Validation of a food frequency questionnaire for Hispanics

The Hispanic population will grow to comprise one fourth of the U.S. population by 2050. Compared with non-Hispanic whites, Hispanics have disproportionately higher rates of obesity, diabetes, and other diet-related conditions. Valid methods for studying the dietary intake of this group are needed.

From June through September 2000, we conducted a study of low-income Hispanic men and women (n = 89) who were recruited for a validation study of the Spanish-language food frequency questionnaire used in the Study of Women's Health Across the Nation. The mean age of the participants was 36.8 years, 42% were male, and 92% had been born in Mexico. Three 24-hour dietary recalls provided the reference data. The food frequency questionnaire was administered by interview, with a portion-size graphic to aid in quantitation. The questionnaire asked about diet in the previous 12 months. Mean nutrient values, correlation coefficients, and the sensitivity and specificity for identifying people with intakes of less than the recommended levels were calculated.

Mean energy and macronutrient intake estimates were significantly higher by the food frequency questionnaire than by the 24-hour dietary recalls. Cholesterol, saturated fat, dietary fiber, iron, vitamin A, and percentage of energy from fat were not significantly different by the two methods. The median of unadjusted correlations was 0.52 and of deattenuated correlations was 0.61. The median sensitivity was 0.62, and the median specificity was 0.76.

The Study of Women's Health Across the Nation Spanish food frequency questionnaire appears to be reasonably valid in assessing the dietary intakes of Hispanics. Correlations tended to be higher than those found in other validation studies in Hispanic populations. Interviewer administration of questionnaires may be necessary in this population.