Background/Purpose: Gait speed has been associated with many clinical outcomes (e.g. frailty, mortality, joint replacement need, etc.) relevant for rheumatologic conditions. Measuring Gait speed (stride length x step count/time) typically requires significant clinician/staff time or a gait lab with specialized equipment. Our formative work was to measure “step count” via smartphones as a first step in making gait speed measurement available for patient home follow-up.

Methods: We developed and tested a mobile App-based method to count steps, comparing a hardware (pedometer) vs. software (“Shake algorithm”) approach. Shake algorithm software allows for adjustment of amplitude (how big a shake equates a step) and refresh rate (how often the software counts a shake/step). We conducted calibration and validation phases. In both, subjects carried iOS and Android phones (one in each front pants pocket) simultaneously, and electronic gait timers measured the time it took to cover pre-measured distances.

Calibration phase: Subjects walked a 20m course at normal and slow (<1.0 m/s) speeds for 9 different shake algorithm settings (x3 walks for each). We determined which setting most closely matched investigator observed step counts. Validation phase: we used the selected shake algorithm settings for the smartphones and recruited subjects to each complete walks (x3) of 6, 10 and 20m at both slow and normal (>1.0m/s) speeds. 

We compared step difference (absolute difference) from observed step counts to hardware (pedometers) and software (shake algorithm) derived step counts. We used generalized estimated equation adjusted (participant level) negative binomial regression models of absolute step difference from observed step counts, to determine optimal settings (calibration) and subsequently to gauge performance of the shake algorithm settings and pedometers across different distances and speeds (validation).

Results: Calibration: 270 observations across 5 individuals were used to determine optimal smartphone shake algorithm settings for slow and fast walking speeds. Validation: Compared to observed step count, the shake service outperformed pedometer across all distances at slow, and at 6m for normal speed on iOS. On the Android phone, the shake service outperformed the pedometer at slow speed for 6 and 20m, and at 20m for normal speed (Table).

Conclusion: Software based approaches such as the shake algorithm, which can have parameters adjusted for optimized measurement of step can be adjusted to outperform fixed hardware such as pedometers. These results will facilitate bringing the capture of performance based metrics such as gait speed to patient home follow-up, using technology users already own (smartphones) to optimize data capture for research and clinical care across rheumatologic conditions.