An interpretable framework for assessing zoonotic spillover risk

Zoonotic diseases pose significant threats to global health, as evidenced by the COVID-19 pandemic. Despite their impact, our understanding of pathogen spillover mechanisms remains incomplete due to data limitations and methodological challenges. Here we integrate machine learning approaches with ecological models to predict and quantify spillover risks globally. We first systematically assess current limitations in ecological epidemiological modeling, then develop a framework that utilizes pathogen emergence events as critical indicators for spillover risk. Through ensemble machine learning combined with causal inference, we map global spillover risk patterns and identify key climatic, environmental, and socioeconomic drivers. We further apply this framework to tick-borne disease systems across Europe, demonstrating that hierarchical environmental constraints—from macroclimatic phenology to landscape configuration—differentially shape vector abundance and disease prevalence. We show that development intensity sets boundaries for tick population establishment, while landscape features determine realized abundance within climatically suitable areas, with effect magnitudes varying across biogeographic contexts. This interdisciplinary approach advances spillover risk assessment and provides evidence-based guidance for One Health strategies integrating environmental, vector, and human health surveillance.