The implications of underreporting in dietary studies

Dietary data from the Western Sydney Dietary Survey 1989-90 (n = 512) was used to investigate: 1. the prevalence and predictors of underreporting of energy intake, 2. the effects on results of excluding data from underreporters for analysis of mean nutrient intakes, and 3. the proportion of energy intake supplied by macronutrients and proportions of subjects who met dietary goals. The proportion whose measured energy intakes from a Food Frequency Questionnaire (FFQ) were below cutpoints for biologic plausibility was 28.5 per cent; it was higher for subjects who had BMI > 25 and were female. Point estimates for mean intakes of energy and nutrients were all greater when data from underreporters were excluded, but nutrient intakes expressed as percentages of energy intake remained largely unchanged. Increases in estimated mean population intake for each nutrient ranged from 7 per cent to 14 per cent for males, and 12 per cent to 17 per cent for females. Estimates of the percentages of the sample who did not meet dietary goals were significantly lower for a number of nutrients when underreporters were excluded. We conclude that: 1. results expressed as a percentage of energy intake are not affected by the exclusion of energy underreporters, and 2. estimates of the proportion of populations meeting some nutrient goals and associations between diet and disease are likely to change meaningfully and significantly with the exclusion of data from underreporters.