Accuracy and Diversity of Wearable Device–Based Gait Speed Measurement Among Older Men: Observational Study

Abstract

Background Gait speed measurements are widely used in clinical practice, as slow gait is a major predictor of frailty and a diagnostic criterion for sarcopenia. With the development of wearable devices, it is possible to estimate the gait speed in daily life by simply wearing the device. Objective This study aims to accurately determine the characteristics of daily life gait speed and analyze their association with sarcopenia. Methods We invited community-dwelling men aged >50 years who had visited the outpatient clinic at a tertiary university hospital to participate in the study. Daily life gait speed was assessed using a wearable smart belt (WELT) for a period of 4 weeks. Data from participants who wore the smart belt for at least 10 days during this period were included. After 4 weeks, data from a survey about medical and social history, usual gait speed measurements, handgrip strength measurements, and dual-energy x-ray absorptiometry were analyzed. Results A total of 217,578 daily life gait speed measurements from 106 participants (mean age 71.1, SD 7.6 years) were analyzed. The mean daily life gait speed was 1.23 (SD 0.26) m/s. The daily life gait speed of the participants varied according to the time of the day and day of the week. Daily life gait speed significantly slowed down with age (P<.001). Participants with sarcopenia had significantly lower mean daily life gait speed (mean 1.12, SD 0.11 m/s) than participants without sarcopenia (mean 1.23, SD 0.08 m/s; P<.001). Analysis of factors related to mean daily life gait speed showed that age and skeletal muscle mass of the lower limbs were significantly associated characteristics. Conclusions More diverse and accurate information about gait speed can be obtained by measuring daily life gait speed using a wearable device over an appropriate period, compared with one-time measurements performed in a laboratory setting. Importantly, in addition to age, daily life gait speed is significantly associated with skeletal muscle mass of the lower limbs.

Digital Phenotyping in Health Using Machine Learning Approaches: Scoping Review

“…The most dominant study design was observational (n=28, 60.9%) [ 17 , 20 , 21 , 23 - 25 , 27 , 28 , 31 , 32 , 34 , 36 - 40 , 42 - 47 , 49 - 51 , 57 , 58 , 60 ], followed by randomized trials (n=10, 21.7%) [ 19 , 22 , 26 , 30 , 35 , 41 , 52 - 55 ], post hoc analyses of RCTs (n=5, 10.9%) [ 18 , 29 , 56 , 59 , 61 ], and post hoc analyses of observational studies (n=3, 6.5%) [ 16 , 33 , 48 ].…”

“…Three studies (6.5%) included patients undergoing hemodialysis [ 25 , 46 , 61 ]. Other clinical areas investigated included circadian rhythms [ 42 ], cough [ 57 ], sarcopenia [ 58 ], physical training [ 39 ], and rheumatoid arthritis and lupus erythematosus [ 41 ].…”

“…The categories used to map how active and passive data were collected in the included studies were as follows: wearable device, mobile app, wearable device + mobile app, wearable device + other, and other. Most (n=25, 54.3%) studies fell under the wearable device category [ 16 - 20 , 22 , 24 , 25 , 32 - 34 , 36 , 38 , 43 , 44 , 46 , 47 , 49 - 51 , 55 - 58 , 61 ]. Many (n=14, 30.4%) studies also collected data using a combination of wearable devices and a mobile app and thus fell under the wearable device + mobile app category [ 21 , 23 , 26 - 30 , 35 , 39 , 45 , 48 , 53 , 54 , 60 ].…”

“…Regarding the data analysis techniques, 22 (47.8%) studies used regression-based statistical methods [ 16 , 20 , 22 , 23 , 28 , 30 , 33 , 35 , 37 , 40 , 41 , 43 , 45 , 48 - 50 , 53 , 54 , 56 , 58 , 61 ], 2 (4.3%) used latent growth analysis [ 18 , 38 ], and 14 (30.4%) used other statistical analysis methods [ 21 , 24 - 26 , 29 , 31 , 32 , 34 , 42 , 44 , 46 , 47 , 52 , 55 ]. One (2.2%) study did not perform any statistical analyses because it was a case report [ 36 ].…”

“…Due to the observational nature of the studies, causal relationships between the passive and active data collected and outcome measures could not be confirmed. Some studies also reported device- and app-related limitations, including short battery life of smartwatches (leading to underestimation of physical activity) [ 21 ], challenges in keeping the app running 24/7 [ 60 ], no measurements of users’ interactions with mobile phone notifications [ 26 ], missing data [ 23 , 30 , 48 , 49 ], and drawbacks in the algorithms tested [ 16 , 32 , 45 , 57 , 58 ]. Another limitation reported was reliance on self-reported data, which included active data collected and those collected for outcome measurements.…”