Gait speed is a crucial indicator of health and function, particularly in aging and disease. Traditional gait assessments, conducted in clinical settings, are episodic and may not accurately reflect real-world gait due to factors like the Hawthorne effect. Advances in wearable technology offer the potential to measure gait in naturalistic settings, providing more comprehensive and objective data. Previous studies have shown discrepancies between controlled and free-living gait assessments, with at-home monitoring potentially offering enhanced sensitivity. While the accuracy of wearable-based gait measurements has been validated, the optimal monitoring duration for reliable estimates remains unclear. This study aimed to validate a single lumbar-worn accelerometer for gait speed measurement, compare in-lab and at-home gait speed measurements, assess the sensitivity of these measures to detect age-related differences, and determine the minimal at-home monitoring duration for reliable gait speed estimation.

Existing literature highlights the limitations of traditional, episodic gait assessments conducted in controlled environments. The Hawthorne effect and the inherent subjectivity of observational scales limit the generalizability of these assessments to real-world situations. The development and validation of wearable sensors for gait assessment has been the subject of several studies, which have shown promise but also pointed to issues like device placement, study population, and population-specific variability that may affect accuracy and reliability. Studies have also indicated that at-home monitoring captures gait variability not seen in clinical settings, offering a potentially richer dataset for detecting clinically meaningful changes and improving patient engagement in research. A key gap in the literature is the optimal at-home monitoring duration for reliable gait speed estimation.

Sixty-five healthy adults (33 younger, 18–40 years; 32 older, 65–85 years) participated. In-lab assessments involved walking three 4-meter laps on an instrumented mat (GAITRite), while wearing six wearable sensors (Opal and APDM). Gait speed was calculated using the GAITRite, APDM's proprietary algorithm, and an open-source algorithm (GaitPy) utilizing data from a single lumbar-mounted sensor. At-home monitoring involved continuous wear of a lumbar-mounted accelerometer (GeneActiv) for approximately one week. Gait speed was calculated using GaitPy for both in-lab and at-home data. Linear mixed-effects models were used to analyze in-lab and at-home gait speed data, considering age group, sex, and other covariates. Bootstrap resampling was used to determine the minimum data (steps and days) required for reliable gait speed estimation and to assess the impact of monitoring duration on detecting age group differences.

In-lab gait speed, estimated using various methods (GAITRite, APDM, GaitPy), did not significantly differ between age groups. However, at-home gait speed showed significant age-related differences, with older adults exhibiting slower speeds. This difference was more pronounced during weekdays than weekends. A moderate correlation was observed between in-lab and at-home gait speed. Bootstrap analysis indicated that at least two to three days of at-home monitoring are necessary to reliably estimate median and 95th percentile gait speed in healthy adults. For detecting age-related differences, two days were needed for median gait speed and only one day for 95th percentile gait speed. GaitPy, using a single lumbar-worn sensor, demonstrated good agreement with the GAITRite gold standard in the lab setting.

The study demonstrates that at-home gait speed assessment using wearable devices is more sensitive to age-related differences than traditional in-lab assessments. The discrepancy likely stems from the influence of daily life activities and cognitive factors on real-world gait, which are not captured in controlled settings. This supports the use of at-home monitoring to capture more realistic gait patterns, which may better reflect the functional capabilities of individuals and provide more sensitive outcome measures in clinical trials. The weak correlation between in-lab and at-home gait speed further underscores the importance of using naturalistic measures for evaluating functional ability. The finding that only a few days of data is sufficient for reliable gait speed estimation is significant for minimizing patient burden and improving trial feasibility.

This study validates the use of a single lumbar-worn accelerometer (GaitPy) for at-home gait speed assessment. At-home monitoring offers a more sensitive measure of gait speed, revealing age-related differences not apparent in in-lab assessments. The minimal monitoring duration of two to three days is clinically practical. Future research should explore the application of this methodology to diverse populations and clinical settings to further validate its efficacy in capturing meaningful differences in gait and functional capacity.

The study included only healthy adults and a relatively short at-home monitoring period. While efforts were made to ensure compliance, the lack of a robust method for determining participant compliance with continuous wear is a limitation. The analysis compared data from randomly selected days to the full dataset; however, further research could explore more robust methods for selecting data subsets. The impact of weekday versus weekend variations on gait was not explicitly addressed in the bootstrap analysis, but was noted in supplementary data.