Mechanistic Understanding of Pandemic Dynamics: A Multiscale Algorithmic Framework.

Area of Science:

• Epidemiology
• Computational Biology
• Data Science

Background:

• Standard epidemiological models struggle with parameter identifiability in early pandemic stages, limiting predictive accuracy with sparse data.
• Curve-fitting population profiles is sensitive to uncertainty, hindering reliable early outbreak assessment.
• A novel framework is needed to overcome these limitations for robust early outbreak analysis.

Purpose of the Study:

• To present a robust, data-efficient framework for early outbreak assessment using multiscale analysis and Computational Singular Perturbation (CSP).
• To overcome the limitations of standard compartmental models in early pandemic stages.
• To identify dominant dynamical drivers and predict epidemic peaks using early-stage data.

Main Methods:

• Utilized multiscale analysis and Computational Singular Perturbation (CSP) for early outbreak assessment.
• Employed a calibrated SEIRD model to identify dominant 'explosive time scales' and critical transitions in epidemic growth.
• Assessed the framework against COVID-19 waves in Greece (Delta and Omicron variants) using short calibration windows.

Main Results:

• CSP successfully identified distinct dynamical drivers for different COVID-19 waves (Delta vs. Omicron).
• The framework accurately predicted the timing of outbreak weakening, demonstrating robustness with early-stage data.
• Demonstrated the ability to disentangle complex epidemic mechanisms and assess intervention efficacy in real-time.

Conclusions:

• Multiscale analysis with CSP provides a powerful, pathogen-agnostic early-warning system for pandemics.
• The framework overcomes limitations of traditional models, offering improved predictive utility with sparse data.
• This approach enables real-time assessment of epidemic dynamics and intervention effectiveness.