Abstract:

Study Objectives:
The Emergency Department (ED) is a setting with a particularly high demand for efficient and clear patient education at discharge given potential reappearance or sequelae of symptoms requiring emergent evaluation. However, the ED also often imposes a high degree of constraint on patient-provider interaction time, and providing appropriate patient education during or at the conclusion remains an important area for improvement. Personalized resources may offer an impactful way to improve patient understanding and satisfaction during or after a healthcare visit. ChatGPT is a novel chatbot developed by OpenAI, which has many potential applications for improving patient education. Here, we aim to evaluate the potential role that ChatGPT could have as a personalized educational resource for patients seeking emergency medical treatment.

Methods:
To study how ChatGPT can be used by patients, we searched through the literature to understand and contextualize the current data on this program’s capabilities and clinical applications. We inputted questions from a range of healthcare topics and concerns relevant to a patient seeking emergency care, and documented ChatGPT’s answers.

Results:
As a text-based interface, ChatGPT allows users to interact with the system using natural written language. This new modality can quickly give personalized responses to queries on a wide range of healthcare topics, clarify discharge instructions, help triage medical problem urgency, and potentially improve patient-provider communication. The information provided describes an initial depth of response appropriate for a lay-person’s introduction to the medical topic. In contrast to a general internet search, ChatGPT summarizes the pertinent information in a conversational format that is fast and easily accessible to the user. It remembers previous conversations, which allows for specific clarification and follow-up for each response. ChatGPT is also free-to-use, available in multiple languages, and is more empathetic than previous chatbots. However, like any new technology, ChatGPT has limitations, including technical issues, a risk for misinformation, a lack of input standardization, and concerns about data privacy.

Conclusion:
This paper is the first to describe the use of ChatGPT as an educational resource for patients seeking emergency medical treatment. Although limitations exist, this free-to-use software offers compelling benefits for patient education. Physicians must understand these benefits and limitations to best guide patients, help guide the development of new AI technologies, and conduct further research with the goal of improving patient care.

Keywords:

Technology, education, medicine, large language models, ChatGPT, patient education.