walker SW-200, and ActiGraph GT3X with low frequency extension (LFE) (hip and wrist) were within 6.2% HC steps. The ActiGraph GT3X with moving average vector magnitude (MAVM) (hip and wrist) and Polar A360 required extremely long bouts (>12 steps) of continuous walking, so almost no steps were recorded. All other devices required 6-10 steps to be taken before recording (See Figure 1). In Part B, for most of these devices, even 1-2 s rest was sufficient to break up a walking bout, resulting in a failure to record steps.
Conclusion: Some activity monitors use step-counting algorithms that screen out steps taken in short walking bouts. Since brief intermittent walking bouts are common in daily life, the inability of activity monitors to record every step is likely to be a major source of error contributing to the undercounting of steps taken throughout the day.