Health professions education aims to provide students with practical clinical experience and skills in person-centred care before graduating, and simulated patient interactions are a potential teaching method. However, these interactions can be resource-intensive and expensive, especially if actors are included. AI assistants like Claude provide a scalable alternative for role-playing dialogue. This post describes the process of using Claude to take on various patient personas for health professions education.
How it works
Claude is an AI system created by Anthropic to have natural conversations where you can use detailed prompts to establish a scenario, character, background, and conversation goals.
With the right prompt, Claude can realistically take on the persona of patients with diverse attributes and medical needs. In this context, the lecturer (or team) writes the initial prompt and then gives it to students, which they copy and paste into the chatbot. Then they interact with the AI / patient persona through their own dialogue.
Crafting the prompt
Effective prompts should clearly define the persona and expected interaction. In this context, you’re not only providing information in the form of a vignette; you’re establishing the worldview of the model by creating all the context in which this interaction takes place.
Here are some tips for prompt writing:
- Establish core patient attributes like gender, age, background. Give the persona a name.
- Summarise the relevant medical history, conditions, and needs of the persona.
- Set the scene. For example, describe where the interaction takes place. Give the persona a backstory by including family, social, and medical history.
- Define the student’s role. Are they a new nurse meeting a patient? A physiotherapy student?
- Clarify the conversation goals. Do you want the student to educate the patient? Conduct an assessment? Deliver difficult news?
- Use natural language. Avoid stiff computer-generated text.
- Provide 3-5 sample dialogue exchanges to demonstrate the desired personality. This helps the model understand what this dialogue ‘should’ look like.
- Encourage open-ended responses from students. They should also ask follow-up questions.
With clear prompting, Claude can portray patients of varying ages, cultures, attitudes, health statuses, communication abilities, and emotional states. This allows for extremely diverse scenario development.
The student experience
The lecturer provides the base prompt defining the persona and situation. Students simply paste this into Claude and then begin adding their own dialogue. Because of the nature of language models every student will have a different interaction, even when starting from the same prompt.
After the virtual interaction, students write a reflection, analysing how they built rapport with the persona, gathered information, delivered patient education, and achieved the care objectives. This reflection is then submitted to instructors for feedback, or used to generate conversation in a peer tutorial.
By exposing students to realistic patient encounters through Claude, they gain valuable communication, empathy, and clinical thinking skills before entering real-world practice.
Example
See below for an example of a very simple interaction.
Conclusion
AI-powered virtual patients like Claude are a promising educational technology. With thoughtful prompt engineering, health professions students can engage with conversational agents that simulate diverse personas, needs, and care scenarios. This provides efficient yet realistic practice interactions and reflection opportunities compared to resource-heavy in-person simulations. While not a complete substitute, AI assistance is a supplemental way to help prepare future nurses with the complex communication and critical thinking skills necessary to provide compassionate, effective patient care.