The implications of metapopulation dynamics on the design of vaccination campaigns.

Control programmes for vaccine preventable diseases typically operate under logistic constraints such as limited resources and in spatially structured populations where the assumption of homogeneous mixing is invalid. It is unclear, therefore, how to maximise the effectiveness of campaigns in such populations. We investigate how to deploy vaccine in metapopulations by comparing the effectiveness of alternative vaccination strategies on reducing disease occurrence (presence/absence), using canine rabies as a model system, and a domestic dog population within a Tanzanian district divided into sub-populations corresponding to villages. We use patch-occupancy models to quantify the contribution of sub-populations to disease occurrence ("risk") and model allocation strategies for a limited number of vaccine doses that prioritize villages based on their size, risk, or the reduction in risk for the entire population that would result from vaccination. We assume that a maximum of 70% of susceptible individuals in a village could be vaccinated, and that only susceptible dogs are vaccinated. The most effective strategy maximised the reduction in risk of the entire population, and was up to 62% more effective than the other strategies. Large, single-pulse campaigns provided the greatest short-term protection, but higher frequencies of smaller pulses were more effective at reducing long-term disease occurrence. Vaccine allocation on a per-dose basis was substantially more effective than a per-village strategy, indicating that operational constraints can reduce control effectiveness. The spatial distribution and abundance of hosts have an important influence on disease dynamics and these results demonstrate that metapopulation models can be used to substantially improve the effectiveness of vaccination campaigns and optimize the allocation of limited control resources.