Modeling excess deaths after a natural disaster with application to Hurricane Maria.

Estimation of excess deaths due to a natural disaster is an important public health problem. The CDC provides guidelines to fill death certificates to help determine the death toll of such events. But, even when followed by medical examiners, the guidelines cannot guarantee a precise calculation of excess deaths. We propose two models to estimate excess deaths due to an emergency. The first model is simple, permitting excess death estimation with little data through a profile likelihood method. The second model is more flexible, incorporating temporal variation, covariates, and possible population displacement while allowing inference on how the emergency's effect changes with time. The models are implemented to build confidence intervals estimating Hurricane Maria's death toll.