Dynamic calibration with approximate Bayesian computation for a microsimulation of disease spread.

Area of Science:

• Epidemiology
• Computational Biology
• Public Health

Background:

• The COVID-19 pandemic highlighted the critical role of infectious disease modeling in informing public health policy.
• Quantifying uncertainty in model predictions remains a significant challenge for effective policy development.
• Real-time data integration is crucial for improving model accuracy and reducing prediction uncertainty.

Purpose of the Study:

• To adapt an existing individual-based COVID-19 model for pseudo-real-time updates.
• To explore the benefits of dynamically recalibrating model parameters with emerging data.
• To assess the impact of updated data on forecast accuracy and uncertainty.

Main Methods:

• Utilized Approximate Bayesian Computation (ABC) for dynamic model recalibration.
• Incorporated pseudo-real-time data streams into a large-scale, individual-based COVID-19 model.
• Analyzed posterior distributions to understand parameter and prediction uncertainty.

Main Results:

• Forecasts of future COVID-19 infection rates were substantially improved with up-to-date observations.
• Uncertainty in model predictions decreased significantly in later simulation windows as more data became available.
• ABC provided valuable insights into parameter uncertainty and its effect on model outputs.

Conclusions:

• Dynamically updating infectious disease models with recent data enhances forecast precision.
• Reduced uncertainty in model predictions is achievable through continuous data assimilation.
• This methodology offers a more robust framework for utilizing infectious disease models in policy-making.