Predicting dietary intakes with simple food recall information: a case study from rural Mozambique.

Abstract

Improving dietary status is an important development objective, but monitoring of progress in this area can be too costly for many low-income countries. This paper demonstrates a simple, inexpensive technique for monitoring household diets in Mozambique. Secondary analysis of data from an intensive field survey on household food consumption and agricultural practices, known as the Nampula/Cabo Delgado Study (NCD). In total, 388 households in 16 villages from a stratified random sample of rural areas in Nampula and Cabo Delgado provinces in northern Mozambique. The NCD employed a quantitative 24-h food recall on two nonconsecutive days in each of the three different seasons. A dietary intake prediction model was developed with linear regression techniques based on NCD nutrient intake data and easy-to-collect variables, such as food group consumption and household size The model was used to predict the prevalence of low intakes among subsamples from the field study using only easy-to-collect variables. Using empirical data for the harvest season from the original NCD study, 40% of the observations on households had low-energy intakes, whereas rates of low intake for protein, vitamin A, and iron, were 14, 94, and 39, respectively. The model developed here predicted that 42% would have low-energy intakes and that 12, 93, and 35% would have low-protein, vitamin A, and iron intakes, respectively. Similarly, close predictions were found using an aggregate index of overall diet quality. This work demonstrates the potential for using low-cost methods for monitoring dietary intake in Mozambique.