Bayesian model selection for COVID-19 pandemic state estimation using extended Kalman filters: Case study for Saudi...

Area of Science:

• Epidemiology
• Mathematical Modeling
• Public Health

Background:

• COVID-19 presented significant global health and economic challenges.
• Mathematical models are crucial for understanding disease transmission and informing public health decisions.
• Quantifying uncertainty in data-driven mechanistic models is vital for public health applications.

Purpose of the Study:

• To evaluate the predictive and state estimation capabilities of the SEIQRD (Susceptible-Exposed-Infected-Quarantined-Recovered-Deceased) model during the COVID-19 pandemic.
• To compare the long-term behavior and suitability of the SEIQRD model against the classical SIRD (Susceptible-Infected-Recovered-Deceased) model.
• To establish a basis for validating and comparing epidemiological models in terms of long-term behavior and required complexity.

Main Methods:

• Utilized Bayesian inference with a nested sampling algorithm.
• Employed recursive state estimation via the Extended Kalman Filter (EKF).
• Applied a systematic methodology for time-varying parameter estimation and uncertainty quantification.

Main Results:

• The proposed methodology, integrated within the EKF, generated predictions closely aligned with observed COVID-19 data (active cases and deaths).
• Obtained credible confidence intervals for epidemiological nonlinear dynamical system model parameters using Saudi Arabia COVID-19 data.
• Demonstrated the effectiveness of the SEIQRD model and the EKF framework for pandemic response.

Conclusions:

• The EKF-implemented framework provides a robust tool for addressing future pandemics by generating reliable predictions and quantifying uncertainties.
• The study highlights the importance of model complexity selection based on data and desired accuracy.
• Accurate parameter estimation and uncertainty quantification are essential for effective epidemiological modeling in public health.