Use of artificial intelligence to support guideline-compliant antibiotic therapy.

On May 29, 2025, Richard Klein presented the latest research findings on the use of large language models (LLM) and retrieval-augmented generation (RAG) in medicine at the 96th Annual Meeting of the German Society of Otorhinolaryngology, Head and Neck Surgery in Frankfurt am Main.

The presentation focused on a system that supports physicians in selecting guideline-compliant antibiotic therapy. While classic LLMs generate answers, they also tend to produce incorrect or incomprehensible results (“hallucinations”), whereas RAG enables the targeted integration of external knowledge sources. This results in reliable, reproducible, and comprehensible recommendations that meet the highest standards of quality and patient safety.

The advantages of RAG technology were particularly emphasized:

• high data protection compliance,
• complete control over the knowledge base,
• transparency in decision-making,
• and independence from external providers.

The invitation to give the presentation came at the suggestion of Prof. Orlando Guntinas-Lichius, with whom the project was jointly initiated. The successful application for the presentation slot underscores the high relevance of the topic for the future of clinical practice.