Skip to main content
SearchLoginLogin or Signup

Review 5: "Yellow Fever in Ghana: Predicting Emergence and Ecology from Historical Outbreaks"

While acknowledging the strengths of the studies, reviewers also offer constructive criticism regarding methodological clarity, data interpretation, and the need for updated references.

Published onMar 23, 2024
Review 5: "Yellow Fever in Ghana: Predicting Emergence and Ecology from Historical Outbreaks"
1 of 2
key-enterThis Pub is a Review of
Yellow fever in Ghana: Predicting emergence and ecology from historical outbreaks
Yellow fever in Ghana: Predicting emergence and ecology from historical outbreaks

Abstract Understanding the epidemiology and ecology of yellow fever in endemic regions is critical for preventing future outbreaks. Ghana is a high-risk country for yellow fever. In this study we estimate the epidemiology, ecological cycles, and areas at risk for yellow fever in Ghana based on historical outbreaks. We identify 2371 cases and 887 deaths (case fatality rate 37.4%) from yellow fever reported in Ghana from 1910 to 2022. Since implementation of routine childhood vaccination in 1992, the estimated mean annual number of cases decreased by 81% and the geographic distribution of yellow fever cases also changed. While there have been multiple large historical outbreaks of yellow fever in Ghana from the urban cycle, recent outbreaks have originated among unvaccinated nomadic groups in rural areas with the sylvatic/savanna cycles. Using machine learning and an ecological niche modeling framework, we predict areas in Ghana that are similar to where prior yellow fever outbreaks have originated based on temperature, precipitation, landcover, elevation, and human population density. We find differences in predictions depending on the ecological cycles of outbreaks. Ultimately, these findings and methods could be used to inform further subnational risk assessments for yellow fever in Ghana and other high-risk countries.Author Summary Yellow fever is a viral hemorrhagic fever transmitted by mosquitoes in Africa and South America through different ecological transmission cycles. While West Africa has had the most cases of yellow fever, less is known about the epidemiology and ecology of yellow fever among countries in this region. Ghana has had multiple yellow fever outbreaks, including a recent outbreak in 2021-2022. In this study we estimate cases and deaths due to yellow fever in Ghana, compare the ecological cycles of outbreaks, and predict future areas at risk based on prior yellow fever cases and environmental conditions. We find that the populations at risk for yellow fever in Ghana have changed over the past century and that different ecological factors influence the risk of future emergence. Understanding these changes and the nuances of yellow fever epidemiology and ecology within countries will be important for future outbreak preparedness.

RR:C19 Evidence Scale rating by reviewer:

  • Reliable. The main study claims are generally justified by its methods and data. The results and conclusions are likely to be similar to the hypothetical ideal study. There are some minor caveats or limitations, but they would/do not change the major claims of the study. The study provides sufficient strength of evidence on its own that its main claims should be considered actionable, with some room for future revision.


Review: The authors have used data on past reported cases of yellow fever (YF) in Ghana alongside several ecological covariates to understand the geographical distribution of the risk of emergence across the country. Using machine learning methods they produced maps of the habitat suitability for YF. While the emergence of YF in the sylvatic and savannah transmission cycles was found to be concentrated in the northern half of the country, when the emergence of urban YF outbreaks was included, all regions of the country contained areas at risk of YF emergence.

Yellow Fever (YF) is endemic in West Africa and causes substantial burden across the region. This study looks in detail at the situation in Ghana, collating information on past cases and outbreaks from 1910 onwards. Using environmental and ecological covariates the authors use machine learning approaches to fit models to the locations of past outbreaks and generate a measure of “YF habitat suitability” across the country, describing the risk of emergence of future outbreaks. Essentially this identifies the areas in the country that are ecologically like those where outbreaks have been recorded in the past. The methodology appears sound. The maps thus generated might be useful to public health and policy makers to identify areas where surveillance or prevention efforts should be targeted, particularly when considered together with further information such as local variations on existing vaccination coverage, and locations of nomadic and other hard-to-reach populations. 

Similar approaches have been employed for both YF and other diseases, however, the present study focuses on Ghana only, and the authors have taken great care to capture epidemiological detail in the underlying dataset, for instance identifying the different epidemiological transmission cycles (urban vs sylvatic/savannah) leading to more robust estimates.

While clearly a lot of care has been taken to curate this dataset, it would be good to discuss the impact of the choices taken in inclusion/exclusion of cases/outbreaks (eg lines 150-155, 184 – 187).

The authors highlight that the introduction of routine childhood vaccination in 1992 has changed the potential for outbreaks over time. Additionally, changes population distribution, land use and climate also have the potential to alter the epidemiology and resulting patterns of occurrence. The authors reflect this by providing maps of the YF occurrence for the pre- and post-1992 periods separately. However, the model fitting appears to have used the full dataset, and it is unclear whether any of the covariates varied temporally to fit a full spatio-temporal model, or if all covariates were averaged through time and used as purely spatial covariates? This seems more likely given that the fitted maps of yellow fever habitat suitability are static. This is a limitation that should be clearly stated and discussed.

I would like to see some more data, probably presented as supplementary information:

  • a map of the locations the model was fitted to

  • a full list of the covariates explored (what are all the bio-climatic variables and land cover categories?)

  • maps of all covariates.

In summary, this preprint shows that the risk of yellow fever (YF) varies geographically throughout Ghana, and patterns of past outbreaks highlight a more frequent emergence in the northern half of the country. By comparing the ecological situation in areas where there have been outbreaks recorded in the past with the rest of the country, this study has generated maps that quantify the YF habitat suitability – the more similar a place is to the areas where YF has been recorded in the past, the higher the chance that YF might arise at this location.

1 of 5
No comments here
Why not start the discussion?