Forecasting Multi-Wave Epidemics Through Bayesian Inference.

Area of Science:

• Epidemiology
• Computational Biology
• Public Health

Background:

• The COVID-19 pandemic presented challenges in predicting outbreak trajectories.
• Timely epidemiological data is crucial for effective public health interventions and resource allocation.

Purpose of the Study:

• To develop and validate a near-real-time Bayesian method for inferring and forecasting multiwave infectious disease outbreaks.
• To assess the utility of the method for short-term forecasting and medical resource planning during the COVID-19 pandemic.

Main Methods:

• A Bayesian approach employing Markov chain Monte Carlo (MCMC) sampling to estimate parameters of one- and multiwave infection models.
• Convolution of infection models with incubation period distributions to create competing disease models.
• Utilization of information-theoretic criteria for model selection in forecasting.

Main Results:

• The method demonstrated robustness to noisy epidemiological data.
• Accurate short-term forecasts with uncertainty bounds were generated for COVID-19 outbreaks.
• The approach successfully identified the transition from single-wave to multiwave outbreaks, indicating failures in containment efforts.

Conclusions:

• The developed Bayesian method offers a simple yet effective tool for near-real-time outbreak forecasting.
• This method can significantly aid in medical resource planning by providing reliable short-term predictions.
• The study highlights the importance of dynamic modeling to capture evolving infectious disease dynamics, such as successive surges.