International congress of nutrition

The aim of the study was to validate the 24-hour recall as a tool for nutrition surveys in Algeria. The reference was the weighing method. The study took place in a rural district of North-Eastern Algeria. The subjects were 50 women of child-bearing age, chosen at random in the district population. The methods, recall and reference, were used to measure food intake over the same two days. We computed differences in paired observations (recall-reference) to detect a level-dependent bias, and a constant bias when the former was absent, and we assessed the precision of the method. We detected a level-dependent bias for lipid, vitamin A and iron intakes. High lipid intake was markedly underestimated with the recall. For vitamin A and iron, the bias was of small magnitude ; it was symmetrical for iron but not for vitamin A. There was no constant bias for energy, protein and carbohydrate intakes. Moreover, for these nutrients, two thirds or more of recall estimates were within plus or minus 20% of the reference. The small differential bias in vitamin A intake disappeared once omissions and additions of foods were removed. Bias in the 3rd tertile of lipid intake was mainly due to an underestimation of portion size in fat-rich dishes, and to lesser extent to errors in the amount of fat in recipes. The 24-hour recall is valid for measuring rural Algerian women's energy, protein and iron intakes but needs to be improved for estimating vitamin A and lipid intake. (Résumé d'auteur