The use of national sample surveys for nutritional surveillance: Lessons from Malawi's national sample survey of agriculture

Abstract

National sample surveys, containing measurements of the weights and heights of children along with other socioeconomic modules, represent one of the important sources of information for nutritional surveillance. The potential uses of such information are political sensitization, targeting by geographic area, targeting by socioeconomic group, and identifying the most promising intervention options according to the presumed causes of malnutrition. The latter two applications depend upon the ability to detect stable associations between nutritional status and socioeconomic factors. This paper examines the extent to which these planning applications are affected by variation in the ecology of malnutrition across different segments of society, using Malawi's National Sample Survey of Agriculture (NSSA). The NSSA is nationally representative of Malawi's smallholder sector and contains information on anthropometrics of underfives and various socioeconomic characteristics of their households. This analysis is based upon 3000 households containing at least one underfive, with one child per household being selected for analysis. Height-for-age Z-scores (HAZ) are calculated using WHO standards, and analyzed using analysis of variance (ANOVA). Variation in nutritional ecology is investigated by examining statistical interactions among socioeconomic variables as they relate to HAZ and by comparing ANOVA models constructed within several sample strata. These strata are defined according to geographic region, size of cultivated area, child's sex, child's age and, for a subsample, erhnic/religious identification. The results demonstrate that significant two-way interactions exist between region, cultivated area, household labor availability and age of the child. For the total sample HAZ declines with increasing cultivated area among young children (<24 months) but improves with increasing cultivated area among older children (⪖ 24 months). the overall trend for young children is region-specific, however, such that no generalizations can be made concerning the effects of region and cultivated area without taking the other factor and child's age into account. Similarly, cultivated area interacts with household labor availability among young children, such that more labor is associated with lower HAZ on plots less than 0.7 ha, but more labor is associated with higher HAZ on plots above 1.5 ha. The existence of variation in nutritional ecology is further borne out by the observed variation in multivariabe ANOVA models constructed within various sample strata. It is concluded that significant variation in nutritional ecology does exist between various segments of society. This limits the usefulness of national sample surveys for selecting intervention options and for targeting interventions according to socioeconomic characteristics.