Patient-Representing Population’s Perceptions of GPT-Generated Versus Standard Emergency Department Discharge Instructions: Randomized Blind Survey Assessment
ABSTRACT
Discharge instructions from the emergency department (ED) to home are crucial for patient communication but often get overlooked. Generative artificial intelligence and large language models (LLMs) offer a solution to create high-quality and personalized discharge instructions. This study aimed to assess patient perspectives of LLM-generated discharge instructions using ChatGPT in the ED.
METHODS
The study involved creating realistic ED encounters and generating LLM-generated discharge instructions using GPT-4 in Azure OpenAI Service. Standard discharge instructions were also created for comparison. These instructions were then presented to Amazon MTurk respondents representing patient populations for assessment.
RESULTS
The findings indicated that respondents favored GPT-generated return precautions over standard instructions, and overall rated GPT-generated instructions as more understandable and satisfactory. There were significant differences in favor of GPT-generated instructions in various subsections, particularly in interpretability of significance and satisfaction.
CONCLUSIONS
The study suggests that LLMs like ChatGPT could help reduce the documentation burden in the ED and improve communication with patients by providing tailored and more readable instructions.
PMID:39094112 | DOI:10.2196/60336
Clinical Trial Benefits and AI-Driven Solutions
Clinical trials are crucial for developing safe treatments, and our AI-driven platform, DocSym, consolidates standards, protocols, and research for clinicians. Streamlining operations is essential in healthcare, and our mobile apps support scheduling, monitoring treatments, and telemedicine to enhance patient care and expand services digitally. Using AI can improve workflows, patient outcomes, and reduce paper routines.
