Corresponding Author: Tobias S. Brett
Recent literature has highlighted the potential to predict disease outbreak through early-warning signals (EWS), summary statistics which undergo characteristic changes as an epidemic transition is approached. In these transitions, disease transmission can shift from limited, stuttering chains of transmission to large-scale outbreaks. Thus successful prediction of epidemic transitions using EWS could lead to early intervention, with powerful implications for public health. While theoretical studies show that detectable trends in EWS precede an epidemic transition, these studies do n
ot consider realistic epidemiological data, which is often imperfect. Epidemiological data come from public health agencies in the form of case reports, aggregating individual cases of disease. These reports can be imperfect due to factors such as under reporting, poor notification protocols, clerical/clinical errors, and socio-political factors. These imperfect data may affect EWS and cause them not to display the trends predicting disease emergence found in theoretical studies, thus questioning the application of EWS to observed data.
To address this, a team of researchers including CEID Members Toby Brett, Eamon O’Dea, Paige Miller, Andrew Park, John Drake, and Pej Rohani sought to assess how imperfect real-world data impacts 10 different EWS’s performance. The team simulated a stochastic SIR model of an emerging pathogen, and corrupted the simulated case reports to represent reporting error. They then quantified the performance of each EWS using the area under the curve statistic, which measures how well an EWS distinguishes between simulations of an emerging disease and a stationary disease. Their results found that while different EWS respond to imperfect data differently, 7 of 10 EWS performed well for most realistic scenarios. They conclude that EWS remain strong candidates for incorporation in disease emergence monitoring systems, and that imperfect data is not a barrier to the use of EWS.
Brett TS, O’Dea EB, Marty É, Miller PB, Park AW, Drake JM, et al. (2018) Anticipating epidemic transitions with imperfect data. PLoS Comput Biol 14(6): e1006204. https://doi.org/10.1371/journal.pcbi.1006204