Bayesian modeling of dynamic behavioral change during an epidemic

Area of Science:

• Epidemiology
• Mathematical Biology
• Public Health

Background:

• Infectious disease outbreaks significantly alter population behavior, impacting transmission dynamics.
• Traditional epidemic models often overlook these crucial behavioral changes, limiting their real-world applicability.
• Accurate modeling requires incorporating dynamic behavioral responses to outbreak severity.

Purpose of the Study:

• To introduce a novel class of data-driven epidemic models that explicitly characterize and estimate behavioral change.
• To develop a framework where time-varying transmission is linked to population "alarm" levels, derived from epidemic history.
• To assess the estimability of population alarm using various modeling techniques.

Main Methods:

• Developed a novel data-driven epidemic model incorporating a time-varying transmission rate.
• Defined population "alarm" as a function of the past epidemic trajectory.
• Investigated parametric and non-parametric (splines, Gaussian processes) approaches for estimating alarm.
• Employed a data-augmented Bayesian framework for estimation with partially observed epidemic data.

Main Results:

• Demonstrated the ability to capture and estimate time-varying transmission dynamics driven by behavioral changes.
• Showcased the utility of the population alarm concept in reflecting public response to outbreaks.
• Validated the model's performance across diverse scenarios and with real epidemic data.

Conclusions:

• The proposed data-driven epidemic models offer a significant advancement by integrating behavioral dynamics.
• The population alarm framework provides a robust method for understanding and modeling real-time behavioral responses.
• This approach enhances the accuracy and utility of epidemic modeling for public health interventions.