Performance of an artificial intelligence model for evaluation of unnecessary central lines, Northern California 2025
We used a large language model integrated in the electronic health record to evaluate unnecessary central lines. It had a 16% sensitivity and 99% specificity for detecting unnecessary lines. Although it missed many unnecessary lines, the high specificity suggests potential as a tool where human review is not feasible.
Authors
- Wajeeha Tariq (ORCID: https://orcid.org/0000-0001-8387-4309)
- Jenna Maco Wick (ORCID: https://orcid.org/0009-0001-0031-1832)
- Mindy M. Sampson
- Sean Lau
- Jorge L. Salinas
- Apoorva Bhaskara (ORCID: https://orcid.org/0009-0004-5315-019X)
Institutions
- Stanford Health Care (US)
- Stanford University (US)
Publication Details
- Journal
- Infection Control and Hospital Epidemiology
- Published
- 2026-04-27
- DOI
- https://doi.org/10.1017/ice.2026.10462
- Primary Topic
- Machine Learning in Healthcare
- Type
- article
- Field-Weighted Citation Impact
- 0.00