Using information theory to optimise epidemic models for real-time prediction and estimation

Area of Science:

• Epidemiology
• Mathematical Biology
• Information Theory

Background:

• The effective reproduction number (Rt) is crucial for tracking infectious disease epidemics.
• Current methods for estimating Rt, like renewal models, rely on heuristic piecewise functions, which can be unreliable.
• A principled method for selecting these functions is lacking, potentially leading to inaccurate real-time assessments.

Purpose of the Study:

• To develop a rigorous and practical scheme for selecting piecewise functions to estimate Rt.
• To improve the reliability and accuracy of real-time Rt inference for epidemic dynamics.
• To provide a method that optimizes short-term prediction and detects significant changes in transmission rates.

Main Methods:

• Utilized the accumulated prediction error (APE) metric from information theory for model selection.
• Derived exact posterior prediction distributions for infected population size.
• Integrated these distributions within the APE framework to identify the best-supported piecewise function.

Main Results:

• The developed APE-based scheme reliably identifies the optimal piecewise function for Rt estimation.
• This method enhances short-term prediction accuracy and facilitates rapid detection of Rt fluctuations.
• Demonstrated that heuristic function choices can be misleading, highlighting the need for formal selection.

Conclusions:

• The APE-based method offers an exact and reliable approach for real-time Rt estimation in epidemics.
• This technique improves epidemic forecasting and intervention evaluation by optimizing model selection.
• The method is computationally efficient and broadly applicable to similar models in related scientific fields.