Abstract
Community mobility, encompassing both active (e.g., walking) and passive (e.g., driving) transport, plays a crucial role in maintaining autonomy and social interaction among older adults. This study aimed to quantify community mobility in older adults and explore the relationship between GPS- and accelerometer-derived metrics and fall risk. METHODS: A total of 129 older adults, with and without a history of falls, were monitored over an 8 h period using GPS and accelerometer data. Three experimental conditions were evaluated: GPS data alone, accelerometer data alone, and a combination of both. Classification models, including Random Forest (RF), Support Vector Machines (SVMs), and K-Nearest Neighbors (KNN), were employed to classify participants based on their fall history. RESULTS: For GPS data alone, RF achieved 74% accuracy, while SVM and KNN reached 67% and 62%, respectively. Using accelerometer data, RF achieved 95% accuracy, and both SVM and KNN achieved 90%. Combining GPS and accelerometer data improved model performance, with RF reaching 97% accuracy, SVM achieving 95%, and KNN 87%. CONCLUSION: The integration of GPS and accelerometer data significantly enhances the accuracy of distinguishing older adults with and without a history of falls. These findings highlight the potential of sensor-based approaches for accurate fall risk assessment in community-dwelling older adults.