Medication information extraction using local large language models
Our findings show that fine-tuned local open-source generative LLMs outperform SOTA methods for medication information extraction, delivering high performance with limited time and IT resources in a real-world clinical setup, and demonstrate their effectiveness on both English and German data. Applying Shapley values improved prediction transparency, supporting informed clinical decision-making.
Authors
- Anette Frank (ORCID: https://orcid.org/0000-0003-4706-9817)
- Dominic M. Schwab
- Christina Kiriakou
- Nicolas A. Geis (ORCID: https://orcid.org/0000-0003-3431-3104)
- Christoph Dieterich (ORCID: https://orcid.org/0000-0001-9468-6311)
- Phillip Richter-Pechanski (ORCID: https://orcid.org/0000-0003-0121-373X)
- Marvin Seiferling
Institutions
- Heidelberg University (DE)
- University Hospital Heidelberg (DE)
Publication Details
- Journal
- Journal of Biomedical Informatics
- Published
- 2025-08-22
- DOI
- https://doi.org/10.1016/j.jbi.2025.104898
- Citations
- 2
- Primary Topic
- Topic Modeling
- Type
- article
- Field-Weighted Citation Impact
- 4.96
Funders
- Deutsche Forschungsgemeinschaft
- Bundesministerium für Bildung und Forschung
- Klaus Tschira Stiftung