From disparate lists to population estimates: A multiple systems estimation workflow for mortality analysis in conflict settings

Quantifying mortality in conflict settings is a challenging but necessary task for health and humanitarian response, historical memory, and accountability. In conflict settings, traditional demographic data sources, like vital statistics, may not be available or of suitable quality for mortality estimation. However, alternative sources documenting the deceased may exist, such as digital trace data, news reports, or data collected by civil society organizations. These sources are not without their own biases. Rather than being statistically representative samples or complete counts, they are convenience samples: they represent the fraction of violence that was documented. However, when multiple sources are taken together, multiple systems estimation (capture-recapture) can be used for mortality estimation. This article describes a workflow for mortality estimation using disparate and statistically biased data sources, motivated by three missing data challenges commonly encountered in conflict settings: missing identifiers to link victims across sources, missing covariates in records of documented victims, and records of victims that are entirely undocumented. The workflow is demonstrated using data about conflict-related homicides during the armed conflict in Colombia from 1985–2018. Beyond its applications to conflict mortality estimation, this workflow can be applied to examine mortality due to other causes or other events of demographic interest.