The reproducibility and validity of a self-administered semiquantitative food frequency questionnaire in subjects from South Dakota and Wyoming.

Most validation studies of food frequency questionnaires have been done in populations capable of reporting their diet with unusual accuracy. In this study of ranchers and subjects selected at random, we compared nutrient intakes as assessed with multiple diet records with those assessed with a self-administered food frequency questionnaire (the Harvard-Willett instrument with 116 food items). One hundred thirty-eight subjects from South Dakota and Wyoming, 64 males and 74 females, completed at least one food frequency questionnaire and multiple 1-day diet records (mean = 5 days) during a 6-month to 1-year period. Of invited subjects, 88% participated; 59% of participants had a high school education or less. For 22 dietary nutrients (excluding supplements) and alcohol, the median crude correlation between intakes based on diet records and the food frequency questionnaire completed before the diet records was 0.42; after adjustment for energy, age, and sex, and after correction for attenuation in the correlation coefficients due to error from a limited number of diet records, the median correlation was 0.52. Correlations for men and women were similar. The validity of the food frequency questionnaire in this population was similar to that reported previously and supports the use of self-administered food frequency questionnaires in studies of general populations.