Predicting mental impairment in sarcopenic elderly women using machine learning and association rules
Abstract views: 85 / PDF downloads: 39
DOI:
https://doi.org/10.5281/zenodo.12601047Keywords:
Sarcopenia, cognitive impairment, elderly women, physical activity, machine-learningAbstract
Sarcopenia, a prevalent condition in the elderly characterized by muscle mass and function deterioration, is associated with increased risks of falls, functional decline, frailty, and mortality. This study investigates the link between sarcopenia and cognitive impairment in elderly women and develops a machine-learning prediction model based on association rules to forecast mental status. A total of 67 community-dwelling women aged 60 and above participated in this cross-sectional study. Cognitive function was assessed using the Mini-Mental State Examination (MMSE), and physical activity levels were measured through self-reported activity logs and the six-minute walk test (6MWT). Regularized Class Association Rules (RCAR) were employed to create a prediction model. Results indicated that weekly walking, increased moderate physical activity, and reduced sitting time were significantly associated with lower severity of mental impairment. Specifically, women with a higher Skeletal Muscle Index (SMI) and consistent moderate physical activity demonstrated better cognitive performance. The RCAR model achieved high accuracy (94%), balanced accuracy (93.9%), sensitivity (92.9%), and specificity (94.9%) in predicting cognitive impairment. These findings emphasize the importance of physical activity in mitigating cognitive decline in sarcopenic elderly women and highlight the potential of machine-learning approaches in developing predictive models for clinical applications. Future research targeting sarcopenia could play a crucial role in improving both physical and mental health in the aging population.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2024 Journal of Exercise Science & Physical Activity Reviews
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.