Predicting Covid-19 pandemic waves with biologically and behaviorally informed universal differential equations

Area of Science:

• Epidemiology
• Mathematical Modeling
• Computational Science

Background:

• The COVID-19 pandemic highlighted the interconnectedness of population behavior and disease transmission dynamics.
• Understanding these coupled behavior-disease systems is crucial for effective pandemic response.

Purpose of the Study:

• To demonstrate the capability of Universal Differential Equation (UDE) models in capturing the feedback loop between population behavior and disease spread.
• To develop and test a UDE model for predicting second waves of the COVID-19 pandemic.

Main Methods:

• Development of a novel UDE model tailored for COVID-19 dynamics.
• Training the UDE model using data to learn the interplay between behavioral responses and disease transmission.
• Evaluating the model's predictive accuracy for subsequent pandemic waves across different populations.

Main Results:

• UDE models successfully learned the coupled behavior-disease dynamics from observational data.
• The developed UDE model demonstrated the ability to predict second pandemic waves.
• Model performance was contingent on incorporating learning biases related to disease transmission and population response.

Conclusions:

• Universal Differential Equations offer a promising approach for modeling complex, coupled behavior-disease systems.
• While UDEs show potential for pandemic wave prediction, further refinement is needed before policy application.
• This study provides insights into the benefits, limitations, and techniques for applying UDEs to real-world coupled systems.