Assessments of food-frequency questionnaires in minority populations.

To assess the performance of food-frequency questionnaires in minority populations, we reviewed published studies of such questionnaires with respect to reports of reliability, bias in mean intake estimates, validity, and factors that might affect these characteristics, including the method by which a questionnaire was developed. Because of the small number of studies and the variation in their methods, the results of the review are not easily summarized and cannot be generalized. Most reliability correlation coefficients were in the range usually reported in evaluations of the performance of questionnaires in nonminority populations (0.5-0.8). Mean intakes derived from questionnaires were generally higher than those estimated with use of referent methods, although this varied according to the population, questionnaire, and nutrient being assessed; moreover, the accuracy of the referent method was questioned by some investigators. Validation correlation coefficients were variable, with some indicating that the questionnaires performed less well in certain populations. Additional research is needed on the performance of food-frequency questionnaires in minority and multicultural populations and on determining how best to develop such questionnaires.