Inferring the risk factors behind the geographical spread and transmission of Zika in the Americas

Background: An unprecedented Zika virus epidemic occurred in the Americas during 2015-2016. The size of the epidemic in conjunction with newly recognized health risks associated with the virus attracted significant attention across the research community. Our study complements several recent studies which have mapped epidemiological elements of Zika, by introducing a newly proposed methodology to simultaneously estimate the contribution of various risk factors for geographic spread resulting in local transmission and to compute the risk of spread (or re-introductions) between each pair of regions. The focus of our analysis is on the Americas, where the set of regions includes all countries, overseas territories, and the states of the US. Methodology/Principal findings: We present a novel application of the Generalized Inverse Infection Model (GIIM). The GIIM model uses real observations from the outbreak and seeks to estimate the risk factors driving transmission. The observations are derived from the dates of reported local transmission of Zika virus in each region, the network structure is defined by the passenger air travel movements between all pairs of regions, and the risk factors considered include regional socioeconomic factors, vector habitat suitability, travel volumes, and epidemiological data. The GIIM relies on a multi-agent based optimization method to estimate the parameters, and utilizes a data driven stochastic-dynamic epidemic model for evaluation. As expected, we found that mosquito abundance, incidence rate at the origin region, and human population density are risk factors for Zika virus transmission and spread. Surprisingly, air passenger volume was less impactful, and the most significant factor was (a negative relationship with) the regional gross domestic product (GDP) per capita. Conclusions/Significance: Our model generates country level exportation and importation risk profiles over the course of the epidemic and provides quantitative estimates for the likelihood of introduced Zika virus resulting in local transmission, between all origin-destination travel pairs in the Americas. Our findings indicate that local vector control, rather than travel restrictions, will be more effective at reducing the risks of Zika virus transmission and establishment. Moreover, the inverse relationship between Zika virus transmission and GDP suggests that Zika cases are more likely to occur in regions where people cannot afford to protect themselves from mosquitoes. The modeling framework is not specific for Zika virus, and could easily be employed for other vector-borne pathogens with sufficient epidemiological and entomological data. Author summary: Since May 2015, when Zika was first reported in Brazil, the virus has spread to over 60 countries and territories, and imported cases of Zika have been increasingly reported worldwide. However, there is still much uncertainty behind the mechanisms which dictated the rapid emergence of the epidemic. This work introduces a novel modeling framework to improve our understanding of the risk factors which contributed to the geographic spread and local transmission of Zika during the 2015-2016 epidemic in the Americas. The model is informed by data on regional socioeconomic factors, mosquito abundance, travel volumes, and epidemiological data. As expected, our results indicate that increased presence of mosquitoes, human hosts, and viruses increase the risk for mosquito-borne virus transmission. Passenger air travel, however, was less impactful, suggesting that travel restrictions will have minimal impact on controlling similar epidemics. Importantly, we found that a lower regional GDP was the best predictor of Zika virus transmission, suggesting that Zika is primarily a disease of poverty.