Gait monitoring is a vital aspect of healthcare, especially with the increasing prevalence of digital health technologies like accelerometry. These technologies offer remote, continuous monitoring, providing a more comprehensive understanding of an individual's health status over extended periods. However, the challenge lies in standardizing gait metrics across diverse devices and populations. This is where the SciKit Digital Health (SKDH) package comes into play, offering a device-agnostic framework to address this issue.
The SKDH gait algorithm has undergone a series of enhancements, aiming to improve its performance against reference standards and reduce the need for manual parameter adjustments. These enhancements are particularly crucial for remote trials, where minimizing participant burden is essential. The lumbar location, despite its technical limitations, offers practical advantages, making it a popular choice for gait monitoring. However, its accuracy can be compromised due to its distal position relative to the feet and ground contact points, especially for spatial metrics.
To address these challenges, the SKDH-gait algorithm has been updated, focusing on reducing bias in spatial metric estimation and improving overall accuracy. These updates are significant and require new validation. This paper provides a step-by-step overview of these improvements and their validation using data from healthy adults and pediatric participants. It first examines each algorithmic component for potential enhancements and then assesses their impact on the complete gait algorithm and the generated metrics.
The results demonstrate robust evidence supporting the validity of the enhancements. The updated gait algorithm, gait v3, showcases high accuracy and reliability in capturing gait characteristics across various speeds and age groups. The findings highlight the importance of continuous gait monitoring, especially in patient populations, where even subtle changes in gait can be clinically meaningful. By reducing measurement error, more accurate algorithms enhance the sensitivity to detect disease progression and treatment response, leading to stronger statistical power in clinical trials.
The paper also emphasizes the practical advantages of the lumbar location, making it an appealing choice for large-scale monitoring and decentralized trial settings. The updated algorithm, with its improved accuracy, has the potential to revolutionize remote gait monitoring, providing a more reliable and efficient way to assess gait in various populations.
