Associations of Race/Ethnicity, Education, and Dietary Intervention with the Validity and Reliability of a Food Frequency Questionnaire: The Women's Health Trial Feasibility Study in Minority Populations
This report describes the associations of race/ethnicity and years of education with the validity, reliability, and bias of a self-administered food frequency questionnaire (FFQ) designed to be sensitive to low-fat, regional, and ethnic dietary patterns. Data were from the Women's Health Trial Feasibility Study in Minority Populations, a randomized clinical trial conducted between 1992 and 1994 to test the feasibility of a low-fat dietary intervention that targeted low-income, black, and Hispanic women. Of 1,015 participants eligible for these analyses, 28.1% were black, 16.2% were Hispanic, and 12.3% had not completed high school. The analyses focused on percentage of energy obtained from fat, and used 4-day food records as the criterion instrument. Validity at baseline, defined as the correlation between FFQs and food records, was lower among blacks than among whites (0.26 vs. 0.49; p < 0.001), did not differ between Hispanics and whites, and was lower among women with fewer years of education (0.19, 0.35, 0.49, and 0.42 for 16 years of education, respectively; for trend, p < 0.05). Six months after randomization, validity increased in most race/ethnicity and education subgroups, and differences across groups became small and statistically nonsignificant. Validity increased significantly among participants receiving the dietary intervention, while increases among control women were somewhat smaller. Reliability, defined as the correlation between baseline and 6-month measures among controls, was similar across racial/ethnic and educational groups. Bias at baseline, defined as the mean value from the FFQ minus the mean from the food record, was 4.6 percentage points of energy from fat; it was lowest among blacks (p < 0.01) and did not differ by years of education. These results suggest that special protocols which address participant training may be necessary when using self-administered FFQs in minority or poorly educated populations.