Related Experiment Video
Updated: Jun 10, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
Assessment of Artificial Intelligence-based Translation Tools for Emergency Department Discharge Instructions
Estee Wu1, Cassandra Mackey2, Simi Jandu2
1Upstate Medical University Norton College of Medicine, Syracuse, New York.
Artificial intelligence (AI) translation tools like ChatGPT-4 show promise for improving emergency department (ED) discharge instructions for patients with limited English proficiency (LEP). However, human oversight is still essential for accuracy and patient safety.
Area of Science:
- Medical informatics
- Natural Language Processing
- Health Communication
Background:
- Limited English proficiency (LEP) affects 8% of US residents, increasing risks of poor care quality and safety due to language barriers.
- Miscommunication during emergency department (ED) discharge can lead to adverse health outcomes, including medication errors and readmissions.
Purpose of the Study:
- To evaluate the quality and safety of AI-generated translations of physician-written ED discharge instructions.
- To assess AI translation performance across varying levels of linguistic complexity.
Main Methods:
- Physician-written ED discharge instructions for four conditions were translated into Albanian, Brazilian Portuguese, and Vietnamese using ChatGPT-4, Microsoft Copilot, and Google Translate.
- Professional medical translators assessed translations for semantic accuracy, fluency, meaning, severity, preference, and formality on 5-point and 3-point scales.
Main Results:
- Overall translation quality scores were similar across AI models, with no statistically significant differences.
- ChatGPT-4 demonstrated superior performance in translating instructions into Albanian and was rated highest for Albanian and Brazilian Portuguese.
- Microsoft Copilot and Google Translate each produced five potentially harmful translation errors, while ChatGPT-4 produced none.
Conclusions:
- AI translation tools offer potential for improving ED discharge instructions for LEP patients.
- ChatGPT-4 exhibited the best overall performance and safety profile among the evaluated AI models.
- Despite AI advancements, human oversight remains critical to ensure accuracy and mitigate risks in medical translation.
Related Concept Videos
Improving Translational Accuracy
Non-equilibrium in the Cell
Issues And Trends In Healthcare Delivery System
Cost Containment
Payment for healthcare services has historically promoted adoption of costly and often unnecessary or inefficient...
Guidelines for Nursing Documentation I
Factual:
The following points emphasize the significance of upholding accurate and unbiased documentation in healthcare.
Discharge Summary Forms
Here's a detailed look at the key components and guidelines for preparing a discharge summary:
Translation
Translation Produces the Building Blocks of Life
Proteins are called the...