Estimates of the reproduction ratio from epidemic surveillance may be biased in spatially structured populations

Accurate estimates of the reproduction ratio (Rt) are crucial for projecting the evolution of an infectious disease epidemic and for guiding the public health response. During epidemics, estimates of Rt typically come from surveillance data, e.g., time series of recorded cases. However, we prove that estimates of the reproduction ratio based on inference from surveillance data can be inaccurate if the population is made up of a network of spatially distinct communities, as the space–mobility interplay may hide the true evolution of the epidemic from surveillance data. Consequently, we prove that traditional epidemiological surveillance may underestimate the reproduction ratio over long periods, even to the points of mistaking growing epidemics as subsiding ones. To address this, we use the spectral properties of the operator describing the epidemic coupling among spatial communities to quantify this error, and to propose a correction that removes it across all epidemic phases, by reweighing surveillance data by the leading eigenvector of this operator. We validate this correction against simulated epidemics and use COVID-19 in France as a case study. Specifically, we find evidence that biased estimates of Rt during the rise of the Alpha variant in early 2021 might have contributed to delaying public health interventions, ultimately making it necessary to enforce the third national lockdown. We show that our correction, if available, would have correctly indicated the growth of the Alpha variant. Beyond this retrospective case study, however, our results apply to any epidemic in which mobility is a driver of circulation. Our findings may help improve epidemic monitoring and surveillance and inform strategies for public health responses.

Ref: Birello P, et.al. Nature Physics 20, 1204-1210 (2024) doi: 10.1038/s41567-024-02471-7