Age and environment-related differences in gait in healthy adults using wearables

“[Abstract] Many traditional assessments of physical function utilized in clinical trials are limited because they are episodic, therefore, cannot capture the day-to-day temporal fluctuations and longitudinal changes in activity that individuals experience. In order to understand the sensitivity of gait speed as a potential endpoint for clinical trials, we investigated the use of digital devices during traditional clinical assessments and in real-world environments in a group of healthy younger (n = 33, 18–40 years) and older (n = 32, 65–85 years) adults. We observed good agreement between gait speed estimated using a lumbar-mounted accelerometer and gold standard system during the performance of traditional gait assessment task in-lab, and saw discrepancies between in-lab and at-home gait speed. We found that gait speed estimated in-lab, with or without digital devices, failed to differentiate between the age groups, whereas gait speed derived during at-home monitoring was able to distinguish the age groups. Furthermore, we found that only three days of at-home monitoring was sufficient to reliably estimate gait speed in our population, and still capture age-related group differences. Our results suggest that gait speed derived from activities during daily life using data from wearable devices may have the potential to transform clinical trials by non-invasively and unobtrusively providing a more objective and naturalistic measure of functional ability.

[Article] Gait speed is considered an informative and reliable clinical measure in a wide range of disease populations; it is often referred to as the sixth vital sign. Previous studies have shown that lower gait speed is associated with cognitive decline, falls, and mortality, and is an important indicator of health and function in ageing and disease5. Conventional gait assessment is performed in the laboratory or clinic, using a combination of observational scales (e.g., functional gait assessment) and performance tests (e.g., 6 min walk test), where individuals perform prescribed walking tests under observation. These types of assessments may not be able to provide a reliable estimate of real-world gait because (1) they are administered episodically, (2) can be subjective in nature, and (3) gait can be altered under observation (Hawthorne effect).

[..] gait speed derived from data captured under continuous free-living conditions is slower than gait speed measured in the clinic in frail elderly or community-dwelling older adults. It has been hypothesized that continuous, at-home monitoring provides a richer and more comprehensive view of an individuals experience with the disease.

[..] Despite growing evidence that at-home monitoring provides a more comprehensive assessment of gait, several hurdles need to be addressed to enable broader clinical adoption. Further clinical research that adopts and validates standardized sensor-based methods in various populations under free-living conditions is needed to translate research findings and novel methods into practice. In addition, questions remain regarding the processing and interpretation of at-home data. An open question is the optimal monitoring duration necessary for reliable characterization of gait under free-living conditions. Obtaining data from multiple days and investigating day-to-day variability of at-home measures is necessary, in order to assess the minimum required acquisition period and obtain reliable real-world estimates. However, additional days of monitoring results in increased patient burden and might reduce compliance in clinical trials, especially for patients suffering from particularly debilitating diseases. The required number of at-home monitoring days still remains arbitrary and can be affected by multiple factors, such as the type of disease, treatment, age, geographical location, and socioeconomic status. In fact, previous studies have used data captured during monitoring durations that ranged from one day to several weeks for their analysis. Several studies have also investigated day-to-day variability of at-home measures. These studies have reported a minimum of three to six days of measurements required to obtain reliable estimates of physical activity, energy expenditure, and heart rate. However, the monitoring duration necessary for deriving a reliable estimate of gait speed under free-living conditions is not well understood. We are aware of only one recent study that reported a minimum monitoring duration of 3 days for reliable estimation of gait speed in slow-walking older adults with sarcopenia.

Herein, we present our work on assessment of gait in healthy younger (18–40 years) and older (65–85 years) adults in both the laboratory and home setting, using a single lumbar-worn wearable accelerometer. We aim to (1) assess the validity of measurements derived using the lumbar-worn wearable device by comparing them with those provided by a system that use multiple wearable devices (APDM) and an instrumented mat (GAITRite), (2) test the sensitivity of the median and 95th percentile gait speed derived from in-lab walk test and continuous at-home monitoring data to detect age-related group differences, and (3) propose a minimal at-home monitoring period for estimating gait speed reliably.

[..] In this study, we confirmed the validity of a lumbar-worn sensor and observed that in-lab gait speed estimated using data from a single lumbar-worn sensor (GaitPy) showed good agreement with an instrumented mat (GAITRite). We further observed a consistent bias for both GaitPy (13%) and APDM (5%, reference six-device system) compared to the instrumented mat. However, GaitPy-derived gait speed had higher variability than APDM, likely the result of reliance on a single device. This suggests that there might be a sensitivity trade-off when utilizing fewer devices for measuring gait speed. Although, based on our previous findings, this trade-off may be minimal. Using gyroscope in combination with the accelerometer could potentially lead to better gait characterization, especially important to capture rotational information for turns and falls, with a cost of higher battery usage. Despite limitations in sensitivity, gait speed estimated using a single lumbar-worn device has been shown to distinguish between disease states, as well as detect the effects of treatment. Additional work investigating the validity of gait feature estimation using a single lumbar-mounted device in various disease populations is still needed. Our results suggest that a single lumbar-worn device can provide sufficient accuracy for monitoring gait under free-living conditions and at the same time minimize participants burden.

[..] In our study, in-lab gait speed derived using either wearable sensors or GAITRite was unable to distinguish between the two age groups, whereas at-home gait speed showed the older group walked significantly slower than younger participants. We further confirmed our findings using uninstrumented measurements, which were collected as part of the traditional Short Physical Performance Battery (SPPB) assessment on the same participants (i.e., the time for participants to perform the walk test timed using a stop watch), and found no differences between age groups based on stop watch time (average time to walk 4-m, younger group: 3.8 ± 0.5 s, older group: 3.7 ± 0.5 s, [..] p = 0.3). Furthermore, other gait metrics such as step time and stride length derived from in-lab assessments failed to differentiate the age groups. In-lab measures are acquired during single visits, whereas at-home measurements enable continuous evaluation over prolonged periods, providing the ability to capture nuanced therapeutic effects. Our findings are consistent with evidence that while a participant might change his/her behavior for a short period of time under observation (e.g., no age group difference for gait speed during in-lab assessments), it is unlikely that they will be able to do so during long periods of passive monitoring under free-living conditions.

[..] In healthy adults, real-world gait speed is expected to decrease approximately 0.03 m/s every decade over the life span of an adult, resulting in a difference of around 0.12 m/s after 40 years. In our study, we observed that at-home gait speed differed by 0.09 m/s on average between the younger and older groups (approximately four decades apart), and that group difference in median gait speed was driven by weekday rather than weekend periods. We also observed that age-based differences existed for various bout length, especially for short (ranging between 10 and 30 s) and medium bouts (ranging between 30 and 60 s), although gait speed increased with bout length (effect of age group: χ2 = 6.62, p = 0.02, effect of bout length: χ2 = 557, p < 10−16). Furthermore, not only gait speed but other gait features differed by age group. Similar trend was observed in patient populations such as PD, where significant group differences in at-home gait features were observed compared to healthy volunteers. Although these are cross-sectional studies, accurate estimation of real-world gait and evaluating its sensitivity to clinically meaningful change; e.g., the disease state or its progression, are extremely important. We have shown that gait metrics derived from at-home monitoring with a single lumbar-worn sensor provided more sensitive information to differentiate two age groups compared to in-clinic assessments, and that there is only a weak association between at-home and in-lab gait speed. This result was recently reported in older adults, and it was also shown that gait speed derived during longer walking bouts in the laboratory appears to be better correlated with at-home measurements. These findings provide further evidence for implementing wearable devices in clinical trials for monitoring physical function at home instead of in-clinic assessments. The use of wearable devices out-of-lab settings not only provides a better assessment of real-world activity, but also brings additional benefits of remote monitoring. For example, it enables to enroll patients who live in remote regions with difficult access to clinic, or reduces the exposure of participants who may be vulnerable to healthcare settings during a pandemic. In summary, remote data capturing offers the ability to design decentralized trials, which may become crucial for a wide variety of clinical trials in the near future.

[..] Our results suggest that at least two to three days of monitoring is required to estimate both median and 95th percentile gait speed in healthy volunteers. When the age groups were analyzed separately, we found only one to two days of data was needed to reliably estimate gait speed in the older cohort, whereas two to three days were needed for the younger cohort. This result suggests more day-to-day variance in gait speed in the younger cohort, as also reflected by the significant effect of day type, as well as age group by day type interaction. In addition, our results suggest that two days of data were necessary to estimate differences between young and old using median gait speed, whereas one day was needed for 95th percentile gait speed. Indeed, the upper and lower ICC bounds were tighter for 95th percentile gait speed compared to median gait speed, suggesting less day-to-day variability of 95th percentile gait speed.”

Full article, Czech MD, Psaltos D, Zhang H et al. NPJ Digital Medicine 2020.9.30