Interindividual and intraindividual variation in repeated measures of 24-hour dietary recall in the elderly.

While several studies have examined the validity of the 24-hour recall as a method for assessment of dietary intake in older people, none has examined the components of variability of dietary data obtained from the 24-hour recalls of older adults. The ability to detect differences in group means, and estimates of correlation and regression coefficients, are all reduced when intraindividual (within-subject) variability in a measure is large in comparison to interindividual (between-subject) variability. The present study considered the components of variance of four 24-hour diet recalls taken bimonthly, for 220 adults whose mean age was 72 years. Two components of variability were estimated for 23 nutritional variables: an interindividual (between-subjects) variance component and an intraindividual (within-subjects) variance component. The ratio of intraindividual to interindividual variance differed substantially as a function of the nutritional parameter being assessed. Methods are presented for determining the number of dietary recalls needed to keep the attenuation of correlation and regression coefficients within a specified level. Power calculations for the test of equal means in two independent groups are also presented.