Background: Wearable, consumer-grade activity trackers have become widely available as a means of monitoring physical activity in the form of step counts. However, step counts may not be accurate in persons with Parkinson disease (PD) due to atypical gait characteristics.
Objective: This study aimed to investigate the accuracy of 4 consumer-grade activity trackers in individuals with PD while ambulating during continuous and discontinuous walking tasks.
Design: This study used a cross-sectional design.
Methods: Thirty-three persons with PD (Hoehn & Yahr stages 1-3) donned 4 models of activity trackers on the less affected side of their bodies. Participants performed 2 continuous walking tasks (2-minute walk tests at comfortable and fast speeds) and 2 discontinuous walking tasks (a simulated household course and an obstacle negotiation course) in an outpatient setting. Bland-Altman plots and intraclass correlation coefficients [ICC(2,1)] were computed as a measure of agreement between actual steps taken (reference standard: video recording) and steps recorded by each tracker.
Results: The accuracy of the activity trackers varied widely, with ICCs ranging from -0.03 to 0.98. Overall, the most accurate device across all tasks was the Fitbit Zip, and the least accurate was the Jawbone Up Move during the simulated household course. All activity trackers were more accurate for continuous walking tasks compared with discontinuous walking tasks. Waist-mounted devices were more accurate than wrist-mounted devices with continuous tasks. Bland-Altman plots revealed that all activity trackers underestimated step counts.
Limitations: All walking tasks were measured over relatively short distances.
Conclusions: In persons with mild-to-moderate PD, waist-worn activity trackers may be prescribed to monitor bouts of continuous walking with reasonable accuracy; however, activity trackers have little utility in monitoring discontinuous walking common in household settings.