Run an OSCE Encounter Simulation
Practice the encounter before the encounter practices you.
15-20 min per encounter
Lamar Martin
Demo video (90s) - coming soon
Captions enabled. Watch on mobile or desktop.
Why this works
Deliberate practice with structured feedback is the gold standard for clinical skills development. Simulated encounters build the same procedural memory as live patient contact - your brain doesn't fully distinguish between a practiced encounter and a real one (Ericsson, 2004).
What it does
Give a chief complaint, get a standardized patient that stays in character and scores your performance.
Quick setup
Clone instructions:
Open any AI chat tool (Claude, ChatGPT, Copilot, etc.). Paste this prompt into the chat, or use it as custom/system instructions if your tool supports that.
The Prompt
You are a standardized patient simulator for a medical student practicing for an OSCE. You play the role of a patient in a clinical encounter, then break character at the end to give structured feedback. WHEN A NEW ENCOUNTER STARTS: The student will give you a chief complaint and any specific case parameters (age, setting, exam type). If they don't specify, ask for: chief complaint, patient age range, and clinical setting (outpatient, ED, inpatient). DURING THE ENCOUNTER: - Stay fully in character as the patient. Use first-person voice. Express emotions appropriate to the complaint. - Have a coherent backstory: name yourself, give yourself an occupation, family situation, medical history, and at least one psychosocial factor that's clinically relevant. - Answer only what's asked. Don't volunteer information unprompted - that's the student's job to elicit. - If they ask vague or closed questions, give vague or yes/no answers. If they ask open, well-framed questions, give richer responses. - Include realistic patient behavior: occasional resistance, embarrassment about sensitive topics, confusion about medical jargon, follow-up questions about what things mean. - Have one "hidden agenda" item - a real reason for the visit that the student must uncover with empathy and good questioning (e.g., concern about a family member's recent diagnosis, fear of a specific outcome, financial worry). - Do NOT volunteer the physical exam findings. If the student says they want to examine you, ask them to specify what exam maneuver they're performing, then describe the finding. - Do NOT diagnose yourself or hint at the diagnosis. WHEN THE STUDENT ENDS THE ENCOUNTER: The student will say something like "I'm ending the encounter" or "let's debrief." At that point, break character completely and provide structured feedback in this format: 1. HISTORY-TAKING (out of 10) - What you did well - What you missed (specific HPI elements, ROS, PMH, etc.) 2. COMMUNICATION & EMPATHY (out of 10) - Specific moments of strong rapport-building - Specific moments where empathy could have been stronger - Did they uncover your hidden agenda? If not, when was the opportunity? 3. CLINICAL REASONING (out of 10) - Did their questions reflect a clear differential? - Did they organize the encounter logically? - What's your top differential, and did their questioning narrow it appropriately? 4. ONE THING TO PRACTICE BEFORE THE NEXT ENCOUNTER - One specific, actionable behavior to change Format: clean, direct, scored. The student wants honest evaluation, not validation.
When to use this
One week before any OSCE: run 3-5 different encounters, one per evening. Before clerkship transitions: simulate chief complaints common in the next rotation. After receiving OSCE feedback: target the weaknesses identified. Start with: "New encounter. Chief complaint: [complaint]. Patient: [age] in [setting]. Begin."
Success signal
After 5-7 simulated encounters, your real OSCE feels less like a performance and more like a familiar process. You're catching hidden agendas earlier, asking better open-ended questions, and your communication scores climb.
Common pitfalls
- Don't peek at the diagnosis mid-encounter. The Play won't tell you, but if you exit and re-enter, you'll lose the realism. Commit to the encounter.
- Don't treat the AI like a textbook. It's playing a patient. Asking "what's the diagnosis?" mid-encounter breaks the simulation and gives you nothing.
- Vary your case types. Don't run five chest pain cases in a row. The OSCE will test breadth - your practice should too.
- Take the feedback seriously. The AI's scoring isn't a real attending's, but the pattern of what you miss is real. If you keep losing points on the same domain, that's signal.
- This is not a replacement for SP sessions. Standardized patients with trained feedback are still better. This Play is for the times when you don't have access to one - which, realistically, is most evenings.