Accuracy and Precision of Three Consumer-Grade Motion Sensors During Overground and Treadmill Walking in People With Parkinson Disease: Cross-Sectional Comparative Study

Abstract

Background Wearable motion sensors are gaining popularity for monitoring free-living physical activity among people with Parkinson disease (PD), but more evidence supporting the accuracy and precision of motion sensors for capturing step counts is required in people with PD. Objective This study aimed to examine the accuracy and precision of 3 common consumer-grade motion sensors for measuring actual steps taken during prolonged periods of overground and treadmill walking in people with PD. Methods A total of 31 ambulatory participants with PD underwent 6-min bouts of overground and treadmill walking at a comfortable speed. Participants wore 3 devices (Garmin Vivosmart 3, Fitbit One, and Fitbit Charge 2 HR), and a single researcher manually counted the actual steps taken. Accuracy and precision were based on absolute and relative metrics, including intraclass correlation coefficients (ICCs) and Bland-Altman plots. Results Participants walked 628 steps over ground based on manual counting, and Garmin Vivosmart, Fitbit One, and Fitbit Charge 2 HR devices had absolute (relative) error values of 6 (6/628, 1.0%), 8 (8/628, 1.3%), and 30 (30/628, 4.8%) steps, respectively. ICC values demonstrated excellent agreement between manually counted steps and steps counted by both Garmin Vivosmart (0.97) and Fitbit One (0.98) but poor agreement for Fitbit Charge 2 HR (0.47). The absolute (relative) precision values for Garmin Vivosmart, Fitbit One, and Fitbit Charge 2 HR were 11.1 (11.1/625, 1.8%), 14.7 (14.7/620, 2.4%), and 74.4 (74.4/598, 12.4%) steps, respectively. ICC confidence intervals demonstrated low variability for Garmin Vivosmart (0.96 to 0.99) and Fitbit One (0.93 to 0.99) but high variability for Fitbit Charge 2 HR (–0.57 to 0.74). The Fitbit One device maintained high accuracy and precision values for treadmill walking, but both Garmin Vivosmart and Fitbit Charge 2 HR (the wrist-worn devices) had worse accuracy and precision for treadmill walking. Conclusions The waist-worn sensor (Fitbit One) was accurate and precise in measuring steps with overground and treadmill walking. The wrist-worn sensors were accurate and precise only during overground walking. Similar research should inform the application of these devices in clinical research and practice involving patients with PD.

Step‐rate threshold for physical activity intensity in Parkinson’s disease

“…Step-rate per walking bout was calculated by dividing total number of steps taken by the 6 minutes and expressed as steps•per minutes. This method of manually counting steps has previously been undertaken in PD and other populations, 6,10 and provides an accurate measurement of actual steps taken while walking.…”

“…Of note, we are aware of one study that has generated cut‐points for quantifying MVPA from the rate of activity counts (ie, arbitrary unit of measurement for quantifying the intensity of bodily acceleration), rather than step counts, using research‐grade accelerometer data in PD 9 ; however, it is not feasible for the general population to use these accelerometers and interpret the output. Data from research‐grade accelerometers are not readily accessible without downloading the data via proprietary software, and such software might not provide accurate data based on existing algorithms in PD 10,11 …”

“…Of note, the development of step‐rate cut‐points for MVPA in mild‐to‐moderate PD is further supported by evidence for the accuracy and precision of commercially available motion sensors. For example, one study examined the accuracy and precision of three commercially available motion sensors for step counts during 6‐minute walking bouts in persons with PD and reported that a waist‐worn motion sensor was the most accurate and precise in measuring step counts compared to wrist‐worn motion sensors during both over‐ground and treadmill walking 10 . Another study examined the accuracy of two wrist‐worn motion sensors during 2‐minute bouts of over‐ground walking in measuring steps and physical activity intensity in persons with PD and reported that a wrist‐worn motion sensor was accurate in measuring steps but not physical activity intensity 12 .…”

“…Data from research-grade accelerometers are not readily accessible without downloading the data via proprietary software, and such software might not provide accurate data based on existing algorithms in PD. 10,11 Of note, the development of step-rate cut-points for MVPA in mild-to-moderate PD is further supported by evidence for the accuracy and precision of commercially available motion sensors. For example, one study examined the accuracy and precision of three commercially available motion sensors for step counts during 6-minute walking bouts in persons with PD and reported that a waist-worn motion sensor was the most accurate and precise in measuring step counts compared to wrist-worn motion sensors during both overground and treadmill walking.…”

“…For example, one study examined the accuracy and precision of three commercially available motion sensors for step counts during 6-minute walking bouts in persons with PD and reported that a waist-worn motion sensor was the most accurate and precise in measuring step counts compared to wrist-worn motion sensors during both overground and treadmill walking. 10 Another study examined the accuracy of two wrist-worn motion sensors during 2-minute bouts of over-ground walking in measuring steps and physical activity intensity in persons with PD and reported that a wrist-worn motion sensor was accurate in measuring steps but not physical activity intensity. 12 Collectively, commercially available motion sensors may be accurate and precise in measuring steps in PD, and this supports the interest in developing step-rate thresholds for quantifying MVPA that can be applied for interpreting the output from such devices for prescribing, monitoring, and tracking physical activity in persons with PD.…”