Explaining rapid reinfections in multiple-wave influenza outbreaks: Tristan da Cunha 1971 epidemic as a case study

Camacho, Anton; Ballesteros, Sébastien; Graham, Andrea L.; Carrat, Fabrice; Ratmann, Oliver; Cazelles, Bernard

AIC c correspond to visible, practically significant differences in reproducing the characteristic second wave, we performed predictive checks by simulating 10 5 time series under each ML-fitted model. Comparison of the behaviour of the best model (Win) with each competitor reveals superimposed dynamics with the AoN model and confirms that the dynamics of the four other models are different with respect to the second epidemic wave ( figure 4a-e, upper panels) . Furthermore, figure 4f shows that the extinction probability increases rapidly at the end of the first wave for the 2Vi, Mut and PPI models, whereas the Win and AoN models appear to be much more robust to stochastic extinctions during the inter-wave period. In the electronic supplementary material, text S4, we supplement a suite of statistical analyses to evaluate and compare the goodness of fit of these predictive simulations to the characteristic second-wave infection dynamics. Overall, these analyses support the view that, out of the models considered, the Win and AoN models explain the observed time series significantly better.