Pandemic influenza has the epidemic potential to kill millions of people.\nWhile various preventive measures exist (i.a., vaccination and school\nclosures), deciding on strategies that lead to their most effective and\nefficient use remains challenging. To this end, individual-based\nepidemiological models are essential to assist decision makers in determining\nthe best strategy to curb epidemic spread. However, individual-based models are\ncomputationally intensive and it is therefore pivotal to identify the optimal\nstrategy using a minimal amount of model evaluations. Additionally, as\nepidemiological modeling experiments need to be planned, a computational budget\nneeds to be specified a priori. Consequently, we present a new sampling\ntechnique to optimize the evaluation of preventive strategies using fixed\nbudget best-arm identification algorithms. We use epidemiological modeling\ntheory to derive knowledge about the reward distribution which we exploit using\nBayesian best-arm identification algorithms (i.e., Top-two Thompson sampling\nand BayesGap). We evaluate these algorithms in a realistic experimental setting\nand demonstrate that it is possible to identify the optimal strategy using only\na limited number of model evaluations, i.e., 2-to-3 times faster compared to\nthe uniform sampling method, the predominant technique used for epidemiological\ndecision making in the literature. Finally, we contribute and evaluate a\nstatistic for Top-two Thompson sampling to inform the decision makers about the\nconfidence of an arm recommendation.\n