Epidemiological Forecasting with Model Reduction of Compartmental Models. Application to the COVID-19 Pandemic

Area of Science:

• Epidemiology
• Mathematical Modeling
• Public Health

Background:

• Predicting infectious disease spread is crucial for public health interventions.
• Existing epidemiological models often require substantial data, limiting their application in early-stage or low-resource settings.
• Accurate short-term forecasting is essential for timely resource allocation and response planning.

Purpose of the Study:

• To develop and evaluate a novel forecasting method for epidemiological health series.
• To predict the number of infected and removed individuals (due to death or recovery) for COVID-19.
• To validate the method's efficacy using data from the COVID-19 pandemic in France.

Main Methods:

• The proposed method utilizes model order reduction of parametric compartmental models.
• It is specifically designed to handle scenarios with limited sanitary data.
• The approach operates on a two-week forecasting horizon at regional and interregional levels.

Main Results:

• The method demonstrated efficiency in predicting infected and removed populations during the two major COVID-19 waves in France (February-November 2020).
• Numerical results confirmed the approach's capability to forecast key epidemiological indicators.
• The model's performance was validated against real-world pandemic data.

Conclusions:

• The developed forecasting method shows promising potential for epidemiological predictions, especially with limited data.
• This approach can support public health decision-making by providing reliable short-term health series forecasts.
• The technique offers a valuable tool for understanding and managing infectious disease outbreaks.