Empirical vulnerability strata in under-five mortality in Cameroon: Evidence from DHS data

Despite substantial declines in under-five mortality in sub-Saharan Africa, child survival remains marked by persistent and socially structured inequalities. Variable-centered survival models document average effects of risk factors but provide limited insight into how multiple disadvantages combine to generate heterogeneous mortality risks. This study aims to identify and characterize empirical configurations of vulnerability in under-five mortality in Cameroon. Using pooled data from the 2011 and 2018 Cameroon Demographic and Health Surveys, we analyze time to death before age 59 months. A random survival forest is used as an exploratory survival-based tool to derive individual risk scores, which serve to stratify children into ordered mortality risk strata and then summarized using weighted Kaplan-Meier estimates. Three distinct mortality risk strata emerge. The highest-risk stratum, with an estimated probability of death before age five of 27.8%, concentrates overlapping disadvantages, including low birth weight, short birth intervals, limited maternal and partner education, lack of antenatal care, home delivery, low media exposure, and constrained maternal decision-making autonomy. These configurations are observed consistently across survey waves, indicating persistent social structuring of mortality risk. Under-five mortality in Cameroon remains sharply stratified along cumulative social, reproductive, and gender-related dimensions. Inequalities in child survival are produced not by isolated risk factors but by socially embedded configurations of disadvantage. This study contributes to demonstrate how survival-based machine learning can be combined with conventional statistical tools to reveal socially interpretable strata of vulnerability, shifting attention from marginal covariate effects to empirical configurations of cumulative mortality risk.