Background: Functional assessments and direct measures of physical performance are standard components of community-based studies of older populations. Estimates of the reliability of these measures are necessary for the assessment of functional change.

Methods: The reproducibility of 13 measures of self-reported function and 11 direct measures of physical performance was assessed. A sample of subjects (N=199; > or =55 yrs) was selected from a larger population-based cohort. Subjects were tested in their homes twice, 48 hours apart, by the same interviewer to replicate study conditions. Age-adjusted kappa statistics were used to assess the reliability of measures of physical function; product moment correlation (Pearson r) and intraclass correlation coefficients (ICC) were used to assess direct measures of performance. A repeated measures model was used to assess learning or practice effects of performance, adjusted for age, sex, general health, and cognitive function.

Results: Age-adjusted kappa statistics were > or = .60 for most self-reported items. ICC ranged from .63 to .92. Significant improvements (practice effects) were found for the chair stand, walking speed, and the 360 degree turn. Measures of grip strength, reaching down, and hand dexterity were found to be reliable, with no significant test effect.

Conclusion: Three commonly used measures (chair stand, walking speed, and 360 degree turn) may be less reliable than previously reported. Sample sizes that may be needed to detect change in these areas of performance may be larger than previously estimated given this level of imprecision. Future studies of reproducibility should assess both the level of agreement and the presence of possible practice effects.