Ordinal and timed balance measurements: reliability and validity in patients with stroke

The aim of this study of 38 patients with stroke was to describe the intrasession and intersession reliability and validity of ordinally graded and timed balance tests. Patients had a mean age and time since stroke of 67 years and 21 days respectively. Balance and gait were both timed with a stop-watch and given an ordinal score. Balance was timed for four different task levels: feet apart, feet together and unilateral stance (either limb first). Subjects who demonstrated the ability to complete the task for 60 seconds proceeded to the next, more difficult, task level. Most subjects remained within an ordinal balance category during a session; 81.6% for session 1 and 86.8% for session 2. The distribution of ordinal balance scores between sessions revealed that 31 of the 38 subjects remained within the same ordinal balance score, with the remainder of subjects scoring higher during session 2. The weighted Kappa coefficient for these scores was 0.858, indicating excellent reliability. The magnitude of the intrasession and intersession intraclass correlation coefficients for unilateral balance times indicated acceptable reliability. The Spearman correlations showed that balance measurements were a valid indicator of walking performance regardless of whether timed or ordinal measurements were used to characterize balance and gait. Given the findings of this study relevant to reliability and validity, the use of the more precise timed balance times cannot be advocated.