Recalibrating disease parameters for increasing realism in modeling epidemics in closed settings

Area of Science:

• Epidemiology
• Computational Biology
• Network Science

Background:

• Homogeneous mixing is a common assumption in epidemic modeling due to its simplicity.
• Agent-based models often assume homogeneous mixing within settings like schools and workplaces due to limited contact data.
• Recent high-resolution interaction data allows for evaluating the accuracy of the homogeneous mixing assumption.

Purpose of the Study:

• To assess the accuracy of the homogeneous mixing assumption in epidemic modeling using real-world contact network data.
• To compare epidemic dynamics simulated on contact networks versus homogeneous mixing models.
• To determine if recalibration of parameters can improve the agreement between these modeling approaches.

Main Methods:

• Stochastic spreading simulations were performed on empirical and synthetic contact networks.
• Simulations were also conducted using a homogeneous mixing hypothesis on populations of equivalent size.
• Epidemiological parameters in the homogeneous mixing model were adjusted to match the prevalence curves from contact network models.
• Agreement was quantified by comparing epidemic peak times, peak values, and total epidemic sizes.

Main Results:

• The homogeneous mixing approach provided good approximations for epidemic peak times and values, with a median relative difference under 20%.
• Accuracy of peak time prediction varied by setting, but peak value prediction was independent of the setting.
• Parameter recalibration showed a linear relationship with the original epidemic parameters and was robust across different settings, groups, and population sizes.

Conclusions:

• Recalibrating epidemiological parameters allows homogeneous mixing models to closely approximate epidemic curves derived from contact networks.
• Using recalibrated homogeneous mixing models can enhance the accuracy and realism of agent-based simulations.
• This approach helps mitigate the inherent biases associated with the simplified homogeneous mixing assumption in epidemic modeling.