We all know ChatGPT and are intrigued by its potential, but using it to its fullest is not always so straightforward. Do you know how to write an effective prompt? Do you know about GPT models and their role in designing clinical scenarios? In this article you will discover how to leverage ChatGPT to create realistic and personalized scenarios, addressing doubts and mistrusts. I will guide you step by step, showing how this technology can simplify the creation of complex scenarios through a customized version of the original GPT model. And I will present a concrete example that you can use right away. ChatGPT can be our ally in clinical simulation.
Imagine you need to create a detailed, customized clinical scenario for your next course, but time is scarce. You find yourself structuring learning objectives and organizing every operational detail. What if there was a tool that could help you do all this in a matter of minutes? ChatGPT is here to turn this stressful situation into a quick, intuitive, and even creative process. Find out how in just a few simple steps.
What is ChatGPT in a nutshell
The introduction of Artificial Intelligence (AI) into the field of healthcare education, particularly in medical simulation, presents a complex challenge but one full of opportunities. AI technologies promise to revolutionize traditional teaching methods by offering innovative tools for creating realistic, accessible, and personalized training scenarios. (Hamilton, 2024)
ChatGPT, developed by OpenAI, has emerged as one of the most versatile and promising AI models in this context. Its ability to understand and generate natural language in a fluid and contextualized manner makes it ideal for simulating complex interactions, such as those between doctor and patient or between different professionals in healthcare settings. (Barra, 2024)
The widespread deployment of ChatGPT, aided by a user-friendly interface and a robust support ecosystem, has contributed to its rapid adoption in various clinical settings. Unlike other AI models, such as Gemini (Google), Claude (Anthropic) or Grok (xAI), ChatGPT is easily integrated into existing platforms and systems, making it an accessible tool even for healthcare professionals with basic technology skills.
The GPT-4o version introduces advanced features such as the creation of customized GPTs: models trained on specific datasets to meet special needs and advanced voice interaction modes (Advanced Voice Mode or AVM).
ChatGPT for Writing Simulation Scenarios
Creating realistic and appropriate scenarios for medical simulation is a complex and challenging task, even for the most experienced trainers. It requires a deep knowledge of instructional design, a solid understanding of clinical simulation dynamics, and the ability to adapt scenarios to specific learning needs. The proper use of SimZones and adherence to the best practice standards of international scientific societies are crucial elements to consider at this stage. (Rodgers et al., 2023)
Here are some key advantages of using ChatGPT in scenario writing:
- Organized structure: ChatGPT can create at user prompts structured textual output according to validated or customized templates, which can later be revised and refined by trainers, allowing them to focus on adapting and improving specific content, rather than creating it from scratch.
- Customization: The ability to insert precise instructions via effective prompts allows ChatGPT-generated scenarios to be adapted to the specific needs of learners, the desired level of complexity, and the training context. For example, the type of learner (medical students, nurses, etc.), operational context (SimZone, hospital ward, etc.) and learning objectives can be specified.
- Efficiency: ChatGPT can significantly reduce the time required for scenario creation, allowing trainers to spend more time planning simulation activities, debriefing and other crucial aspects of the training process.
How to write an effective prompt
To take full advantage of ChatGPT’s potential, mastering the art of prompting is essential. An effective prompt is clear, concise, and rich in detail, guiding the model toward generating relevant, high-quality content.
Try, for example, writing “Create a respiratory failure scenario” and assess how generic and insufficient the AI output can be.
Here are some suggestions for creating effective prompts:
- Clarity in prompts: define precisely what you want to achieve from ChatGPT, specifying the type of scenario, learning objectives, target audience, and desired level of detail.
- Use of open questions: encourage ChatGPT to provide elaborate and creative answers, avoiding closed questions that limit its ability to generate original content.
- Explicit indication of context: provide ChatGPT with all the information it needs to contextualize the scenario, such as the clinical setting, participants’ roles, and any resources available.
It is critical to remember that ChatGPT is a tool, not a substitute for the facilitator. Careful and critical review of both the prompt and the output generated by ChatGPT is essential to ensure the accuracy, completeness, and appropriateness of the content. Human verification is essential to identify and correct any errors or “hallucinations” generated by the model, especially regarding medical information and drug dosages. (Violato et al., 2023)
What are GPT models and what are they used for?
In addition to classic interaction in conversational form, ChatGPT also offers the ability to create customized models, called GPTs, that allow specific information to be integrated and generate contextualized responses based on a predefined body of knowledge.
To maximize the effectiveness of custom GPTs, it is critical to follow certain guidelines in data preparation and model design, ensuring content accuracy and relevance.
For example, Masters et al. (2024) suggest to:
- Clearly define the purpose and context of use of the customized GPT.
- Provide detailed instructions to the model through practical examples of questions and situations.
- Use specific files or documents as references to increase the accuracy of answers.
- Integrate external APIs to extend the functionality of the model, if necessary.
To create a custom GPT in ChatGPT, follow these steps:
1. Login and Start Customization:
- Once logged into ChatGPT, in the left sidebar, click on “Explore GPT” to access the library of available templates.
- Within this section, select “Create a GPT” to start the customization process.
2. GPT definition:
- Assign a descriptive title to your GPT that reflects its function (e.g., “Scenario Creation Assistant” or “Chest Pain Patient Simulator”).
- Choose a representative image for your GPT. You can use the built-in tool to generate one automatically or upload a custom one.
3. Instructions and Context:
- Provide detailed instructions on the purpose of the GPT and the contexts in which it should be used.
- Include practical examples of questions or situations to guide GPT behavior. For a patient simulator, provide examples of questions students might ask to gather information about the patient’s medical history and symptoms.
4. Advanced Settings:
- Uploading Files: Upload files or documents that GPT will use for reference to provide more accurate answers.
- External API Integration: If necessary, integrate external APIs to extend the functionality of the GPT. For example, to connect the GPT to other services, such as Google Calendar or Slack, to send reminders or notifications.
5. Visibility and Storage:
- Decide the visibility of your GPT:
- Only you: accessible only within your ChatGPT account.
- Only people with a link: shareable via a unique URL.
- Public: visible to all ChatGPT users.
- Finally click “Save” to complete the GPT creation process.
ChatGPT’s flexibility in creating custom GPTs allows templates to be tailored to the specific needs of different users. However, it is critical to pay attention to the privacy of the information used to train and configure custom GPTs.
Scenario Designer: a GPT for simulationists
A concrete example of a customized GPT is “Scenario Designer,” developed by Dr. Barra for the SIMNOVA Simulation Center to simplify the creation of educational scenarios. This model, introduced in early 2024, was initially tested in Peer Education courses for Trainers and later presented to the simulation community during the SESAM 2024 congress in Prague.
“Scenario Designer” guides the user through a series of prompts aimed at defining scenario parameters:
- Output language: the user can specify the language in which he/she wishes to obtain the scenario.
- Learning objectives : the training objectives that the scenario aims to achieve are defined.
- Target audience of learners: the target audience to which the scenario is addressed (e.g., physicians in training, nurses, etc.) is identified.
- Operational context: the environment in which the simulation and SimZone will take place is specified.
The template, based on a predefined or user-customizable template, generates output structured in several sections, each designed to facilitate the planning and organization of a complete simulation.
Benefits of the GPT Scenario Designer
- Detailed and sequential content review: Structuring the output facilitates scenario review and modification.
- Reduction of trainers’ workload: automatic generation of the base scenario frees trainers from repetitive tasks.
- Standardization of scenarios : the use of a template ensures consistency and uniformity in scenario structure.
- Improved efficiency: splitting into focused sections reduces the amount of information the template has to process for each request.
Access “Scenario Designer” in 5 steps
1 Access the GPT Store: available from the ChatGPT sidebar
2 Search for the model: you can search by name (“Scenario Designer”) or by author (Federico Lorenzo Barra)
3 Select the model : to view its functionality
4 Add to the collection: the GPT will appear in the left sidebar of the interface, identified by an icon and a title
5 Activate the template: selecting it will take you to a dedicated chat that uses the specialized template for scenario creation
Conclusions
The integration of ChatGPT into medical simulation represents a promising frontier for healthcare training. With the right support and rigorous supervision, these technologies can help make medical education more effective, accessible and inclusive by providing students with realistic and well-structured learning opportunities. However, it is critical to continue to explore the limitations of these tools and conduct further research to validate their effectiveness. It will also be crucial to address challenges related to ethics and accuracy to ensure that these technologies are used in a safe and beneficial way for all participants in the educational process.
Bibliography
GPT-4 Technical Report. Retrieved from https://arxiv.org/abs/2303.08774
Hamilton A. Artificial Intelligence and Healthcare Simulation: The Shifting Landscape of Medical Education. Cureus. 2024;16(5):e59747. Published 2024 May 6. doi:10.7759/cureus.59747
F.L. Barra, A. Costa, G. Rodella, F. Semeraro, L. Carenzo, Shaping the future of simulator interactions: The role of ChatGPT’s Advanced Voice Mode, Resuscitation (2024), doi: https://doi.org/10.1016/j.resuscitation.2024.110452
Rodgers DL, Needler M, Robinson A, et al. Artificial Intelligence and the Simulationists. Simul Healthc. 2023;18(6):395-399. doi:10.1097/SIH.0000000000000747
Violato E, Corbett C, Rose B, Rauschning B, Witschen B. The effectiveness and efficiency of using ChatGPT for writing health care simulations. International Journal of Healthcare Simulation. 2023. doi:10.54531/wjgb5594.
Masters K, Benjamin J, Agrawal A, MacNeill H, Pillow MT, Mehta N. Twelve tips on creating and using custom GPTs to enhance health professions education. Med Teach. 2024;46(6):752-756. doi:10.1080/0142159X.2024.2305365OpenAI. (2023).
READ ALSO
