Data Integration Methods to Improve Small Area Mortality Estimation for Tribal and Rural Populations - Tribal Nations often lack accurate, granular, and comprehensive information on the true impact of mortality indicators across American Indian/Alaska Native (AI/AN) population groups needed for different use cases. Misclassification error in state death certificates is disproportionately 30% higher for the AI/AN population compared to other demographic groups due to higher proportions of persons with a range of backgrounds within AI/AN populations. Additionally, US census reported AI/AN population statistics suffer from severe under-reporting at every geographic resolution due to small sample sizes, challenges in data collection, cost, and privacy algorithms. Importantly, official mortality and population statistics do not comprehensively or accurately enumerate distinct AI/AN populations defined by (1) Self- reported AI/AN based on U.S. Census defined classification (census category), (2) Tribal citizenship with a Tribal Nation (legal category), and (3) Residence on Tribal lands (geographic category). To address this data gap, our approach combines Bayesian hierarchical small area disease mapping with measurement error methodologies to produce small area mortality estimates across AI/AN groups for both rural and non-rural areas with improved accuracy. Within the comprehensive methodological framework, model components are customized to produce population-specific mortality estimates with increased accuracy across multiple populations. We apply our methodological approach to obtain county-year trends in AI/AN opioid mortality rates (OMRs) for the 77 counties in Oklahoma between years 2015-2024 across the distinct AI/AN population groups and county-specific rural status. In Phase 1, we estimate county-year OMRs for Tribal citizens by linking data from Oklahoma state death certificates with Cherokee Nation Tribal Registry data. In Phase 2, we estimate county-year OMRs for self-affiliated AI/AN populations using misclassification rates from Phase 1 and assess trends for total populations-at-risk in counties inside and outside Tribal boundaries. Of utmost importance to the study is the independent application and use of the proposed methodology by local health departments. We will develop a data science pipeline consisting of both educational and computational resources for local health analysts to adopt and implement the proposed methodology as a mortality self-monitoring resource.
