Early warning signs

Abstract

Epidemiology Modeling an emerging infectious disease is an inexact science. At an early stage of an epidemic, we only have sparse data, little knowledge of the mechanisms driving emergence, and an urgent need to devise control measures that will be effective. Using epidemiological incidence reports, Brett and Rohani have developed a detection algorithm for disease (re)emergence that is agnostic to the mechanisms involved. This supervised statistical learning algorithm was trained on data collected for mumps outbreaks in England and resurgent pertussis in the United States. The algorithm successfully anticipated reemergence of mumps 4 years in advance, which would have given plenty of time for mitigation efforts to be implemented. The algorithm also performed well for vector-borne diseases, including dengue in Puerto Rico, and predicted the rapid emergence of plague in Madagascar. The success of this approach stems from the common statistical properties of incidence data across disease emergence contexts and has obvious application for monitoring waves of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) reemergence. PLOS BIOL. 18 , e3000697 (2020).