The Effect of Misspecifying Latent and Infectious Periods in Space-Time Epidemic Models

Abstract

Individual level models (ILMs) are a class of models that can be applied to epidemic data to help in the understanding of the spatio-temporal dynamics of infectious diseases. Typically, these models are analyzed in a Bayesian framework using Markov chain Monte Carlo (MCMC) methodology. Here, we test the effect of misspecifying the latent and infectious period in such a model. We do this by simulating data from a simple spatial ILM, and then fitting various misspecified models to the simulated data. The fitted models serve as a basis for investigating the effect of the misspecification of latent and infectious periods on model parameter estimates, as well as estimates of the basic reproduction number.Additionally, we analyze how a given preventative control strategy, optimized via simulation from a fitted model with assumed latent and infectious periods, is affected by such misspecification. We observe bias in the estimation of model parameters as latent and infectious periods become more misspecified, as well as a significant deviation in estimates of the basic reproduction number from those observed under the true model. Where the misspecification results in a higher basic reproduction number estimate, we also find that a more stringent control policy is required to achieve a given policy goal.