Wearable Sensor-Based Prediction Model of Timed up and Go Test in Older Adults

Abstract

The Timed Up and Go (TUG) test has been frequently used to assess the risk of falls in older adults because it is an easy, fast, and simple method of examining functional mobility and balance without special equipment. The purpose of this study is to develop a model that predicts the TUG test using three-dimensional acceleration data collected from wearable sensors during normal walking. We recruited 37 older adults for an outdoor walking task, and seven inertial measurement unit (IMU)-based sensors were attached to each participant. The elastic net and ridge regression methods were used to reduce gait feature sets and build a predictive model. The proposed predictive model reliably estimated the participants’ TUG scores with a small margin of prediction errors. Although the prediction accuracies with two foot-sensors were slightly better than those of other configurations (e.g., MAPE: foot (0.865 s) > foot and pelvis (0.918 s) > pelvis (0.921 s)), we recommend the use of a single IMU sensor at the pelvis since it would provide wearing comfort while avoiding the disturbance of daily activities. The proposed predictive model can enable clinicians to assess older adults’ fall risks remotely through the evaluation of the TUG score during their daily walking.