gms | German Medical Science

Introduction: Commercial smartwatches can access health information, which is becoming increasingly common in daily life. The aim of this study was to validate the accuracy of step counts from a commercial wearable, the Garmin^® Vívosmart 4 smartwatch, with the CE-certified (Class 1m medical device) SENS motion^® accelerometer system [1] in patients with ulcerative colitis (UC). The data are from the FilgoColitis study, a prospective, open-label, multi-centre, non-interventional, 24-month observational therapy study in ulcerative colitis (UC).

Methods: Step count data were collected from a subset of FilgoColitis participants using two wearable devices: the Garmin^® Vívosmart 4 smartwatch and the SENS motion^® accelerometer [2]. SENS motion^® data is available at 15-minute intervals using the steps of continuous walking [3]. Garmin smartwatch data was retained if heart rate, exercise, or steps were available for a full 15-minute interval. Our analysis included a variety of approaches, such as summarising step count data using medians and interquartile ranges (IQRs) as well as calculating Lin's concordance correlation coefficient (CCC) of step counts between both devices over different time intervals. Furthermore, Bland-Altman analysis and equivalence test were performed, using the mean step count of both devices for each patient.

????Results: We analysed a total of 43,002 comparisons from 40 patients. 20,749 (48.25%) had a step count of 0 measured by both devices. For the remaining 22,253 (51.75%) data points, the median and IQR of measured steps were 71 (18, 191) for Garmin and 70 (17, 187) for SENS motion in a 15-minute interval. The median difference and IQR between SENS motion and Garmin were -3 (-30, +28). The CCC increased from 0.83 at the 15-minute level to 0.93 at the daily level. Bland-Altman analysis showed a bias of 6 steps and a limit of agreement (LOA) of -39 to +51 steps. Equivalence tests showed that step count equivalence can only be assumed when the equivalence bounds are at least 13 steps.

Discussion: Our results indicate a substantial agreement between continuous step counts recorded by the Garmin^® and SENS motion^® devices in the real-world context of a non-interventional study in UC patients. Bland-Altman analysis showed a slight bias towards SENS motion, with a relatively wide LOA. This may be due to the different placement of the device, sensor sensitivity to detect steps but could also be due to measurement artefacts. Our analysis revealed that a discrepancy of 13 steps must be considered to ensure that both devices are treated as equivalents. This threshold should be accounted when addressing specific research questions. On the other hand, the increase in CCC at longer time intervals indicates that the differences in step count diminish throughout the day.

Conclusions: This study shows that commercially available fitness trackers can be a valuable tool for measuring steps, e.g. in patients with UC. They offer a convenient alternative to more specialised and expensive medical devices. Wearable devices can collect additional health metrics that can contribute to a comprehensive assessment of an individual's health status, which could improve the care of UC patients.

The authors declare that they have no competing interests.

The authors declare that a positive ethics committee vote has been obtained.