Intrinsic randomness in epidemic modelling beyond statistical uncertainty.

Area of Science:

• Epidemiology
• Mathematical Biology
• Infectious Disease Modeling

Background:

• Current infectious disease risk frameworks primarily address epistemic uncertainty (imperfect knowledge).
• Aleatoric uncertainty (intrinsic randomness) in epidemics is often unquantified due to the observation of single outbreaks.
• This gap limits accurate risk assessment and policy decisions.

Purpose of the Study:

• To develop a framework for characterizing both aleatoric and epistemic uncertainty in infectious disease outbreaks.
• To explicitly decompose aleatoric variance into mechanistic components and analyze its temporal dynamics.
• To highlight the underestimation of risk when aleatoric uncertainty is ignored.

Main Methods:

• Utilized a time-varying general branching process model.
• Decomposed aleatoric variance into specific mechanistic factors contributing to epidemic uncertainty.
• Analyzed the dynamic growth of forecasting uncertainty over time.

Main Results:

• Substantial outbreak uncertainty can arise without superseding events or overdispersed offspring distributions.
• Aleatoric forecasting uncertainty increases dynamically and rapidly.
• Forecasting solely on epistemic uncertainty leads to significant underestimation of true epidemic risk.

Conclusions:

• Failure to incorporate aleatoric uncertainty provides a misleadingly low assessment of potential infectious disease risk.
• Accurate epidemic risk management necessitates quantifying both aleatoric and epistemic uncertainty.
• The proposed method demonstrates the extent of underestimation using historical epidemic data.