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

Viruses, such as dengue, Zika, yellow fever and chikungunya, depend on mosquitoes for transmission. Their epidemics typically present periodic patterns, linked to the underlying mosquito population dynamics, which are known to be driven by natural climate fluctuations. Understanding how climate dictates the timing and potential of viral transmission is essential for preparedness of public health systems and design of control strategies. While various alternative approaches have been proposed to estimate local transmission potential of such viruses, few open‐source, ready to use and freely available software tools exist. We developed the Mosquito‐borne Viral Suitability Estimator (MVSE) software package for the R programming environment. MVSE estimates the index P, a novel suitability index based on a climate‐driven mathematical expression for the basic reproductive number of mosquito‐borne viruses. By accounting for local humidity and temperature, as well as viral, vector and human priors, the index P can be estimated for specific host and viral species in different regions of the globe. We describe the background theory, empirical support and biological interpretation of the index P. Using real‐world examples spanning multiple epidemiological contexts, we further demonstrate MVSE's basic functionality, research and educational potentials.

LG) 1617 Americas with the overall average accuracy remaining above 85% even for prediction 42 windows of up to 12 weeks. 43 44 Conclusions 45Sensitivity analysis illustrated the model performance to be robust across a range of 46 features. Critically, the model performed consistently well at various stages 47 throughout the course of the outbreak, indicating its potential value at any time during 48 an epidemic. The predictive capability was superior for shorter forecast windows, and 49 geographically isolated locations that are predominantly connected via air travel. 50The highly flexible nature of the proposed modeling framework enables policy 51 makers to develop and plan vector control programs and case surveillance strategies 52 which can be tailored to a range of objectives and resource constraints. 53 54 Keywords 55 Zika, epidemic risk prediction, dynamic neural network 56 57 58 59 60 4 Background 61 The Zika virus, which is primarily transmitted through the bite of infected Aedes 62 aegypti mosquitoes (1), was first discovered in Uganda in 1947 (2) from where it 63 spread to Asia in 1960s, where it since has caused small outbreaks. In 2007 ZIKV 64 caused an island wide outbreak in Yap Island, Micronesia (3), followed by outbreaks 65 in French Polynesia (4) and other Pacific islands between 2013 ̶ 2014 where attack 66 rates where up to 70% (5-7). It reached Latin America between late 2013 and early 67 2014, but was not detected by public health authorities until May 2015 (8) and since 68 affected 48 countries and territories in the Americas (9-11). Since there is no 69 vaccination or treatment available for Zika infections (12, 13), the control of Ae. 70 aegypti mosquito populations remains the most important intervention to contain the 71 spread of the virus (14). In order to optimally allocate resources to suppress vector 72 populations, it is critical to accurately anticipate the occurrence and arrival time of 73 arboviral infections to detect local transmission (15). 74 75 Whereas for dengue, the most common arbovirus infection, prediction has attracted 76 wide attention from researchers employing statistical modelling and machine learning 77 methods to guide vector control (16-29), such real-time machine learning based 78 models do not yet exist for Zika virus. Early warning systems for Thailand, Indonesia, 79 Ecuador and Pakistan have been introduced and are currently in use (30-34). In 80 addition to conventional predictions based on epidemiological and meteorological 81 data (20, 35, 36), more recent models have successfully incorporated search engines 82 (37, 38), land use (39), human mobility information (40, 41) and spatial dynamics 83 (42-44), and various combinations of the above (45) to improve predictions. Whereas 84 local spread may be mediated by overland travel, continent wide spread is mostly 85 driven by air passenger travel between climatically synchronous regions (46-52). 86 87 The aims of our work are to 1) present recurrent neural networks for time ahead 88 predictive modelling as ...

The ongoing Zika epidemic in the Americas has challenged public health surveillance, response, and control systems. Even as the epidemic appears to be near its end in the Americas, it is unclear whether substantial Zika virus transmission may still be ongoing. This issue is exacerbated by large discrepancies in local case reporting and significant delays in detecting outbreaks due to surveillance gaps. To uncover locations with lingering outbreaks in the Americas, we investigated travel-associated Zika cases diagnosed in the United States and Europe to identify signatures of transmission dynamics that were not captured by local reporting. We found that a large and unreported Zika outbreak occurred in Cuba during 2017, a year after peak transmission in neighboring countries, with cases still appearing in 2018. By sequencing Zika virus from infected travelers, we show that the 2017 outbreak in Cuba was sparked by longlived lineages of Zika virus introduced from multiple places in the Americas a year prior. Our data suggest that while aggressive mosquito control in Cuba may initially have been effective at mitigating Zika virus transmission, in the absence of vaccines, herd immunity, or strong international coordination, such control measures may need to be maintained to be effective. Our study highlights how Zika virus may still be 'silently' spreading in the Americas and provides a framework for more accurately understanding outbreak dynamics.