Child Anthropometry and Mortality In Malawi: Testing for Effect Modification by Age and Length of Follow-up and Confounding by Socioeconomic Factors

As shown in a review of 28 studies, anthropometric measurements of preschool-aged children are consistently related to the risk of subsequent mortality in community-based studies from Asia and Africa (Pelletier 1994). Although the results are consistent at this general level, a number of important questions remain concerning the relationship. The purpose of this study is to address two of these questions using data from a similar study conducted in rural northern Malawi: 1) In relation to screening, are the anthropometry-mortality relationships affected by the child's age and the interval between measurement and death ("length of follow-up") and 2) In relation to policy implications, is the anthropometry-mortality relationship due to confounding by socioeconomic factors, especially when considering mild-to-moderate malnutrition. The results reveal that mortality prediction is significantly affected by child's age and length of follow-up, but the strength and direction of this effect modification varies across the four commonly used anthropometric indicators [weight-for-age (WA), height-for-age (HA), weight-for-height (WH) and arm circumference-for-age (ACA)]. An important result for public health practice is that there are no statistically significant differences in prediction across these four indicators when applied to young children (6-23 mo) and employing a 1-y follow-up period. As regards confounding, the results indicate that the anthropometry-mortality relationship is not due to confounding by socioeconomic factors when all grades of malnutrition are considered. When only mild-to-moderate malnutrition is considered, statistically controlling for confounders reduces most of the anthropometric predictors to nonsignificance (probability values to > 0.20), but the strength of the association (odds ratio) remains of the same order of magnitude. However, when effect modification by child's age and length of follow-up is taken into account, the effect of mild-to-moderate malnutrition (WA and WH) remains statistically significant for young children dying with 1 y of follow-up.