Predicting falls in older women

Abstract

Falls among older people are common and their occurrence is associated with detrimental effects on physical and psychosocial functioning. However, falls are not an inevitable consequence of ageing and there is growing evidence of effective interventions to prevent them. Accurate screening methods to identify high-risk populations are important if such strategies are to be cost-efficient. Epidemiological studies have identified a diverse group of risk factors for falls of different types in a variety of settings and patient groups. These have proved useful in delineating high-risk groups and have propagated a range of risk assessment tools for falls. Without an accepted taxonomy for the reporting of trials testing these instruments, direct comparison of results has been difficult. In frail older people, 'multi factorial assessment tools' have achieved some utility in the discrimination of fallers from non-fallers, whereas performance-based 'functional mobility assessments' appear to be more suited to predicting falls in groups of more active elders. The predictive value of these measures has been hampered by the complex and dynamic interaction between attendant risk factors and their variable influence in populations of different frailty profiles. Furthermore, current indices used in the prediction of falls are built upon statistical methodologies employing logistic regression, which fail to account for the breadth and depth of these associations in populations at risk of falling. Statistical representations more consistent with the complex modelling required in the design of falls risk assessment trials, such as tree classification techniques, may provide better results in future studies that aim to generate accurate predictors of falls.