Forecasting Chikungunya spread in the Americas via data-driven empirical approaches

Abstract

Background: Chikungunya virus (CHIKV) is endemic to Africa and Asia, but the Asian genotype invaded the Americas in 2013. The fast increase of human infections in the American epidemic emphasized the urgency of developing detailed predictions of case numbers and the potential geographic spread of this disease. Methods: We developed a simple model incorporating cases generated locally and cases imported from other countries, and forecasted transmission hotspots at the level of countries and at finer scales, in terms of ecological features. Results: By late January 2015, >1.2 M CHIKV cases were reported from the Americas, with country-level prevalences between nil and more than 20 %. In the early stages of the epidemic, exponential growth in case numbers was common; later, however, poor and uneven reporting became more common, in a phenomenon we term "surveillance fatigue." Economic activity of countries was not associated with prevalence, but diverse social factors may be linked to surveillance effort and reporting. Conclusions: Our model predictions were initially quite inaccurate, but improved markedly as more data accumulated within the Americas. The data-driven methodology explored in this study provides an opportunity to generate descriptive and predictive information on spread of emerging diseases in the short-term under simple models based on open-access tools and data that can inform early-warning systems and public health intelligence.