Model selection and parameter estimation for dynamic epidemic models via iterated filtering: application to rotavirus...

Area of Science:

• Epidemiology
• Mathematical Modeling
• Computational Statistics

Background:

• Dynamic models are crucial in infectious disease epidemiology but often incorporate variability subjectively.
• Inference for stochastic transmission models is challenging due to intractable likelihoods from partial observability.
• A systematic approach is needed for model selection and parameter inference in epidemic modeling.

Purpose of the Study:

• To demonstrate a systematic approach for model selection and parameter inference in dynamic epidemic models.
• To address the challenge of adequately including variability in epidemiological models.
• To estimate the basic reproduction number (R0) for rotavirus using German surveillance data.

Main Methods:

• Inference was performed for six partially observed Markov process models with varying levels of assumed variability.
• Iterated filtering methods, implemented in the R package pomp, were used for inference in stochastic transmission models.
• The approach was illustrated using German rotavirus surveillance data from 2001 to 2008.

Main Results:

• A systematic framework for model selection and parameter inference in dynamic epidemic models was demonstrated.
• Practical challenges associated with the iterated filtering methods were discussed.
• A model-based estimate for the basic reproduction number (R0) was calculated using the rotavirus data.

Conclusions:

• The study provides a robust framework for incorporating and selecting variability in epidemic models.
• Iterated filtering offers a viable method for inference in partially observed stochastic epidemic models.
• The findings contribute to more accurate infectious disease modeling and parameter estimation, such as R0.