Anticipating the risk and spatial spread of measles in populations with high MMR uptake: using school-household networks to understand the 2013 - 2014 outbreak in the Netherlands.

Measles outbreaks are still routine, even in countries where vaccination coverage exceeds the guideline of 95%. Therefore, achieving ambitions for measles eradication will require understanding how unvaccinated children interact with others who are unvaccinated. Here we propose a novel framework for modelling measles transmission to better understand outbreaks in high uptake situations. The high importance of school- and home-based transmission to overall outbreak dynamics is well established. Making use of this, we created a network of all primary and secondary schools in the Netherlands based on the total number of household pairs between each school. A household pair are siblings from the same household who attend a different school. We parameterised the network with individual level administrative household data provided by the Dutch Ministry for Education and estimates of school level uptake of the Mumps, Measles and Rubella (MMR) vaccine. We analyse the network to establish the relative strength of contact between schools. We simulated measles outbreaks on the network and evaluated the model against empirical measles data per postcode-area from a large outbreak in 2013 (2766 cases), comparing the model to alternative models that do not account for specific network structure or school-level vaccine uptake. Our network analysis shows that schools associated with low vaccine uptake are highly connected, particularly Orthodox-Protestant schools (Coleman Homophily Index = 0.63). Simulations on the Network were able to reproduce the observed size and spatial distribution of the historic outbreak much more clearly than the alternative models, with a case weighted Receiver Operating Condition sensitivity of 0.94 for the data-driven network model and 0.38 and 0.23 for the alternative models. Further, we establish that variation in local network properties result in clear differences in final size of outbreaks seeded in orthodox-protestant-affiliated and other schools with low MMR coverage. Our framework indicates that clustering of unvaccinated children in primary schools connected by unvaccinated children in related secondary schools lead to large, connected clusters of unvaccinated children. Using our approach, we could explain historical outbreaks on a spatial level. Our framework could be further developed to aid future outbreak response.