Criteria for evaluation of measurement properties of clinical balance measures for use in fall prevention studies.

Work Package 3 of the Prevention of Falls Network Europe has evaluated measurement properties of clinical balance measures to be used to: (1) select participants for interventions with the goal to prevent falls in older people, and (2) assess the results of such intervention on balance function. Inclusion in a fall prevention study may be based on measures identifying subjects who have impaired balance or increased risk of future falls. We propose that an appropriate statistical method to analyse discriminative ability of a balance measure is discriminant analysis or logistic regression analysis. The optimal cut-off score is best determined by plotting a receiver-operating-characteristic curve for different cut-off values. The evaluation of predictors for risk of future falls should be based on a study design with a prospective data collection of falls. Sensitivity to change is a measurement property needed to evaluate the outcome of an intervention. The standardized response mean is frequently encountered in the literature and is recommended as a statistical measure of sensitivity to change in the context of an intervention study. Adequate reliability is a prerequisite for consistent measurement. Relative reliability may be reported as an intraclass correlation coefficient and absolute reliability as the within-subject standard deviation (s(w)), also called standard error of measurement. When measurement error is proportional to the score, calculation of a coefficient of variation can be considered. In a second paper, the authors will evaluate clinical balance measures for use in fall prevention studies based upon criteria recommended in this report.