Country-Specific Modeling of Verbal Autopsy Misclassification for Improved Child and Neonatal Cause-Specific Mortality Estimates at Low- and High-Resolution Causes - The primary objective of this K99/R00 application is to develop Dr. Sandipan Pramanik’s (PI) research skills to improve our understanding of disease burden among children under age 5 in low and middle-income countries (LMICs). The K99 phase of the proposed research will support PI in accomplishing four training objectives. This will enable his seamless transition into an independent global health investigator, enhancing mortality surveillance through impactful solutions that integrate multi-source data. First, he will broaden his understanding of factors impacting child and neonatal mortality in LMICs, along with foundational statistical and demographic surveillance methods. Second, by conducting field visits, he will gain firsthand experience and insights into on-site data collection, method/software implementation, and associated challenges. He will also arrange direct studies to acquire knowledge on different causes of death (CODs) diagnoses (such MITS, verbal autopsy (VA)), and their medical background. Third, by visiting collaborators, he will enhance his expertise through research presentations and discussing competing/complementary methods. Fourth, he will seek professional development opportunities, emphasizing grant writing and collaboration. Based on the training objectives and utilizing unique access to essential rich data sources (CHAMPS, COMSA), this proposal is timely and pivotal for enhancing mortality estimates, focusing on children under age 5. With the acquired skills from training, the PI will be well-positioned to successfully achieve the three research aims, expanding VA-calibration applicability. Utilizing paired MITS-VA COD data and tackling systematic bias and heterogeneity, Aim 1 will develop a concise country-specific VA misclassification modeling framework, enhancing mortality estimation covering detailed causes. In a first, Aim 2 will design efficient Bayesian tests to effectively compare hypotheses about different effects in VA misclassification. In contrast to classical tests, this will facilitate more efficient and clear statistical decision-making. Using national-level estimates, Aim 3 will create a novel statistical downscaling framework. This will enhance inference for sub-populations at the sub- national level, including age groups, sexes, and geographic regions, ensuring coherence with population-level estimates. Complementing Aims 1–2, it will also distribute open-source software and apply proposal outputs to various projects and grants, with immediately improved mortality estimates in Mozambique and Sierra Leone. This proposal will accelerate the worldwide effort to improve child mortality surveillance. It also aligns with the US government and international community’s objective to assess the effectiveness of age and disease- specific interventions, such as malaria vaccine2 or azithromycin3, in combatting the primary causes of mortality among them. The collaborative efforts between the Departments of Biostatistics and International Health at Johns Hopkins, involved in initiatives like COMSA in Mozambique and CHAMPS, offer an ideal environment to receive the research support and resources to attain his training and research objectives.
