Estimating the COVID-19 infection fatality ratio accounting for seroreversion using statistical modelling

Area of Science:

• Epidemiology
• Biostatistics
• Infectious Disease Modeling

Background:

• The infection fatality ratio (IFR) is critical for assessing the burden of coronavirus disease 2019 (COVID-19).
• Previous IFR estimates often failed to account for uncertainties and biases, such as antibody waning (seroreversion).
• Accurate IFR quantification requires integrating time delays (infection to seroconversion, death, seroreversion) and test performance characteristics.

Purpose of the Study:

• To develop a robust statistical framework for estimating age-specific IFR by addressing uncertainties and biases.
• To incorporate factors like seroreversion, time delays, and test performance into IFR calculations.
• To provide a reproducible method for calculating IFR in different populations and epidemic waves.

Main Methods:

• A Bayesian statistical model was developed to integrate multiple factors influencing IFR.
• The model was applied to simulated data and ten international serologic studies.
• Consideration was given to time from infection to seroconversion, time to death, time to seroreversion, and serologic test accuracy.

Main Results:

• Seroreversion significantly impacts IFR estimates over longer time periods but is less critical in short-term, first-wave dynamics.
• Disaggregating serologic surveys by regional disease burden can refine estimates of serologic test specificity.
• The overall IFR estimates across different settings ranged from 0.49% to 2.53%.

Conclusions:

• A robust statistical framework was established to manage uncertainties in IFR parameters.
• The developed methods and code are available for application to new datasets and future epidemics.
• This approach enhances the accuracy of IFR estimation, crucial for public health response.