Harnessing Large Language Models in Neonatal IVH: Exploring RAG Methodology for Prognostic Variable Discovery

The authors aimed to evaluate the ability of large language models (LLMs) to autonomously synthesize literature and extract prognostic variables related to neonatal intraventricular hemorrhage (IVH) outcomes. Their pilot study found that LLMs could identify key predictors such as gestational age and birth weight, while also highlighting the need for human validation due to potential inaccuracies in the data synthesis process. The findings suggest a foundation for developing AI-assisted clinical decision support tools, although significant research gaps remain in understanding resolution predictions and complications.