A Bayesian modelling framework with model comparison for epidemics with super-spreading.

Area of Science:

• Epidemiology
• Mathematical Biology
• Statistical Modeling

Background:

• Epidemic transmission is often heterogeneous, characterized by super-spreading events or individuals.
• Inferring super-spreading typically relies on secondary case data (offspring distribution), which is often unavailable.
• Incidence time-series data offer a more accessible alternative for epidemic analysis.

Purpose of the Study:

• To develop and validate a flexible multi-model framework for analyzing epidemic transmission dynamics using incidence time-series.
• To differentiate between homogeneous transmission, super-spreading events, and super-spreading individuals.
• To provide a disease-agnostic tool for public health and infectious disease management.

Main Methods:

• A framework of five discrete-time, stochastic, branching-process models was developed.
• Bayesian inference with Markov Chain Monte-Carlo methods was used for parameter estimation.
• Model comparison was performed using Bayes factors and importance sampling for marginal likelihood estimation.

Main Results:

• The framework successfully identified the correct model and accurately inferred parameters (e.g., basic reproduction number) from simulated data.
• Application to SARS and COVID-19 incidence data consistently identified the same transmission model and mechanism across different time series.
• Inferred estimates align with previous studies using secondary case data.

Conclusions:

• The developed modeling framework effectively analyzes epidemic dynamics using incidence time-series data, even in the presence of super-spreading.
• This approach provides a valuable, disease-agnostic tool for quantifying super-spreading's contribution to transmission.
• Accurate quantification of super-spreading is vital for informing infectious disease control and public health interventions.