Deriving Small Area Mortality Estimates Using a Probabilistic Reweighting Method

Small area health estimates are important for studying environmental exposure, disease transmission, and health outcomes at the local scale. Yet, to protect privacy, the majority of publicly available health data are aggregated within larger spatial units such as states or counties. This article describes a method to generate small area mortality estimates from individual microdata that are available only for larger geographic entities. The mortality estimates are based on the probabilistic reweighting and spatial allocation of a population constructed by combining the individual-level microdata with census tract–level summary data. The generated mortality counts can be used to explore local mortality patterns and identify clusters of mortality from various causes. Validation of the allocated death counts against actual restricted-use census tract–level death counts suggests that the estimated counts reliably duplicate the total mortality patterns found in the actual data. The allocations of cause-specific mortality outcomes are less accurate, however.