On food frequency questionnaires: the contribution of open-ended questions and questions on ethnic foods

BACKGROUND: Epidemiologic studies of diet must balance the goal of comprehensiveness with that of minimizing subject and study burden. Information on the impact of certain additional questionnaire elements may aid in such decisions. METHODS: We administered a food-frequency questionnaire in a large multiethnic cohort of women (Study of Women's Health Across the Nation) in 1996-1997. Ancillary questionnaires asked about "any other foods you eat at least once a week" and about consumption of 8-16 ethnic foods included for respondents of Hispanic, Chinese, or Japanese ethnicity. We assessed the impact of these 2 ancillary questionnaires by examining the mean nutrient amount they contributed and their effects on rank-order correlations and categorizations of nutritional estimates. We examined the impact of ethnic foods both within the relevant ethnic group and among non-Hispanic whites. RESULTS: Rank-order correlations (rs) between nutrient estimates with and without the open-ended question were 0.99 for almost all nutrients and all ethnic groups. Nutrient amounts added by the open-ended question rarely exceeded 2% of the total. Weighted kappa statistics for quintile classifications were greater than 0.9. Similarly, asking white respondents about ethnic foods added little to nutrient estimates, and all correlations between estimates with and without the ethnic foods were at least 0.95. The effect of ethnic foods on nutrient means within the relevant ethnic group was more substantial, although most correlations between estimates with and without the added foods were still above 0.9. CONCLUSIONS: Open-ended food questions added little to nutrient estimates or rankings in any of the ethnic groups. Specific questions about ethnic foods were useful only within ethnic groups, for which they increased mean nutrient estimates and had modest impact on ranking.