Bayesian sequential data assimilation for COVID-19 forecasting.

Area of Science:

• Epidemiology
• Computational Biology
• Bayesian Statistics

Background:

• Epidemic forecasting is challenged by non-autonomous dynamical systems influenced by complex factors like human behavior and viral evolution.
• Accurate, long-term epidemic prediction remains difficult due to these inherent system complexities.

Purpose of the Study:

• To introduce and validate a Bayesian sequential data assimilation and forecasting method for non-autonomous dynamical systems.
• To apply this novel method to forecast the ongoing COVID-19 pandemic.

Main Methods:

• Developed a Bayesian sequential data assimilation framework integrating transmission, epidemic, and observation models.
• Dynamical systems were employed to represent epidemic spread, with observation models defining likelihood functions.
• Forecasts were sequentially updated using a sliding window of epidemic data and calibrated prior distributions.

Main Results:

• Demonstrated the forecasting method's performance using a SEIR (Susceptible-Exposed-Infectious-Recovered) model with COVID-19 data from Mexican localities.
• The sequential data assimilation approach proved effective in updating epidemic forecasts as new data became available.

Conclusions:

• The proposed Bayesian sequential data assimilation method provides a viable approach for epidemic forecasting in complex, non-autonomous systems.
• This method offers a practical balance between fitting observed data and predicting the future dynamics of epidemics like COVID-19.