Identification of Fall-Risk Factor Degradations Using Quality of Balance Measurements

Abstract

Falls concern a third of the people aged over 65y and lead to the loss of functional ability. The detection of risks factors of falls is essential for early intervention. Six intrinsic risk factors of fall: vision, vestibular system, joint range of motion, leg muscle strength, joint proprioception and foot cutaneous proprioception were assessed with clinical tests before and after temporarily degradation. Standing balance was recorded on a force plate. From the force plate, 198 parameters of the centre of pressure displacement were computed. The parameters were used as variables to build neural network and logistic regression model for discriminating conditions. Feature selection analysis was performed to reduce the number of variables. Several models were built including 3 to 10 conditions. Models with 5 or less conditions appeared acceptable but better performance was found with models including 3 conditions. The best accuracy was 92% for a model including ankle range of motion, fatigue and vision contrast conditions. Qualities of balance parameters were able to diagnose impairments. However, the efficient models included only a few conditions. Models with more conditions could be built but would require a larger number of cases to reach high accuracy. The study showed that a neural network or a logistic model could be used for the diagnosis of balance impairments. Such a tool could seriously improve the prevention and rehabilitation practice.