LLM Discharge Summaries Show High Adoption in Dutch Hospital
A pilot study utilizing mixed methods was conducted at an academic hospital in the Netherlands to assess an Electronic Health Record (EHR)-integrated Large Language Model (LLM) for drafting discharge summaries. Over a period of nine weeks, 21 residents and 4 physician assistants generated a total of 379 summaries. The text produced by the LLM was incorporated into 58.5% of admissions, while identifiable content from the LLM appeared in 29.1% of the final discharge letters. Participants reported an 86.9% decrease in documentation time and a 60.9% reduction in administrative workload. Following the pilot, 91.3% expressed intent to continue using the system, highlighting the need for accurate measurement of documentation time for future assessments.
Key facts
- 379 discharge summaries generated
- 21 residents and 4 physician assistants participated
- 9-week pilot study
- 58.5% of admissions had LLM text copied
- 29.1% of final letters contained identifiable LLM content
- 86.9% reported reduced documentation time
- 60.9% reported reduced administrative workload
- 91.3% intent to use after pilot
Entities
Locations
- Netherlands