This paper introduces a benchmark to evaluate the epidemiology of Large Language Models, specifically focusing on their observational distribution knowledge, which is crucial for understanding and improving their real-world applicability.