Examining the impact of protracted conflicts on mortality in humanitarian emergencies : using small-scale surveys and conflict data from Yemen

Many protracted crises are largely in resource poor settings affecting billions of people around the world. Efforts are being made globally to mitigate the impact of protracted crises on affected populations. However, these places typically lack the required health information systems for the collection of birth and death records needed to conduct a comprehensive assessment of the public health status and needs of the population. In particular, Yemen has been experiencing drawn out conflict, exposing children and adults to untold hardship and diseases as well as the risk of losing their lives increases as the crisis continues. A comprehensive nationwide analysis of the impact of the protracted crisis in Yemen on mortality is lacking. However, over the course of the crisis in Yemen, series of small-scale surveys have been conducted on regional/zonal levels for planning purposes by humanitarian organizations operating on ground. Because of the costs and the impracticability of conducting an indepth, nation-wide assessment, we have attempted to conduct an effective and robust investigation for a better understanding of the impact of the conflict on mortality by combining the sparse, publicly available data. We started by assessing the methods used for the collection and presentation of mortality data obtained from the small-scale survey. We found no strict adherence to standardized methodology guidelines, and reporting of mortality and sample size data. Adherence to methodological guidelines and complete reporting of surveys in humanitarian settings will vastly improve the estimation of mortality rate and also uptake of key data on health indicators of the affected population. We also assessed the change in number of deaths as a result of the conflict using a Bayesian mixture model approach. Despite an increase in the estimated number of deaths attributed to the conflict, we observed uncertainty surrounding the point estimates, which calls for caution while interpreting the result from our data. The difference observed could range from fewer deaths, or more deaths during the conflict when compared to period before the conflict. In addition, we developed a simple and parsimonious classification of insecurity level using publicly available data. We examined the relationship between patterns in under-five mortality and level of insecurity using a Bayesian finite mixture model, and we found that high number of under-five deaths are clustered around regions/zones experiencing high levels of insecurity.(SP - Sciences de la santé publique) -- UCL, 202