Anticipating Outbreaks with Imperfect Data

Anticipating disease emergence is a challenge with clear public health ramifications. Theoretical studies have already demonstrated that epidemic transitions are in principle preceded by detectable temporal trends in statistics (early-warning signals). We investigated the robustness of these early-warning signals under simulated realistic disease reporting scenarios, testing the effects of case reporting error, reporting probability, and weekly and monthly aggregation of case reports. Our case report data were simulated by combining a stochastic SIR model with a model of reporting error.

We found that seven of ten common statistics used as early warning signals perform well for realistic reporting scenarios, and are strong candidates for incorporation in disease emergence monitoring systems.

Our paper, published in PLoS Computational Biology, is now available online:

Brett, T.S., E.B. O’Dea, É. Marty, P.B. Miller, A.W. Park, J.M. Drake & P. Rohani. 2018. “Anticipating epidemic transitions with imperfect data.” PLoS Computational Biology 14(6): e1006204. https://doi.org/10.1371/journal.pcbi.1006204.