Dietary assessment: developing, validating, and improving a 24-hour Web food report questionnaire

Abstract

Developing, Validating, and Improving a 24-hour Web Food Report Questionnaire Jay Robert Desai Precise and accurate assessment of usual energy and nutrient intake among free living populations remains a major challenge for the study of diet and disease. The food frequency questionnaire (FFQ) carries a relatively low cost and participant burden, but it is much less accurate than multiple 24-hour telephone recalls (24TR). However, 24TRs are usually prohibitively expensive for large studies, and carry a high participant burden. Consequently, there remains a great need for alternative self-report approaches. Emerging web technology and internet penetration provide an opportunity to reduce dietary measurement error while increasing participant and investigator convenience. We developed the 284-item 24-hour Web Food Report Questionnaire (WFRQ) and conducted a pilot comparative validation study (n=51) examining energy and nutrient intake estimates from three WFRQs and the NCI Dietary History Questionnaire (NCI-DHQ) compared to three interviewer-administered 24TRs. Mean intake estimates were similar between the 24TR and the WFRQ but significantly lower with the NCI-DHQ for many nutrients. Correlations with the 24TR ranged from 0.24 to 0.73 and were typically lower with the NCI-DHQ than the WFRQ. Compared to estimated energy requirements the 24TR, WFRQ, and NCI-DHQ under-reported energy intake by 15%, 19%, and 25%, respectively. On average the WFRQ took 10 minutes to complete. A more in depth examination of reported foods and food amounts consumed, comparing 24TRs and WFRQs collected for the same day (n=46), revealed that the 24TR better captures foods used as additions, food-types, and for some foods, portion amounts. Incorporating more food detail and more effective portion size aids may improve the validity of the WFRQ. At the same time, refining the mapping of the WFRQ food items and portion categories to our food and nutrient database may also reduce measurement error. The WFRQ `beta' version is promising for research use in large studies involving free living populations.