Validity and Reproducibility of a Food Frequency Interview in a Multi-Cultural Epidemiology Study

Purpose: There is limited support for the validity and reproducibility of dietary assessment in culturally diverse populations. The goal of this study was to evaluate the comparative validity and reproducibility of a Food Frequency Questionnaire (FFQ) used in the observational, multi-cultural Insulin Resistance Atherosclerosis Study (IRAS).

Methods: Women (n = 186) were approximately equally distributed by ethnicity from one urban center (African Americans and non-Hispanic whites) and one rural center (Hispanics and non-Hispanic whites). The IRAS FFQ was modified from the National Cancer Institute Health Habits and History Questionnaire to include ethnic and regional foods. Validity was assessed by comparing dietary values, including supplements, obtained from the FFQ to the average intake estimated from a series of 8 24-hour dietary recalls collected by telephone over the same 1-year period. Reproducibility was assessed among women who reported no change in their usual diet (n = 133) by comparing data from the original IRAS FFQ (in-person) with the FFQ administered for the validity study (two to four years later, by telephone).

Results: Correlation coefficients for validity were statistically significant for most nutrients (mean r = 0.62 urban non-Hispanic white, 0.61 rural non-Hispanic whites, 0.50 African American, 0.41 Hispanic) and did not differ among subgroups of obesity or diabetes status. The median correlation coefficient for the total sample was 0.49. Correlations were lower for women with less than 12 years of education (mean r = 0.30; median r = 0.25). The lower correlations among Hispanics was largely explained by the lower educational attainment in that sample. For reproducibility, the mean correlation for nutrients evaluated was r = 0.62 (median r = 0.63) and did not differ for subgroups.

Conclusions: Although educational attainment must be considered, the IRAS FFQ appears to be reasonably valid and reliable in a diverse cohort.