Modelling the large-scale yellow fever outbreak in Luanda, Angola, and the impact of vaccination

Background: Yellow fever (YF), transmitted via bites of infected mosquitoes, is a life-threatening viral disease endemic to tropical and subtropical regions of Africa and South America. YF has largely been controlled by widespread national vaccination campaigns. Nevertheless, between December 2015 and August 2016, YF resurged in Angola, quickly spread and became the largest YF outbreak for the last 30 years. Recently, YF resurged again in Brazil (December 2016). Thus, there is an urgent need to gain better understanding of the transmission pattern of YF. Model: The present study provides a refined mathematical model, combined with modern likelihood-based statistical inference techniques, to assess and reconstruct important epidemiological processes underlying Angola’s YF outbreak. This includes the outbreak’s attack rate, the reproduction number (R 0), the role of the mosquito vector, the influence of climatic factors, and the unusual but noticeable appearance of two-waves in the YF outbreak. The model explores actual and hypothetical vaccination strategies, and the impacts of possible human reactive behaviors (e.g., response to media precautions). Findings: While there were 73 deaths reported over the study period, the model indicates that the vaccination campaign saved 5.1-fold more people from death and saved from illness 5.6-fold of the observed 941 cases. Delaying the availability of the vaccines further would have greatly worsened the epidemic in terms of increased cases and deaths. The analysis estimated a mean R 0 ≈ 2 . 6 - 3 . 4 and an attack rate of 0.09-0.15% (proportion of population infected) over the whole period from December 2015 to August 2016. Our estimated lower and upper bounds of R 0 are in line with previous studies. Unusually, R 0 oscillated in a manner that was “delayed” with the reported deaths. High recent number of deaths were associated (followed) with periods of relatively low disease transmission and low R 0, and vice-versa. The time-series of Luanda’s YF cases suggest the outbreak occurred in two waves, a feature that would have become far more prominent had there been no mass vaccination. The waves could possibly be due to protective reactive behavioral changes of the population affecting the mosquito population. The second wave could well be an outcome of the March-April rainfall patterns in the 2016 El Niño year by creating ideal conditions for the breeding of the mosquito vectors. The modelling framework is a powerful tool for studying future YF epidemic outbreaks, and provides a basis for future vaccination campaign evaluations. Author summary: An epidemic model for the transmission of yellow fever virus (YFV) in urban areas is formulated and implemented to study the 2016 yellow fever (YF) outbreak in Luanda, Angola. We explore the complex vector-host dynamics of this system taking into account mosquito abundance, vaccination and asymptomatic infections in the human population, that are generally not included in other modelling studies of YF. The model successfully fits the time series of weekly reported YF cases and deaths during the epidemic. This allows us to study the impact of the vaccination campaign in Luanda and hypothetical “delayed vaccination” scenarios. The transmission of YFV appeared to be oscillatory having a wave-like pattern in the basic reproduction number (R 0 ( t )). The oscillations are hypothesized to be due to human reaction to the reported deaths, as has been noted for other human infectious diseases, and the second wave also possibly due to El Niño rainfall patterns. We conclude that the lives saved due to the vaccination campaign before August 2016 should have been approximately 370 (i.e., approximately five-fold of the observed 73 deaths), and would have been far larger extrapolating beyond August 2016.