Measuring standing balance in adults: reliability and minimal real difference of 14 instrumental measures

The study provides estimates of reliability and minimal real difference (MRD) (i.e. the minimal change significantly different from zero, expressed in the original units) of 14 parameters obtained from six motor tasks of standing balance on the EquiTest (dynamic) and the Balance Master (static) force platforms. The different tasks and parameters allowed an assessment of balance in three domains: quiet standing, perturbed standing and multidirectional leaning. Fifteen healthy adults (eight men and seven women; age 22-42 years) were studied at baseline and retested 1 and 3 weeks later. The significance level was set at P-value less than 0.05 and adjusted for multiplicity within sets of tests reflecting the same balance domain (Bonferroni corrections). Repeated analysis of variance modelling revealed a moderate yet significant time trend across the three time points, suggesting a practice effect for the mean of one out of the 14 parameters. Changes across pairs of time points did not reach significance (Tukey's post-hoc test). Test-retest reliability across the three time points and across pairs of time points was estimated for each parameter using the intraclass correlation coefficients (ICCs) (model 3.1; range 0-1, none to perfect reliability). Across the three time points, the ICCs ranged from 0.21 to 0.85 (>0.60 in nine out of the 14 cases). The MRDs were computed from the ICCs. For all 14 parameters showing a time trend, absolute changes (root mean squares of differences) were minimal. Thus, albeit overestimated for one parameter, the MRDs provide useful thresholds for changes to be interpreted as dependent on therapeutic interventions.