Developing a multivariate prediction model of falls among older community-dwelling adults using measures of...

To examine whether measures of neuromuscular control and proprioceptive acuity were predictive of falls in an older community-dwelling population and to develop a multivariate prediction model.

Fifty-eight adults aged above 60 living independently in the community were recruited for a prospective falls study. On entry, they undertook a Sensory Organisation Test (SOT) and an Active Movement Extent Discrimination Assessment (AMEDA) and completed a short fall risk questionnaire. Participants were monitored for falls over the subsequent 12 months. Prior to analysis, falls were classified into three categories based on the difficulty of the activity being undertaken and the demands of the environment in which the fall occurred. Logistic regression was used to predict the probability of a fall.

For falls occurring under the least challenging circumstances, the model fitted using the AMEDA score and two of the questions from the fall risk questionnaire, related to balance and confidence, achieved a specificity of 87% and sensitivity of 83%. Falls occurring in more challenging circumstances could not be predicted with any accuracy based on the variables recorded at inception.

This study highlights the importance of considering the heterogeneous nature of falls. Poorer proprioceptive acuity appears to play a role in falls occurring where neither the environment nor the activity is challenging, but not in falls occurring in other circumstances. Falls in the least-challenging circumstances affected 15% of participants, but this group was considerably more likely to have multiple falls, increasing their vulnerability to adverse consequences.