Causal Inference Unveils How Forest Coverage Mitigates Excess of Snakebite Cases During Rainfall Seasons in Colombia

Abstract

Background: Snakebite envenoming, a neglected tropical disease, poses a significant threat in tropical countries with limited antivenom availability, especially affecting rural populations. Understanding risk factors for increased cases is vital for public health response. This study employs causal inference to assess the association between rainfall and snakebite surges in Colombia.

Methods: Utilizing monthly snakebite cases data (2007–2021), we conducted an ecological study employing a causal inference approach to estimate the association between rainfall and excess snakebite cases. Additionally, we included nine atmospheric and oceanic indices and forest coverage as confounders, and rural GDP as an effect modifier. The causal association was estimated using machine learning models.

Findings: High rainfall significantly causes an excess in snakebite cases in Colombia. Rural GDP was identified as a collider variable, thus its effect over causal inference on the directed acyclic graph was deprecated. A one standard deviation increment in monthly rainfall (134.65 mm) correlated with a 2.1% rise in excess cases (95% CI: 1.3 – 2.9). Forest coverage exhibited an inverse relationship with the impact of rainfall on excess cases, where this impact is negative in regions with dense forests.

Interpretation: Our findings stress the need for tailored health strategies during rainfall seasons. Increased rainfall elevates snakebite risk, only in areas with low forest presence. High forest coverage changes the dynamics, emphasizing the importance of addressing deforestation to deeply understand snakebite dynamics. Our study informs proactive planning for the evolving burden of snakebite incidents associated with climate change and habitat transformation.

Keywords: Snakebite, Climate, Rainfall, One health, Epidemiology, Causal inference, Machine learning, Neglected tropical disease