Measles epidemics in pre-vaccine era United States cities: Natural experiments in metapopulation dynamics

Background/Question/Methods

In ecology, historical measles epidemics have long provided insight into the
dynamics of an idealized two-species system, particularly through work on the
classic England and Wales dataset.  Here we present a new epidemiological
dataset consisting of 24 years of pre-vaccine era weekly case reports of
measles in 85 United States cities.  This dataset is remarkable for the wide
geographic extent and heterogeneity of its cities; it provides a series of
natural experiments in a spatially isolated yet interconnected metapopulation.

The dynamics of each measles epidemic is non-linear, and is influenced both by
within and between-city factors.  Within-city factors include demographic
history, particularly births and immigration, as well as the city's recent
epidemic history. The primary influence of other cities is through importation
of infected individuals.

We assess the relative contribution of between and within city factors by
examining epidemic timing compared to mean metapopulation disease incidence
using wavelet analysis.

Results/Conclusions

We show that timing of yearly epidemic outbreak is
highly variable between both years and cities, and that geographic isolation or
proximity does not predict epidemic timing or year-to-year timing variance.
Significantly, smaller cities display greater variance of epidemic timing,
indicating less influence of the metapopulation at large.  We conduct a
bootstrap analysis by shuffling individual city-years (September to August) in
three ways: one, between years and within cities, to disrupt inter-city
influences; two, between cities within years, to disrupt intra-city influences;           
and three, between both cities and years, to construct a null distribution.
Correspondingly, we use three ensembles of synthetic datasets to demonstrate
the nested effects of within and between city factors on epidemic timing
variance.

Surprisingly, and unlike previous work, we detected no spatial structure of
epidemic timing here.  The observed high variance of epidemic timing in smaller
cities, as well as the lack of spatial synchrony,  could play a major role in
disease persistence in metapopulations at large, even in the absence of
populous focal cities.