A Hierarchical Bayesian Model for Estimating Age-Specific COVID-19 Infection Fatality Rates in Developing Countries

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Accurate COVID-19 infection fatality rate (IFR) estimates, particularly age-specific rates, are crucial for understanding disease impact and resource allocation.
• Existing methods struggle to synthesize age-stratified seroprevalence and death data, especially with inherent uncertainties from sampling and imperfect diagnostics.
• Heterogeneity in population age structures across locations necessitates tailored IFR estimations.

Purpose of the Study:

• To introduce a novel Bayesian hierarchical model for estimating age-specific IFR and seroprevalence.
• To develop a method that continuously models IFR as a function of age.
• To account for uncertainties arising from seroprevalence sampling variability and imperfect serology tests.

Main Methods:

• Developed a Bayesian hierarchical model to estimate IFR as a continuous function of age.
• Simultaneously modeled test assay characteristics, serology data, and death data, often available in age-binned formats.
• Utilized hierarchical modeling to share information across 26 developing country locations, improving estimates with limited data.

Main Results:

• Seroprevalence showed minimal variation across age groups in the studied locations.
• The infection fatality rate (IFR) at age 60 exceeded high-income country estimates in most analyzed developing countries.
• The model successfully integrated heterogeneous data sources and reflected inherent uncertainties.

Conclusions:

• The novel Bayesian model provides a robust framework for estimating age-specific IFR, accounting for data limitations and population structures.
• Findings highlight a potentially higher burden of severe COVID-19 outcomes in older populations in developing countries.
• The approach facilitates more accurate comparisons of COVID-19 impact across diverse global settings.