Within host dynamics have been largely overlooked in epidemic modeling.  However, variability in pathogen growth rates and death rates within a host can affect between host transmission and thus epidemic dynamics.  Furthermore, tradeoffs between transmission within hosts and between hosts may affect natural selection.  The standard approach taken to modeling epidemics is to assume that within host dynamics are deterministic.  For small population sizes, however, stochasticity can have important effects on dynamics.  Similarly, when the number of infected individuals is small, as in the early stage of an epidemic, stochasticity is important.  Ultimately, to understand the evolution of virulence, we need to account for drift, and thus, we need to incorporate stochasticity both within and between hosts.  The study system we use is an insect nucleopolyhedrovirus, because insects have easily modeled immune systems and the outcome of infection is easily scored.  The model that we use is a stochastic SIR model with a time delay between infection and death that is determined by a birth-death model.

Results/Conclusions

The model shows that allowing for stochasticity within hosts can strongly increase variability in the fraction of the population that ultimately becomes infected.  Furthermore, the model shows that as the difference between birth rates and death rates increase, the variability in epidemic outcomes decreases.  We use these results to understand the impact of stochasticity on host-pathogen coevolution, estimating model parameters to make predictions about the effects of natural selection and genetic drift in the evolution of the gypsy moth system.