Bayesian parameter inference for dynamic infectious disease modelling: rotavirus in Germany.

Area of Science:

• Epidemiology
• Mathematical Biology
• Biostatistics

Background:

• Infectious disease dynamics are often modeled using ordinary differential equations, requiring parameter calibration from data.
• Current methods frequently fix parameters based on external information, potentially causing biased inference and unexamined uncertainty.
• This approach limits the understanding of disease transmission and model accuracy.

Purpose of the Study:

• To develop a Bayesian inference framework for more data-driven calibration of epidemic models.
• To reduce dependence on external parameter quantification and examine the impact of parameter fixing.
• To improve the accuracy and reliability of infectious disease modeling.

Main Methods:

• Developed a Bayesian inference framework incorporating residual autocorrelation and model averaging.
• Applied the framework to age-stratified weekly rotavirus incidence data in Germany (2001-2008).
• Utilized a susceptible-infectious-susceptible (SIS) model with stochastic reporting of new cases.

Main Results:

• Estimated rotavirus detection rates significantly higher in eastern German states (19.0%) compared to western states (4.3%).
• Infectiousness of symptomatically infected individuals was over 10 times higher than asymptomatically infected individuals (95% CI: 8.1–19.6).
• Demonstrated the substantial impact of pre-fixing parameters on model-predicted transmission dynamics.

Conclusions:

• The novel Bayesian framework provides valuable epidemiological insights into rotavirus transmission.
• Data-driven calibration reduces bias and uncertainty compared to methods relying on fixed parameters.
• Findings highlight regional disparities and transmission characteristics crucial for public health interventions.