Updates to the zoonotic niche map of Ebola virus disease in Africa

David M. Pigott, Anoushka Millear, Lucas Earl, Chloe Morozoff, Barbara A. Han, Freya M. Shearer, Daniel J. Weiss, Oliver J. Brady, Moritz U. G. Kraemer, Catherine L. Moyes, Samir Bhatt, Peter W. Gething, Nick Golding, Simon I HayeLife 2016 · DOI

TL;DR

This paper updates the zoonotic niche map for Ebola Virus Disease (EVD) in Africa by incorporating new occurrence data and expanding the consideration of potential bat reservoir species. The updated map provides a more refined understanding of areas at risk for zoonotic EVD transmission, crucial for strengthening surveillance and response efforts.

What Problem Does This Paper Address?

The 2014 West African Ebola epidemic highlighted critical gaps in our understanding of EVD's epidemiology, particularly concerning its zoonotic origins and the role of animal reservoirs. While a previous zoonotic niche map existed, it was based on limited data, specifically focusing on only three bat species known to be RNA-positive for Ebolavirus. This study addresses the need for a more comprehensive and up-to-date assessment of EVD risk by:

  • Incorporating new occurrence data: Including recent human and animal EVD cases that have emerged since the previous mapping effort.
  • Expanding the scope of potential reservoirs: Moving beyond just RNA-positive bats to include species with serological evidence of infection and those identified through trait-based machine learning as potential filovirus carriers. This acknowledges that identifying reservoirs is complex and that current knowledge might be incomplete.
  • Improving predictive accuracy: By integrating these new data and a broader range of potential reservoir species, the study aims to create a more robust and accurate map of areas at risk for zoonotic EVD transmission.

Methods

The study employs a species distribution modeling approach, specifically using boosted regression trees (BRTs). This is a standard technique for predicting the geographic distribution of species based on environmental variables. The key novelties and refinements in this study's methodology include:

  • Updated Occurrence Data: The database of EVD occurrences was updated with six new animal records and one new human outbreak record.
  • Expanded Bat Reservoir Species Groupings: Potential bat reservoir species were categorized into three groups based on the strength of evidence:
  • Group 1: Bats with confirmed Ebolavirus RNA detection (the original three species).
  • Group 2: Bats with serological evidence of EVD infection.
  • Group 3: Bats identified through trait-based machine learning as similar to known filovirus carriers.
  • Consolidated and Individual Bat Layers: The BRT model was run in two configurations:
  • Using three consolidated layers representing the suitability for each of the bat groupings.
  • Using individual layers for each of the identified bat species.
  • Covariate Selection: The model incorporated environmental covariates such as elevation, mean evapotranspiration rate, enhanced vegetation index (EVI), and land surface temperature (LST).
  • Population-at-Risk Estimation: A thresholding approach was used to convert the continuous suitability surface into a binary at-risk map, and human population density was used as a proxy for spillover frequency in at-risk areas.

Datasets

The study utilized the following datasets:

  • Human and Animal Occurrence Data: New human and animal EVD occurrence records were incorporated, supplementing the existing database.
  • Bat Occurrence Records: These were sourced from the Global Biodiversity Information Facility (GBIF) and supplemented by literature searches.
  • Environmental Covariates: Data on elevation, mean evapotranspiration rate, enhanced vegetation index (EVI), and land surface temperature (LST) were used.
  • Human Population Data: Gridded population data for Africa (WorldPop Project) was used to estimate populations at risk.

A new data portal for sharing these updated maps is also discussed, aiming to facilitate wider access and use of the generated risk maps.

Key Insights and Ideas

  • Broader Bat Diversity Matters: Including a wider range of bat species, not just those with direct RNA detection, significantly influences the predicted zoonotic niche. This suggests that our understanding of EVD reservoirs is still evolving and that a more inclusive approach is necessary.
  • Geographic Expansion of Risk: The updated map shows an expansion of areas predicted to be at risk for zoonotic EVD transmission, notably including Kenya. This highlights the dynamic nature of disease risk and the importance of continuous data updates.
  • Importance of Environmental Factors: Enhanced Vegetation Index (EVI) remains a strong predictor of zoonotic EVD transmission, underscoring the link between ecological conditions and disease risk.
  • Data Integration is Key: The study demonstrates the feasibility and value of integrating new occurrence data and expanding the consideration of reservoir species into existing modeling frameworks. This iterative approach is crucial for maintaining up-to-date risk assessments.

Key Results

  • 23 Countries Identified as At-Risk: Based on a thresholding approach that captures 95% of the occurrence dataset, 23 countries were identified as being at risk for zoonotic EVD transmission.
  • Improved Model Fit: The updated model showed comparable performance (similar AUC values) to the previous iteration, indicating that the inclusion of new data and expanded reservoir species did not compromise the model's predictive power.
  • Shift in Predictor Importance: The relative importance of Group 1 bat distributions shifted from fifth to second among the top predictors, indicating their increased influence in the updated model.
  • Specific Regions Show Increased Suitability: The map update indicated increased environmental suitability for zoonotic EVD transmission in regions like Cameroon, Gabon, Republic of Congo, and mainland Equatorial Guinea.

Limitations and Future Work

  • Environmental Suitability vs. Spillover: The authors reiterate that environmental suitability does not automatically equate to spillover events. Other factors, such as human behavior, population density, and the susceptibility of other animals, play a crucial role.
  • Data Limitations: The study acknowledges that the identification of reservoirs is a complex process, and isolation of Ebolavirus from living animals has been rare. The potential for false negatives in sampling efforts remains.
  • Future Work: The paper suggests that future work should focus on:
  • Rigorously identifying reservoir species: Continued efforts are needed to pinpoint the exact species and their roles in sustaining viral transmission.
  • Incorporating human behavior and other factors: Future models could benefit from including covariates that account for human behavior, patterns of susceptibility in other animals, and land-use changes to provide a more nuanced understanding of spillover risk.
  • Developing more robust data collection: The potential for different pathogenic species of Ebolavirus to have differing distributions cannot be explored more rigorously due to insufficient data.
  • Utilizing the new data portal: The developed data portal is intended to facilitate further research and inform policy by providing accessible risk maps.

Summary updated

Abstract

As the outbreak of Ebola virus disease (EVD) in West Africa is now contained, attention is turning from control to future outbreak prediction and prevention. Building on a previously published zoonotic niche map (Pigott et al., 2014), this study incorporates new human and animal occurrence data and expands upon the way in which potential bat EVD reservoir species are incorporated. This update demonstrates the potential for incorporating and updating data used to generate the predicted suitability map. A new data portal for sharing such maps is discussed. This output represents the most up-to-date estimate of the extent of EVD zoonotic risk in Africa. These maps can assist in strengthening surveillance and response capacity to contain viral haemorrhagic fevers.
BibTeX
@inproceedings{pigott2016updates,
  author = {David M. Pigott and Anoushka Millear and Lucas Earl and Chloe Morozoff and Barbara A. Han and Freya M. Shearer and Daniel J. Weiss and Oliver J. Brady and Moritz U. G. Kraemer and Catherine L. Moyes and Samir Bhatt and Peter W. Gething and Nick Golding and Simon I Hay},
  title = {{Updates to the zoonotic niche map of Ebola virus disease in Africa}},
  booktitle = {eLife},
  year = {2016},
  doi = {10.7554/elife.16412},
}
Figure