Abstract
A food frequency questionnaire was designed to assess the usual nutrient intake of New Zealand adults. National diet survey data was used as a basis for compiling the food list.
The optimal design and analysis of the 132 item food frequency questionnaire was determined by a series of analyses using diet intake data from a sample of 101 female undergraduate Nutrition students (mean age 21 years).
The performance of the food frequency questionnaire was assessed by comparison with seven-day diet records. The nutrients used in this comparison were ones of special consideration for New Zealanders according to the Nutrition Taskforce (1991), i.e carbohydrate, starch, sugars, fat, saturated fat, polyunsaturated fat, protein, dietary fibre, calcium, iron, zinc, thiamin and ascorbic acid.
The comparison methods used were mean difference, standard deviation of the difference, classification into the same or adjacent quintile, gross misclassification, correlation coefficients and actual values for surrogate categories.
There is a need for further investigation into the sensitivity of the 'classification into the same of adjacent quintile' method for assessing the ability of the food frequency questionnaire to classify subjects. The use of a statistical test to assess the significance in changes in classification is necessary. The use of more than five categories, e.g. 7 or 9, requires consideration for a more sensitive measure of the food frequency questionnaire's ability to classify individuals.
General principles were formulated according to the effect of the design and analysis issues on the food frequency questionnaire's performance.
The food frequency questionnaire's ability to estimate the group mean intake was enhanced through use of a categorical response scale format (rather than an open-ended scale), age and sex specific serving sizes (rather than common standard measures) and adjustments for the total servings of fruit, vegetables and meat.
The food frequency questionnaire's ability to classify individuals was enhanced through use of a categorical response scale format (rather than an open-ended scale), use of common standard measures as serving sizes, and asking the respondent to specify small, medium or large serving sizes.