Using secure artificial intelligence agents integrated within the electronic medical record for the evaluation of blood culture appropriateness—Northern California, 2025
We evaluated large language model (LLM)-based agents integrated with the electronic medical record to assess blood culture appropriateness. While sensitivity was high, specificity remained low. Performance was shaped by prompt phrasing, sycophantic behavior, and semantic triggers, reflecting both the potential and limitations of LLMs in real-world clinical decision support.
Authors
- Guillermo Rodríguez-Nava (ORCID: https://orcid.org/0000-0001-9826-7050)
- Jorge Salinas (ORCID: https://orcid.org/0000-0002-9193-820X)
- Mindy Marie Sampson (ORCID: https://orcid.org/0000-0001-5724-5875)
- Eugenia Miranti (ORCID: https://orcid.org/0000-0002-0654-8937)
- Wajeeha Tariq (ORCID: https://orcid.org/0000-0001-8387-4309)
- Timothy Keyes (ORCID: https://orcid.org/0000-0003-0423-9679)
- Erika Paola Viana Cardenas (ORCID: https://orcid.org/0000-0002-6399-890X)
- Nerissa Ambers (ORCID: https://orcid.org/0009-0008-9878-1824)
Institutions
- Stanford Health Care (US)
- Stanford University (US)
Publication Details
- Journal
- Infection Control and Hospital Epidemiology
- Published
- 2025-11-11
- DOI
- https://doi.org/10.1017/ice.2025.10349
- Primary Topic
- Bacterial Identification and Susceptibility Testing
- Type
- article
- Field-Weighted Citation Impact
- 0.00