Real-time calibrated models could have sharpened pandemic decision support in England

Infectious disease models predict how outbreaks spread and affect populations, but during the COVID-19 pandemic in the United Kingdom, formal uncertainty quantification techniques were not routinely used to inform policymakers, according to a new paper published in PLOS Computational Biology.

The authors, including Daniel Williamson, Trevelyan J. McKinley, Xiaoyu Xiong, James M. Salter, Robert Challen, Leon Danon, Ben Youngman and Doug McNeall, argue that in a fast-moving pandemic, models and policy change on timescales that make traditional uncertainty quantification methods impractical, if not impossible, to implement. These methods require calibrating model parameters and accounting for structural discrepancies between the model and the reality it simulates, a process that can be too slow when decisions must be made quickly.

To address this, the researchers present an alternative formulation of the calibration problem that embeds model discrepancy directly within the structure of the model and appropriately assimilates data within the simulation. This restructuring allows uncertainty quantification to be used to calibrate the model in real time, producing disease trajectories that account for both parameter uncertainty and model discrepancy.

The team applied these ideas to an age-structured COVID-19 model for England and demonstrated the types of information the approach could have produced to feed into policy support prior to the lockdown of March 2020.

By showing how calibrated predictions can be generated under the time constraints of an unfolding epidemic, the work outlines a practical route for embedding uncertainty quantification into operational pandemic modelling, where the speed of policy decisions has historically outpaced the pace of traditional calibration workflows.

Subject of Research: Real-time uncertainty quantification and calibration of infectious disease models for pandemic decision support

Article Title: On real-time calibrated prediction for complex model-based decision support in pandemics: Part 1

Article References: Williamson, D., McKinley, T. J., Xiong, X., Salter, J. M., Challen, R., Danon, L., Youngman, B., & McNeall, D. (2026). On real-time calibrated prediction for complex model-based decision support in pandemics: Part 1. PLOS Computational Biology, 22(10), e1014868. https://doi.org/10.1371/journal.pcbi.1014868

Image Credits: AI Generated

DOI: 10.1371/journal.pcbi.1014868

Keywords: uncertainty quantification, infectious disease modelling, COVID-19, model calibration, pandemic policy, model discrepancy, age-structured model, England, real-time prediction, decision support, PLOS Computational Biology, real-time