Bayesian data assimilation for estimating instantaneous reproduction numbers during epidemics: Applications to...

Area of Science:

• Epidemiology
• Computational Biology
• Statistics

Background:

• Accurate estimation of time-varying epidemiological parameters like the reproduction number (Rt) is crucial for understanding infectious disease transmission.
• Current methods often suffer from lagging observations, averaging effects, and inadequate uncertainty quantification.

Purpose of the Study:

• To develop a robust Bayesian data assimilation framework for accurate and timely estimation of time-varying epidemiological parameters.
• To introduce the 'DARt' (Data Assimilation for real-time Rt) system to address limitations in current Rt estimation methods.

Main Methods:

• A Bayesian data assimilation framework incorporating observation delays to jointly infer infections and Rt.
• Instantaneous updates with new observations and a model selection mechanism to capture abrupt changes.
• Bayesian smoothing for quantifying and reducing uncertainty in parameter estimates.

Main Results:

• The DARt system effectively addresses time misalignment from lagging observations.
• It overcomes averaging drawbacks through instantaneous updates and adaptive model selection.
• Bayesian smoothing significantly quantifies and reduces estimation uncertainty.

Conclusions:

• The DARt system offers a state-of-the-art solution for real-time Rt estimation during emerging epidemics.
• It provides accurate and timely insights into disease transmission dynamics, validated by COVID-19 data.
• This framework presents a promising approach for epidemiological surveillance and control strategies.