Rethinking sepsis prediction in the era of large language models

Automated sepsis prediction models have historically struggled due to the heterogenous clinical presentation of sepsis. Unlike traditional methods, large language models (LLMs) offer a novel way to integrate clinical context from text-based data into clinical prediction tasks, leading to recent groundbreaking performance in sepsis prediction. As LLMs become increasingly powerful, health systems must rethink their approach to clinical and data workflows to effectively integrate LLMs into their clinical environments.