I recently heard the term “synthetic patient”—used to describe a ChatGPT-powered simulated patient designed for clinical communication training and OSCE practice—and immediately thought, “Ooo, I’ve never heard that term before! Is this a new, widely used label? Is this the 2025 remix of the old-school ‘virtual patient’? The hot, new glow-up of 'digital patients'?"
Naturally, curiosity got the best of me. One Google search turned into many… and yes, I even asked the expert itself: "Hey ChatGPT, what is a synthetic patient?”
Here’s what I learned:
What Is a Synthetic Patient?
A synthetic patient is a digitally or physically created representation of a patient used for training, research, or testing in healthcare education. These are not real people, but are designed to simulate the conditions, behaviors, and responses of real patients.
It then broke it down into three primary types:
- Digital Synthetic Patients
- These are used mostly in virtual environments and often powered by AI:
- Think chatbots or avatars that simulate interviews, emotional responses, or medical histories.
- Often embedded into simulation platforms or telehealth training tools.
- Flexible in age, symptoms, and personality.
- Example: An AI-driven virtual patient that speaks and reacts during a simulated telehealth visit. This is definately where Spark fits in.
-
Physical Synthetic Patients
- These are the manikins and task trainers we see in sim labs:
- Used for hands-on training in procedures like CPR, catheter insertion, or trauma scenarios.
- Some have sensors, vitals, and even voices (especially when paired with AI tools like SimVox).
- Example: A full-body manikin that breathes, bleeds, blinks, and talks during an emergency simulation. This could be where ALEX makes the cut.
-
Data-Based Synthetic Patients
- These live in spreadsheets, not sim labs:
- Synthetic datasets mimic real patient records, used for research and AI training.
- Vital for testing diagnostic algorithms or EHR platforms—without real patient data.
- Example: A synthetic dataset used to train a model to detect early signs of heart disease.
So… Is “Synthetic Patient” the New “Virtual Patient”?
Maybe. The rise of generative AI might also be driving the need for sharper terminology.
We may be entering a phase where the language we've relied on—like “virtual patient”—no longer feels specific enough. These new AI-driven tools can listen, speak, respond, and even adapt on the fly. That’s a big leap from the static, pre-scripted scenarios that originally defined virtual patient simulations.
This question brought to mind a 2015 article that still feels surprisingly relevant: “Virtual Patients – What Are We Talking About?” by Kononowicz et al. In it, the authors unpack a longstanding challenge in healthcare simulation: the term “virtual patient” is incredibly broad—and often confusing. It’s been used to describe everything from simple multiple-choice case studies to high-fidelity animated avatars. As a result, the same term could refer to very different tools, technologies, and learning experiences.
Maybe it's not enough any more to say virtual patient—
New Terms for a New Era in Simulation
As we move from scripted simulations to generative patients that listen, speak, and adapt—our language might have to catch up.
“Synthetic Patient” might carve out a niche meaning and emerge as the go-to term for:
- AI-driven or AI-enabled
- Dynamically generated
- Non-human by design
I suppose time will tell! Maybe we’ll see “synthetic patient” make its official debut in Version 4 of the SSIH dictionary, or maybe it ends up just referring to spreadsheet-based datasets.
But regardless of what we call them—virtual, synthetic, or even intelligent patients—the direction is clear: We’re entering a new era of healthcare education, where simulations are no longer static tools, but smart, adaptive partners in learning.
