Teaching residents to think like detectives
When Dr. Emily Thomas needed a postpartum hemorrhage case for her family medicine residents, she had to work around one gap: InvolveXR doesn’t yet have a way to visually show active bleeding. It’s on the roadmap. In the meantime, Thomas turned that into an opportunity. She built the entire diagnosis into the conversation, relying on the platform’s AI to carry the clinical picture through dialogue rather than a visual cue, and the case turned out stronger for it.
Thomas teaches at the West Virginia School of Osteopathic Medicine. She’s an internist by training, not an obstetrician, so she brought in a family medicine colleague to check the clinical details. What she designed was hers alone: a scenario where nothing visual tells the resident what’s wrong. The patient has to.
The case opens with a 33-year-old who just delivered a large baby after a long, complicated labor. She’s awake and talking. She mentions she’s been feeling weak and dizzy every time she stands, and that she’s noticed blood in the toilet, though she’s not sure if that’s normal. Her vitals tell a different story: low blood pressure, a racing heart.
Here’s what that encounter sounds like.
Nothing about the encounter announces the diagnosis. Residents have to ask the right questions, notice that what the patient is saying doesn’t match her numbers, and build the diagnosis themselves. Thomas calls it teaching residents to be good detectives.
The case runs on three states. Recognize the hemorrhage, start fluids and the right medication, and the patient recovers. Miss it or move too slowly, and she decompensates: more tachycardic, unresponsive, blood pressure dropping further. For findings the platform can’t yet show physically, like a uterine exam, Thomas built in guidance her operators can relay on request, so the clinical picture stays whole no matter what a resident reaches for.
She keeps the AI tightly scripted on the facts that matter. Running several simulation rooms at once means one inconsistent detail can derail a debrief, so the hard clinical facts stay locked while the conversation itself stays natural and responsive.
That design choice paid off. The debrief wasn’t about whether residents caught the hemorrhage. It was about how they got there, working through transfusion thresholds, uterotonic medications, and placental exams, because those questions came out of what they noticed, not what the simulation handed them.