Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AI Managed Emergency Documentation with a Pretrained Model

David Menzies, Sean Kirwan, Ahmad Albarqawi

TL;DR

Discharge letters from emergency departments are essential but frequently suffer from delays and inconsistencies. The authors fine-tuned the Davinci GPT-3 model (MedWrite) for medical writing and evaluated text and voice interfaces against manual typing with 19 emergency physicians, demonstrating notable time savings while preserving content quality. Key contributions include a data-and-feedback–driven fine-tuning pipeline, a privacy-preserving data processing flow, and evidence that AI-assisted generation can reduce clinician workload in ED documentation. The work suggests practical potential for AI-driven administrative tooling in EDs, contingent on real-world deployment and continuous model refinement to mitigate hallucinations and integration challenges.

Abstract

This study investigates the use of a large language model system to improve efficiency and quality in emergency department (ED) discharge letter writing. Time constraints and infrastructural deficits make compliance with current discharge letter targets difficult. We explored potential efficiencies from an artificial intelligence software in the generation of ED discharge letters and the attitudes of doctors toward this technology. The evaluated system leverages advanced techniques to fine-tune a model to generate discharge summaries from short-hand inputs, including voice, text, and electronic health record data. Nineteen physicians with emergency medicine experience evaluated the system text and voice-to-text interfaces against manual typing. The results showed significant time savings with MedWrite LLM interfaces compared to manual methods.

AI Managed Emergency Documentation with a Pretrained Model

TL;DR

Discharge letters from emergency departments are essential but frequently suffer from delays and inconsistencies. The authors fine-tuned the Davinci GPT-3 model (MedWrite) for medical writing and evaluated text and voice interfaces against manual typing with 19 emergency physicians, demonstrating notable time savings while preserving content quality. Key contributions include a data-and-feedback–driven fine-tuning pipeline, a privacy-preserving data processing flow, and evidence that AI-assisted generation can reduce clinician workload in ED documentation. The work suggests practical potential for AI-driven administrative tooling in EDs, contingent on real-world deployment and continuous model refinement to mitigate hallucinations and integration challenges.

Abstract

This study investigates the use of a large language model system to improve efficiency and quality in emergency department (ED) discharge letter writing. Time constraints and infrastructural deficits make compliance with current discharge letter targets difficult. We explored potential efficiencies from an artificial intelligence software in the generation of ED discharge letters and the attitudes of doctors toward this technology. The evaluated system leverages advanced techniques to fine-tune a model to generate discharge summaries from short-hand inputs, including voice, text, and electronic health record data. Nineteen physicians with emergency medicine experience evaluated the system text and voice-to-text interfaces against manual typing. The results showed significant time savings with MedWrite LLM interfaces compared to manual methods.
Paper Structure (13 sections, 1 equation, 2 figures, 2 tables)

This paper contains 13 sections, 1 equation, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. System Development
  3. System Architecture
  4. Data and Evaluation
  5. Methods
  6. Results
  7. Discussion
  8. Limitations
  9. Conclusion
  10. Declarations
  11. Data Availability
  12. Author Contributions
  13. Appendix

Figures (2)

  • Figure 1: High level system design of medical discharge letter generation system highlighting the data flow.
  • Figure 2: Compare the average completion times for discharge letters using manual typing and AI-assisted generation. The AI-assisted generation uses simulated EHR input using text-based interface or voice dictation.