Venomous snakebite caused 63,400 deaths (95% UI 38,900–78,600) and 2.94 million YLLs (1.79 million–3.74 million) in 2019, which makes it the deadliest NTD according to GBD 2019.11 Over time, the global age-standardized rate of death has decreased by 36% (2–49), which shows progress; however, this annual rate of change would be insufficient to accomplish WHO’s 2019 goal of halving the burden by 2030.4
The greatest venomous snakebite mortality occurred in south Asia, and specifically India, where we estimated over 50,000 deaths occurred in 2019. These estimates are consistent with previous research conducted with verbal autopsy mortality surveys, which were the source of data in India in our process as well.9,10,17 The high mortality in India is an example of ecological factors, socioeconomic factors, and health system shortcomings intersecting to create a vulnerable population to preventable snakebite death. After venomous snakebite occurs, the probability of death increases if antivenom is not administered within six hours.18 However, in south Asia, many seek out traditional healers or attend clinics with insufficient education about how to treat venomous snakebites or lack the antivenom to administer life-saving treatment.18–21 Victims who do reach a hospital often have insufficient access to dialysis, ventilators, and blood transfusions, which are essential to deal with the complications of envenoming.20,22 Interventions to secure more rapid antivenom delivery need to be coupled with preventive strategies like increased education and health system strengthening in rural areas.
Sub-Saharan Africa had the second greatest mortality with 6790 deaths (95% UI 5040–10,100) and 314,000 YLLs (219,000–521,000), equivalent to age-standardized rates of 1.2 deaths (0.9–1.6) per 100,000 and 36.9 YLLs (27.3–54.6) per 100,000. In the WHO’s 2019 updated Strategy for Prevention and Control of Snakebite Envenoming in sub-Saharan Africa, updated and precise epidemiological data were outlined as a need moving forward to better guide appropriate and efficient implementation of antivenom interventions.23 However, there are no robust verbal autopsy or vital registration systems in the area; precise measurement of deaths is difficult and statistical modeling is required. Recently, political determination to curb venomous snakebites in sub-Saharan Africa has improved, and we hope these estimates prompt further support for antivenom distribution and detailed epidemiological studies on the extent of venomous snakebite in sub-Saharan Africa.
Our ensemble modelling framework allowed us to test multiple covariates for their association with venomous snakebite mortality providing important insights on the disease epidemiology. Environmental indicators such as living at a lower elevation and latitude and socioeconomic indicators like education had strong negative associations with venomous snakebite mortality (Appendix Figure 4). We found education had a more negative association for males, while urbanicity was more strongly negative for females. These findings aligned with previous research that reported higher venomous snakebite mortality in females than males in rural areas.10 We show that at a population level, interventions for rural areas focused on antivenom delivery should be supplemented with education for agricultural workers to increase awareness of high-risk behaviors and mitigation strategies. With better epidemiological data, more data-driven implementation of proven interventions can be achieved, like the use of education, rapid emergency transport for agricultural workers, as well as rigorous evaluation of innovative interventions like antivenom delivery via drones to at-risk rural locations.24–27
When paired with the recent analysis by Longbottom et al. that mapped the vulnerability to snakebite envenoming, our high-level estimates present a complementary assessment of the drivers behind venomous snakebite mortality, and especially highlight gaps in antivenom access in many areas of the world.28 In some places, their results intersected with locations we estimated to have high mortality rates, such as central and eastern sub-Saharan Africa, which Longbottom et al. estimated have significant vulnerability due to poor health system infrastructure and the presence of snakes. Conversely, we found that high rates of mortality also occur in areas that Longbottom et al. did not estimate to have high vulnerability, such as India. This is likely due to the existence of antivenom for the “big four” snakes (Bungarus caeruleus, Daboia russelii, Echis carinatus, Naja naja) that cause the majority of envenomations in the country, while the vulnerability estimates were focused on exposure to snakes that do not have antivenom treatments.18,28 Our mortality estimates demonstrate that venomous snakebite death depends on more than just the existence of antivenom, but also its dissemination to rural areas and the health system capacity of the area to provide supportive care to victims with secondary complications such as neurotoxic respiratory failure or acute kidney injury requiring dialysis.29 Future studies should improve the resolution of mortality at a more detailed spatial level, and combine metrics of human-snake interactions, health system capabilities, and disease burden. A more granular spatial level will also reveal disparities not captured in this analysis. Greater temporal resolution incorporating the seasonality of venomous snakebites, especially in south Asia where the incidence of bites increases during the rainy season, would also be useful for decision makers.
Limitations and strengths
In this analysis, we incorporated an extensive amount of ICD-coded VR and VA data that has previously not been utilized in global snakebite estimates. However, even in this dataset there was sparsity across some locations, especially in sub-Saharan Africa and Southeast Asia where there are few robust in-country data reporting systems. Despite sparse data, our estimate of 6790 deaths (95% UI 5040–10,100) in sub-Saharan Africa aligns closely with the meta-analysis by Chippaux and colleagues, which estimated there were 7331 (5149—9568) annual deaths.7 Both studies have the same problems of data scarcity, are likely underestimates of the true number of deaths, and emphasize the urgent need for better epidemiological assessments to provide a more accurate assessment of the true disease burden due to venomous snakebite in sub-Saharan Africa, South Asia, and Southeast Asia.
Verbal autopsy and vital registration are also both imperfect methods for counting venomous snakebite deaths and represent another limitation in our study. We could still be underestimating the true magnitude of death if the distinctive signs of snakebite, or the snake itself, were not seen when the bite occurred. For example, in Cambodia, only a single verbal autopsy study including venomous animal mortality has been conducted to our knowledge in the country,30 which did not find a single death due to venomous snakebite, despite the presence of multiple venomous snakes in the country.14 This highlights the need for improved focused venomous snakebite surveillance in areas where venomous snakes are known to be endemic.
Alternatively, official death statistics have been shown to miss many venomous snakebite deaths or miscode them as another cause. Studies comparing verbal autopsy community-based studies and official records frequently find that official records undercount the number of deaths that actually occurred.9,10,21 Acknowledging these limitations in vital registration data, we attempted to use post-processing steps like redistribution of ill-defined causes of death to attempt to account for underreporting.31 However, given that many venomous snakebite deaths occur in rural settings in countries without strong cause of death surveillance or vital registration systems, underreporting likely still occurred and our estimates are potentially underestimates, given the limitations of the epidemiological data.
To improve future studies, questions related to venomous snakebites should be incorporated into regular health surveys that are already being conducted across sub-Saharan Africa and south Asia. Injury surveillance, such as the use of District Health Information System 2, has also shown promise and could be adapted to snakebites to create real-time epidemiological information.32 Increased collaboration between researchers and local health institutions should be prioritized to bolster the availability of data, demonstrate the unmet need for antivenom, and rigorously monitor and evaluate interventions.
Our analysis also relied on the WHO venomous snake distribution map to decide which locations could reliably be identified as having venomous snakes of medical importance and which did not. It was important for our results to be ecologically feasible, and this database represented the most complete list of venomous snakes capable of causing mortality that we could find. However, while it is updated iteratively, it is not complete and only contains approximately 200 venomous snakes deemed medically important, out of 600 venomous snakes. While these other 400 snakes may not cause fatalities regularly, they could cause fatal envenomation in rare cases. If a country only contained one of these 400 venomous snakes that was capable of a rare fatal envenomation and not one of the 200 medically important snakes, then we would be erroneously zeroing out that location. For example, there is the Solomons Coral Snake (Salmonelaps par) in Solomon Islands, that has no recorded fatal envenomations but there are case reports of near lethal bites.33 Conversely, there were countries where we had official health statistics data that recorded an ICD-coded death due to venomous snakebite in Chile and New Zealand. Based on review of the WHO venomous snake distribution database and venomous snake habitats, we agreed with the WHO venomous snake distribution database that there were no endemic venomous snakes despite these recorded deaths.