Correction: Machine Learning Approach For Frailty Detection In Long-Term Care Using Accelerometer-Measured Gait And Daily Physical Activity: Model Development And Validation Study
Frailty is a common and serious condition among older adults in long-term care, making early detection crucial. This research explored how wearable sensors and artificial intelligence could offer a new way to identify frailty by monitoring daily physical activity and walking patterns. The study involved equipping older adults with a small, wearable device called an accelerometer, which is similar to the technology found in smartphones that tracks movement. These devices collected data during a short walking test and over several days of everyday activities. The data, including various aspects of how a person walks (gait) and their overall activity levels, was then fed into several machine learning models. Machine learning is a type of artificial intelligence that allows computers to learn from data without being explicitly programmed, enabling them to find patterns and make predictions. The findings suggested that these models could effectively distinguish between frail and non-frail older adults. Specifically, the study highlighted that subtle changes in dynamic gait—such as how stable or symmetrical a person’s steps are—were more sensitive indicators of frailty than simpler measures like walking speed. Advanced AI analysis further revealed that frail older adults tend to have more varied, complex, and uneven (asymmetric) walking patterns. While an earlier interpretation indicated that frail individuals had higher gait symmetry scores, subsequent correction clarified that frail individuals actually exhibit lower gait symmetry scores, meaning their walking is indeed more asymmetrical. This work underscores the potential of using smart technology to improve frailty detection and ultimately enhance care for older adults.
Source: link to paper