Wearable Device and Algorithm for Fall Prevention: Monitoring Fall Risk Using Foot Motion Measured by an In-Shoe Motion Sensor

Falls are a significant public health concern worldwide, particularly for older adults. In this study, we aim to create an effective quantitative indicator for long-term daily fall risk assessment suitable for use with in-shoe motion sensors (IMSs). We used the likelihood of an individual being a faller (ranging from 0 to 1) as a representation of the quantitative fall risk indicator. This indicator was constructed using data from 40 nonfallers and 24 fallers, all female older adults with distal radius fractures (DRFs) caused by falls within 6 months (denoted as 0 and 1, respectively). The indicator was determined using Fisher's discriminant, naïve Bayesian, or logistic regression approaches. Predictors were derived from four types of gait parameters and four types of physical ability metrics, all of which can be measured or estimated using IMS-measured gait. To prevent overfitting, we performed principal component analysis (PCA) on the predictors. To validate our method, we assessed Pearson's correlation between clinical experts-rated fall risk scores and the constructed fall risk indicator using a separate group of 19 older female adults. The Pearson's correlation obtained was 0.766. We successfully developed a fall risk indicator using IMS-measured gait, demonstrating the potential for IMSs to achieve long-term daily fall risk assessment.