South KoreaHealthcare and life sciences2024 – ongoingSeverance Hospital
Y-KNOT emergency department discharge documentation
Severance's Y-KNOT reduced median discharge-note writing time to 32.0 seconds in testing
Problem
Emergency physicians had to assemble discharge notes from clinical records under time pressure.
Approach
Severance and its partners developed an on-premises bilingual language model connected to electronic health records. The project began in June 2024, with routine clinical service starting that November. Physicians could request drafts within their existing record system and edit them before saving.
Results
Takeaway
The timing result came from a controlled evaluation involving 6 physicians and 50 cases. Draft errors remained, including an invented procedure that one physician did not remove.
Categories
- Size
- Large enterprise
- Industry
- Healthcare and life sciences
- Business function
- Internal work and knowledge managementProduction and quality
- Innovation type
- Process innovation
- AI technology
- Generative AI
- Adoption
- Partner co-development
- Evidence
- Independent study
Sources
Source summaries
The hospital-affiliated researchers compared manual and AI-assisted documentation in controlled testing. Median writing time fell from 69.5 to 32.0 seconds; an audit also identified omissions and an invented procedure.
The project team documented the hospital's on-premises model and integration with electronic records. It dated project initiation to June 2024 and routine clinical service to November 2024.
Full details are in the original sources.