Moving Beyond Model Life Table Systems: A New Method for Direct Estimation of Age-Specific Mortality

Austin E. Schumacher, University Of Washington
Christopher J. L. Murray, University of Washington
Aleksandr Aravkin, University of Washington
Peng Zheng, University of Washington

To enhance the estimation of age-specific mortality rates, particularly in low and middle-income countries (LMICs) lacking comprehensive civil registration and vital statistics, we introduce a novel Bayesian hierarchical modeling approach. This model directly models age-specific mortality from diverse data sources such as vital registration systems, sample registration systems, complete birth histories, summary birth histories, sibling survival histories, and household death recall. Traditional methods, reliant on model life table systems, often misrepresent mortality patterns due to their generalization from disparate data contexts, inadequately reflecting local health realities such as lower old age mortality rates in sub-Saharan Africa. We propose a method for directly modeling age-specific mortality through correlated splines and kernel regression for residual smoothing. Preliminary comparisons against previous Global Burden of Disease (GBD) estimates and United Nations World Population Prospects (WPP) demonstrate our model's superior performance, particularly in locations with previously sparse data. This methodology not only offers a more precise and flexible tool for demographic and public health research but also holds significant potential for informing effective health interventions and policy-making by providing a clearer insight into mortality patterns across age groups and locations over time.

Keywords: Data and Methods, Mortality and Longevity, Multi-level modeling , Bayesian methods

See extended abstract.