Can household-based food consumption surveys be used to make inferences about nutrient intakes and inadequacies? A Bangladesh case study

Household Consumption and Expenditures Surveys (HCES) are increasingly being used to make inferences about individual food consumption, despite the fact that they collect food data at only the household level. Usually the analysis assumes that the household's food is distributed among its members in direct proportion to each member's share of the household's total energy requirements; what is referred to as the adult male equivalent (AME) approach. Using the 2011-2012 Bangladesh Integrated Household Survey which collected consumption data for all members of rural households using combined 24-hour recall (24HR) and food weighing methods this study directly compared probability-based average estimates of intakes, intake gaps, and the prevalence of inadequacies as estimated by the 24HR and those calculated using the AME approach for energy and each of four micronutrients: vitamin A, iron, zinc, and calcium. At the population level, for iron, vitamin A, and calcium, more than 97% of all individuals had the same estimated prevalence of adequate or inadequate nutrient intakes using 24HR and AME-based estimates. In the case of energy and zinc, roughly 77 and 83% of the sample population had identical adequacy statuses, respectively. The magnitude of inadequacies (the nutrient gaps) differed by 8 percentage points for energy and less than 3 percentage points for the four micronutrients. Disaggregating intakes and inadequacies by age groups revealed that the vast majority of variance between the two methods was highly concentrated in the first few years of life. Children 3 years of age and younger constituted 7.5% of the population but had 13.1% of the inconsistencies in intake adequacy status. 54% of children 3 and under had at least one pair of inconsistent prevalence estimates. While there are important differences in the levels of estimated energy and micronutrient intakes using 24HR and the AME-based estimates for children 3 and under, the results are remarkably comparable for the rest of the population.