Comparison between household budget survey and 24-hour recall data in a nationally representative sample of Polish households.

OBJECTIVE Household budget survey (HBS) data are used regularly for nutritional epidemiological purposes. The validity of HBS data, however, is not well established. The aim of this project was to compare HBS and individual nutrition survey (INS) data in a nationally representative sample of Polish households. DESIGN Estimates of food consumption and nutrient intake were compared between household food acquisition data collected over 1 month and a single 24-hour recall collected from every household member in a nationally representative sample of Polish households surveyed between September and November 2000. To facilitate the comparison, INS food consumption data excluded food eaten away from home and were modified using a computer program to estimate food 'as purchased' (including disaggregation of recipe data) and to allow for wastage. SETTING Poland. SUBJECTS Participants were 3716 individuals in 1215 households (representing co-operation rates of 86.2% and 89.2%, respectively). RESULTS Good agreement was shown between median estimates of foods such as potatoes, vegetables (including processed), meat, meat products and poultry, and animal fats (excluding butter), but agreement was poor for bread and rolls, fruit, vegetable fats and oils, eggs and six other food groups. Estimates of energy and nutrient intake were within +/-10% with the exceptions of polyunsaturated fats, potassium and vitamin C. CONCLUSIONS Possible reasons for differences in findings between the two surveys include survey bias (e.g. social approval bias leading to overreporting of fruit), seasonal variations (e.g. high potato purchases between September and November) and aspects of the methodology (e.g. HBS data were based on records collected over 1 month, whereas 24-hour recall data were based on recalls collected from all household respondents on only 1 day and averaged for each household type). HBSs provide useful data for epidemiological research, but findings need to be interpreted in the light of other data regarding consumption, and numerous factors that may affect consumption need to be taken into account.