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

Real-time calibration of an age-structured COVID-19 model for England embeds model discrepancy in the model structure and assimilates data to generate disease trajectories accounting for parameter uncertainty.

• Why it matters: Traditional uncertainty quantification was not routinely used to inform UK policymakers during COVID-19, and fast-changing conditions make conventional UQ impractical.
• What they did: The paper reformulates calibration by embedding model discrepancy within the model and assimilating data during simulation, then uses uncertainty quantification to calibrate in real time.
• The result: Applied to England’s age-structured COVID-19 model, the approach could have produced policy-relevant information prior to the March 2020 lockdown.