Misclassification of nutrient intake of individuals and groups using one-, two-, three-, and seven-day food records

Freudenheim, J. L (Dept of Social and Preventive Medicine, SUNY, Buffalo, NY 14214), N. E. Johnson, and R. L. Wardrop. Misdassification of nutrient intake of individuals and groups using one-, two-, three-, and seven-day food records. Am J Epidemiol 1987; 126:703-13.In this study, 6,844 food records were collected during three years (1979-1982) from 106 volunteer Wisconsin women, aged 35-65 years. Subjects recorded all intake of food, and vitamin/mineral supplements on a structured, precoded form. One-, two-, three-, and seven-day records were compared with usual intake of calcium, kilocalories, vitamin A and vitamin C. Usual intake was calculated using 37-72 food records per subject. Estimates of group means from a small number of records were not significantly different from mean usual intake (p > 0.05). Correlations with usual intake ranged from 0.43-0.64 and from 0.71-0.90 for the one day and the seven-day estimates, respectively. For the one-day record, 43-67% of subjects were correctly classified to the extreme quintiles of intake, 52-78% for the seven-day record. Classification was least good for vitamin A better for other nutrients with lower intraindividual variance. However, overall agreement with usual classification of assignment to quintiles even with the seven-day record was less than 55% for all four nutrients. Effects at extremes of intake might be more easily analyzed than dose-response relations.