A mechanistic model of snakebite as a zoonosis: Envenoming incidence is driven by snake ecology, socioeconomics and its impacts on snakes

Abstract

Snakebite is the only WHO-listed, not infectious neglected tropical disease (NTD), although its eco-epidemiology is similar to that of zoonotic infections: envenoming occurs after a vertebrate host contacts a human. Accordingly, snakebite risk represents the interaction between snake and human factors, but their quantification has been limited by data availability. Models of infectious disease transmission are instrumental for the mitigation of NTDs and zoonoses. Here, we represented snake-human interactions with disease transmission models to approximate geospatial estimates of snakebite incidence in Sri Lanka, a global hotspot. Snakebites and envenomings are described by the product of snake and human abundance, mirroring directly transmitted zoonoses. We found that human-snake contact rates vary according to land cover (surrogate of occupation and socioeconomic status), the impacts of humans and climate on snake abundance, and by snake species. Our findings show that modelling snakebite as zoonosis provides a mechanistic eco-epidemiological basis to understand snakebites, and the possible implications of global environmental and demographic change for the burden of snakebite.

Fig 1. The snakebite sieve, an adaptation of the conceptual framework for zoonoses of Plowright et al. [13], shows how different factors align and result in what we recognise as snakebite burden.

Species and occupational identities are meant to be schematic and do not represent the characteristics of our study system.

Fig 2. Snakebite patterns predicted by the spatial model (median of posterior estimates).

Insets in the top right corner of each map show a scatter plot of our model estimates and Ediriweera et al. [23]. Right hand side panels, from top to bottom show the autoregressive random effects, root mean square error for each Sri Lankan district between our model and data used, and the residuals with original survey data analysed by Ediriweera et al. [23].

Fig 3. Partial responses of snakebite incidence to human population density and snake abundance per land cover class in the model of snakebites.

The contour lines in each 2D plot indicate the expected incidence level highlighted with the colour scale at that combination of human population and snake abundance.

Fig 4. Envenoming patterns predicted by the spatial model of the probability that a bite results in envenoming (median of posterior estimates).

The top right corner of the left side maps show the relationship with the data used to fit the models, and the correlation coefficient adjusted for spatial autocorrelation. Right hand side panels, from top to bottom show the conditional autoregressive random effects, root mean square error for each Sri Lankan district between our model and data used, and the residuals with the original survey data analysed by Ediriweera et al. [23].