Performance of a short dietary questionnaire to assess nutrient intake using regression-based weights

Abstract

To evaluate the performance of a short dietary questionnaire, using weights to estimate nutrient intake. Using dietary data collected in 1993-1995 from a large cohort of French women, stepwise regression analysis was used to identify the food groups that best predicted nutrient intakes, resulting in a short list of twenty-three foods. This list was used to design a twenty-three-item dietary questionnaire. Nutrient intake was estimated from the answers to the twenty-three questions, applying weights to each response. Weights were calculated from the large database as regression coefficients of the nutrient intake against the twenty-three food groups. In 2005-2006, 103 women responded (at a 1-year interval) to both the short questionnaire and a previously validated dietary history questionnaire. Intakes of twenty nutrients and energy estimated from these two questionnaires were compared. French adult female population. For developing the instrument, 73 034 women aged 41-72 years; for testing, 103 women aged 55-80 years in 2005. Mean nutrient intakes generally differed by less than 10 % between the two methods. Correlation coefficients of nutrient intakes ranged from 0.23 for vitamin D to >0.65 for Mg, vitamin B3 and alcohol. For most nutrients, at least 70 % of subjects fell into the same or an adjacent quintile when classified by either of the two questionnaires. In light of both its strengths and limitations, this short questionnaire could be used in French adult women to obtain some general nutritional information, notably for adjustment purposes when response to an extensive questionnaire cannot be obtained.