Rapid Nutrition Surveys: How Many Clusters are Enough?

On the basis of theoretical considerations, population-based nutrition surveys of 30 clusters of 30 children should provide reasonably valid estimates of the prevalence of malnutrition with at least 95 per cent confidence that the estimated prevalence differs from the true value by no more than 5 per cent. In areas of famine in Africa, where an urgent need often exists for rapid nutritional assessment to determine the extent and severity of the problem, visiting 30 sites is often logistically difficult. To determine the effects of using fewer than 30 clusters on the validity and precision of the estimated level of undernutrition, we used data from the 1983 Swaziland National Nutrition Survey and from rapid nutrition surveys performed in 1984 and 1985 in Burkina Faso, Guinea, and Niger. Fewer than 30 clusters may result in point prevalence estimates that differ dramatically from the true prevalence and, in most instances, are less precise. In contrast, little is gained by collecting more than 30 clusters. In summary, around 30 clusters provides relatively valid and precise estimates of the prevalence of undernutrition, and every effort should be made to obtain the logistic support required to study this number of clusters.