Optimization of hepatological clinical guidelines interpretation by large language models: a retrieval augmented generation-based framework
Problem
Clinical decision support systems need accurate interpretation of medical guidelines for chronic Hepatitis C virus management.
Why it matters
Accurate guideline interpretation at the point of care affects treatment decisions for a chronic infection affecting millions.
What the AI did
Built a customized LLM framework using GPT-4 Turbo with retrieval augmented generation and prompt engineering, converting guidelines into LLM-friendly structured formats, and ran an ablation study across five experimental setups versus baseline.
Results
not stated in source
Impact
Evidence supports framework development and evaluation, but no measured accuracy figures are given in the abstract.
Record
| Venue | npj Digital Medicine |
|---|---|
| Published | 2024 |
| Work type | article |
| Authors | Simone Kresevic, Mauro Giuffrè, Miloš Ajčević, Agostino Accardo, Lory Saveria Crocè, Dennis Legen Shung |
| Citations | 201 (OpenAlex, at import time) |
| References | 49 |
| Model or method | OpenAI GPT-4 Turbo with RAG |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41746-024-01091-y |
| Open access | gold |
| Keywords | LLM, RAG, clinical guidelines, Hepatitis C, CDSS, GPT-4 |
Importance rating
real but narrower advance
CDSS guideline interpretation for chronic HCV care is a real clinical advance, but abstract gives no scale or outcome evidence beyond accuracy improvement in one framework study.