Measles transmission and surveillance in open subcommunities

Abstract

We're often in a situation where we have models of measles transmission that are stuck at a coarse population scale. This paper is about increasing their resolution or, said differently, creating consistent models for the population's subcommunities. We pay special attention to the fact that even if the full population is isolated, subcommunities are open to measles importations and related consequences of population movement. It turns out that the resolution issue can be productively framed as a constrained optimization problem, where the constraint is aggregation to the coarse model. That mathematical scaffolding allows us to make epidemiologically motivated approximations concentrated at the large population scale, where they're more valid, which leaves the estimates at the fine subcommunity scale atypically unstructured. We develop the idea in the context of the pre-vaccination-era 60 city UK dataset, building transmission models for all 60 cities. We find that they have intuitive seasonality, volatility, and importation elements, but that they depend on nontrivial dynamics in their case reporting probabilities. Potential interpretations of those surveillance processes are discussed in some detail, and we also describe how they influence inference on the balance between competing cryptic transmission and transient elimination dynamics.