The quality and safety of using generative AI to produce patient-centred discharge instructions
Patient-centred instructions on discharge can improve adherence and outcomes. Using GPT-3.5 to generate patient-centred discharge instructions, we evaluated responses for safety, accuracy and language simplification. When tested on 100 discharge summaries from MIMIC-IV, potentially harmful safety issues attributable to the AI tool were found in 18%, including 6% with hallucinations and 3% with new medications. AI tools can generate patient-centred discharge instructions, but careful implementation is needed to avoid harms.
Authors
- Kristian Stanceski (ORCID: https://orcid.org/0000-0001-5165-4096)
- Julie Ayre (ORCID: https://orcid.org/0000-0002-5279-5189)
- Sarita Lo (ORCID: https://orcid.org/0000-0002-5272-9958)
- Marguerite Tracy (ORCID: https://orcid.org/0000-0003-1247-6412)
- Vasi Naganathan (ORCID: https://orcid.org/0000-0001-7243-0796)
- Adam G. Dunn (ORCID: https://orcid.org/0000-0002-1720-8209)
- Shiyin Zhong (ORCID: https://orcid.org/0009-0007-2485-0332)
- Sam Khadra
- Jinman Kim (ORCID: https://orcid.org/0000-0001-5960-1060)
- X. Zhang (ORCID: https://orcid.org/0009-0006-6798-1204)
- Linda Koria
Institutions
- The University of Sydney (AU)
- Sutherland Hospital (AU)
- Royal Prince Alfred Hospital (AU)
- Concord Repatriation General Hospital (AU)
- Blacktown & Mount Druitt Hospital (AU)
- Sydney Local Health District (AU)
Publication Details
- Journal
- npj Digital Medicine
- Published
- 2024-11-20
- DOI
- https://doi.org/10.1038/s41746-024-01336-w
- Citations
- 25
- Primary Topic
- Machine Learning in Healthcare
- Type
- article
- Field-Weighted Citation Impact
- 8.68