Human insecurity and child deaths in conflict: evidence for improved response in Yemen

Background Since the beginning of the ongoing conflict in Yemen, >23 000 air strikes and >100 000 fatalities have been recorded. Data from Yemen Data Project linked >1300 child fatalities and >900 child injuries to air raids. However, there is little literature on the effect of the protracted armed conflict on the pattern of child mortality using data from small-scale surveys. We aimed to identify the pattern of the death rate for children aged <5 years ('under-5') and its relationship with human insecurity in Yemen. Methods We created a human insecurity index (i.e. severely insecure vs insecure) for the 22 governorates in Yemen from 2015 to 2019, using data from the Armed Conflict Location and Event Database. We matched this insecurity index with the corresponding under-5 mortality data from the Complex Emergency Database. We analysed the relationship between the under-5 death rate (U5DR) and the insecurity level using a Bayesian finite mixture model in order to account for unobserved heterogeneity in clustered finite subsets of a population. Results We extracted 72 surveys and 77.8% (n = 56) were included in this study. The mean of the recall period for mortality was 106 days with a standard deviation of 93 days. We identified two subpopulations: Subpopulation I-high average number of child deaths and Subpopulation II-low average number of child deaths. The log posterior mean of the U5DR is 1.10 (95% credible intervals: 0.36, 1.82) in the severely insecure group in Subpopulation I and 3-fold the estimate in Subpopulation II. However, in Subpopulation II, we found no association between the insecurity level and the U5DR. Conclusion The pattern of child deaths is crucial in understanding the relationship between human insecurity and the U5DR.