Reproducibility of reports of frequency of food use in the Tecumseh Diet Methodology Study.

Frequency reports in 1967-1969 for 83 foods were compared to frequency reports for the same foods asked retrospectively in 1982-1983 about use in 1967-1969 for 1,184 respondents aged 45-64 in the Tecumseh Community Health Study cohort. Food frequency reports of the current (1982-1983) diet were also compared to baseline reports to indicate diet stability and to retrospective reports to indicate the influence of current diet on retrospective reporting. Comparisons were also made for 13 food groups and vitamin A and C indices. Short-term reproducibility of seven foods, measured by two reports 1-3 weeks apart in 1967-1969, was compared to long-term reproducibility for the same foods. Short-term reproducibility was consistently greater than long-term. Reproducibility was associated positively with stability of use. For the two nutrient indices and nine of the ten food groups examined, mean intake estimated from the retrospective report differed significantly from mean intake from the baseline report. Agreement between retrospective and current reports exceeded agreement between retrospective and baseline reports, indicating that retrospective reports were greatly influenced by current behavior; yet, the retrospective reports were better indicators of baseline consumption than were the current reports. While misclassification from use of the retrospective report to measure baseline consumption was large, risk differences may still be detected with usual sample sizes.