Abstract

Although complex cluster surveys, such as the Demographic and Health Survey, are commonly used to study under-five mortality in sub-Saharan Africa, there has been little research using the data structures to study the connections between individual- and community-level unusual mortality outcomes in the region. This chapter applied univariate and multivariate diagnostic statistics for logistic regression to examine relationships between outlier children and outlier villages to under-five mortality outcomes in Malawi. The study examined a nationally representative random sample of 17,286 children in 850 villages in Malawi using data from the 2015–2016 Malawi Demographic and Health Survey. The analyses were conducted using R software version 4.3.0. The results revealed that 9 villages had outlier under-five mortality rates that were at least four times the national average rate. All these villages were located in rural areas in the southern and central regions of the country. It was also noted that there was a direct link between individual-level outlier children and outlier villages in terms of the under-five mortality rate. In each of the outlier villages identified, outlier children were also detected at the individual level. These outlier children were found to have died despite having a low predicted probability of dying according to the mixed model. The findings suggest that using diagnostic statistics for multivariate models can help identify communities with unusual child mortality outcomes and the characteristics of the children contributing to this. This information could be valuable in planning appropriate interventions to reduce child mortality in the identified high-burden communities within the population.

Original language	en
Pages (from-to)	119-136
Publication status	Published - 2025