Responding to the threat of urban yellow fever outbreaks.

When in April, 2016, WHO declared the yellow fever epidemic in Angola a global threat, it was because yellow fever appeared in Luanda, the capital city of Angola, causing a rapidly spreading urban outbreak due to the massive movement of people to and from the city and easy access to international airports, with daily connections to Asia, Europe, and the Americas. Nearly 45 years had elapsed since a similar urban yellow fever epidemic occurred in Angola in 1971 (a smaller one occurred in 1988); in that interval, urbanisation has increased at record rates, with more than 62% of the population now living in urban areas. For reasons that are still poorly understood, the yellow fever virus, which is maintained in a transmission cycle involving non-human primates and arboreal mosquitoes, crosses into a far more threatening human-to-human transmission cycle involving urban and domestic Aedes aegypti mosquitoes. The distribution of A aegypti is now the widest ever recorded and it is extensive in all continents, including parts of North America and Europe, with more than 3 billion people at risk. Urbanisation with resulting increased population densities, further enhanced by man-made larval habitats, amplifi es A aegypti-transmitted diseases, including those caused by the dengue, chikungunya, and Zika viruses. Among these viral infections, yellow fever is of major concern because the lethality of this haemorrhagic fever is 20–50%, rivalling that of Ebola virus disease. Of particular concern is that urban yellow fever has the potential for rapid spread via international travellers to vulnerable countries where A aegypti mosquitoes are present. Indeed, for the fi rst time, such spread happened during the Angola outbreak when travellers infected with yellow fever in Angola entered China between March and April, 2016, thereby putting 1·8 billion largely unvaccinated people in Asia at risk. Fortunately, no autochthonous cases occurred. In late 2015 and the fi rst months of 2016, the yellow fever outbreak in central Africa had a high reproductive rate (4·8 new people infected for every case). Although an eff ective vaccine, which protects against infection within 7–10 days after vaccination, has been available since the 1930s, implementation of an emergency vaccination campaign to contain the rapidly expanding outbreak was hampered by limited vaccine supplies and problems in the delivery of vaccinations. In such a setting it is essential to identify the areas at greatest risk of infection, to inform vaccine prioritisation decisions. Consequently, the analyses by Moritz Kraemer and colleagues reported in The Lancet Infectious Diseases provide an important contribution with many practical implications. Although a WHO working group had previously identifi ed a set of mainly ecological criteria relevant to the transmission of yellow fever virus and systematically applied these criteria to classify areas with risk for transmission of yellow fever virus, the model developed by Kraemer and colleagues adds mobility patterns and demographic determinants that govern transmission and spread of the virus in the region. The investigators used standard logistic models that discriminated districts with high risk of invasion from others. Notably, population density was a dominant predictive factor for onward transmission, corroborating previous fi ndings. The spread of yellow fever in Angola was driven by high population density, including in locations that were distant from the origin of the outbreak in Luanda. Furthermore, the team captured diff erent aspects of connectivity and were able to infer regular daily commuting patterns, longer-term movements, and general human diff usion processes. Their gravity model assumed that relative fl ow between districts is a log-linear function of the population sizes of the districts and the distance between them, thereby emphasising large population centres such as the capital cities Luanda (Angola) and Kinshasa (DR Congo), which were the epicentres of the epidemic. A radiation model additionally took account of the draw from other populations within the same radius of the districts considered, as well as the population sizes and distance of the origin and destination locations. Models based on travel volumes as shown by Kraemer and colleagues are increasingly used to predict international spread and identify the most vulnerable receiving areas or countries for infectious diseases such as infl uenza H1N1, polio, dengue, Zika virus, and yellow fever, and should become the basis for similar studies aiming to predict regional and international spread. The modelling techniques used by Kraemer and colleagues allowed the analysis of near real-time data to inform the control of an ongoing outbreak. If, in midFebruary, the 50 districts with the highest calculated probability of infection had been targeted, their model would have correctly identifi ed 27 (84%) of the 32 districts Lancet Infect Dis 2016