A practical method for collecting 3-day food records in a large cohort

Recent studies suggest that diet records are more valid measures of nutrient intake than are food-frequency questionnaires. However, food records are considered unsuitable for large studies due to the need to train participants and to review and correct completed records.

We evaluated a self-administered 3-day food record protocol in Washington State. One hundred men and women age 50-76 years were mailed a food record and serving-size booklet. Sixty-five people returned a completed food record and were subsequently interviewed to obtain missing information. The food records were analyzed with and without added information from the interview.

The most common error was incomplete description, which affected 8% of recorded foods. Differences in mean nutrient intake between the uncorrected and corrected records were within 5%, and nutrient estimates from the 2 methods were highly correlated.

This streamlined protocol yielded data comparable to those collected by more burdensome protocols, suggesting that the use of food records may be feasible in large cohort studies.