Abstract
In regions burdened by significant disease and constrained resources, such as sub-Saharan Africa, identifying communities with particularly atypical public health outcomes can enhance the optimisation of available resources during interventions. It is essential to understand the specific characteristics of individuals contributing to these unusual health outcomes within the outlier communities to determine the most effective interventions. While diagnostic statistics have been developed to detect grouped outliers in clustered survival data, there is scarcity of research focusing on the contribution of individual subjects to these outlier groups, particularly concerning binary outcome data. This paper adapts diagnostic statistics developed for clustered time-to-event data within mixed-effects logistic regression to identify outlier villages with elevated child mortality rates in Malawi and to analyse their characteristics. The findings indicate that nine villages exhibited child mortality rates that were at least four times higher than the national average, mostly located in the rural southern and central regions of the country. In each of these outlier villages, the study identified children who died despite possessing a low predicted probability of death according to the model. This research demonstrates how residuals from hierarchical survival models can be utilised to connect higher-level and individual outliers within a mixed-effects logistic regression framework, allowing for a comprehensive analysis of unusual binary outcome data at both the community and individual levels.