Measuring children's diets: evaluation of dietary assessment techniques in infancy and childhood.

Epidemiological studies of dietary habits in infancy and childhood face a number of difficulties, which are more or less specific for these ages. In connection with studies on dietary habits of Swedish infants and children an evaluation of different dietary assessment techniques was performed. Breastfeeding data obtained in retrospective interviews at six months had good validity. The reliability of breastfeeding and weaning data decreased over time. Short questions on food frequencies, often used in research and clinical practice, were shown to be a poor screening instrument and suffered from biases when used in groups of four- and eight-year-old children. Group mean estimations of dietary intake of four- and eight-year-old children obtained by 24-hour recalls were close to those of seven-day records from the same individuals. Dietary intake in a 13-year-old group according to 24-hour recall and dietary history differed significantly; the dietary history gave much higher estimations. The reliability of dietary history in a small group of 13-year-old boarding school children was fairly good. The internal validity of food recording was examined by use of chemical analysis of duplicate portions. Generally, there was good agreement between the records and the analysed duplicates. The intake of a number of nutrients varied during the week, often showing a maximum towards the weekend. The intake of vitamin C was higher in winter time, otherwise no seasonal variation was found. The intra-individual variation in dietary intake is illustrated and its consequences for dietary studies are discussed.