Beverly Slate, a high-fidelity mannequin, sits in a simulation lab hospital bed. In the room with her are four of Bryant University’s Physician Assistant students. One talks with Beverly at her bedside while the others, seated by the door, watch. Students ask Beverly how she’s feeling, discuss yesterday’s lab results, talk about the next steps, and ask if she has any questions. On the opposite side of the wall, looking through a one-way mirror, are Associate Director of Clinical Education Jodi Cusack, MHS, PA-C, and Stephanie Manchester, the voice actor pretending to be Beverly.
As students are learning, Beverly is not an easy-going patient. She’s confrontational, cuts off the PA students when they talk, and speaks mostly with sass. Despite this obstacle, students continue to make strides to ensure she feels understood and gets the help she needs.
Known as an Immersive Clinical Simulation (ICS), this four-day exercise centers on narrative medicine — an interdisciplinary practice that integrates storytelling in medical care by listening to and interpreting patients’ stories to understand an illness's emotional and psychological dimensions. Ordering consultations and handling the situation as if it were a real-world medical environment case, the students’ objectives within the ICS are to determine Beverly’s disease state and create a treatment plan.
“We try to dial up the realism as much as we can,” says Director of Medical Simulation and Assistant Clinical Professor Stephen Sherman, ’19 MSPAS, PA-C.
To effectively capture this authenticity, Sherman has been leveraging artificial intelligence.
A force multiplier in world-building
Developing a plausible medical simulation calls for world-building.
To set the scene, PA faculty members tell students they are at Bryant University Hospital, a massive academic medical center with access to a variety of specialists and testing. Sherman invents approximately a dozen characters for students to interact with and constructs pretend medical records and medical notes for their patient, which students access through their patient’s electronic medical record.
“When I started this exercise four years ago, I did it by hand. I wrote consult note after consult note thinking of my own medical background, but I'm not a specialist in everything, so I’d have to call people within specific areas to get their thoughts and verify information,” says Sherman, emphasizing that it could take six to eight hours to put together a single consult note; with ChatGPT, it takes approximately 30 minutes.
This year, Sherman used AI to assist with world-building and found it to be a force multiplier in the amount of content and type of content he’s been able to create. For instance, during the ICS, students needed to get professional advice from Savannah Thode, a fictional gastroenterologist. Having previously developed the gastroenterologist character, Sherman uploaded student-generated medical notes into ChatGPT and asked the following: “Write a consult note for this patient based on these student notes. The patient has ‘x’ disease and is presenting with ‘x’ lab values. Pretend like you're an irritable and very busy attending physician who thinks this is a baloney consult.” ChatGPT generated Dr. Thode's note — adding a bit of orneriness into the text.
After students discovered an abnormal finding on chest X-ray, Sherman prompted ChatGPT to write an additional note from the perspective of a junior attending on the pulmonology service who was anxious and wanted to be incredibly thorough. The result was a longer, rambling note. Sherman later asked ChatGPT to generate lab results and found that the bot generated not only realistic lab values but provided explanations and commentary to help him keep track of the lab trends internally.
“It feels like there is so much more personality in terms of different dynamics that students are going to deal with in the real world,” says Sherman, who verifies AI outputs with industry professionals.
Leveling it up
Following their conversation with Beverly, students head to an adjacent room to enter a new lab request into the simulation’s electronic medical record — another area where Sherman has integrated AI.
Wanting students to interact with the electronic medical record and technical support in a realistic way, Sherman used chatbot-generated code to create a simple messaging system, similar to that found in commercial medical record systems like Epic; this allows him to communicate with the students as the various characters in the simulation.
RELATED ARTICLE: From paratrooper to PA instructor, Matthew Lavoie is preparing others to save lives
“I have so much power, especially through coding tasks, to expand the realism than I ever had before this technology came along,” says Sherman, who’d never dabbled in coding prior to using it for this simulation. “You just tell the chatbot what you want as if you were explaining it to an expert who's then able to take your layman's ideas and translate them into a professional product.”
Sherman adds that he’s intentional on not giving students run-of-the-mill scenarios and, instead, is set on dialing up the challenge. To make the experience more realistic, he’s using AI to level up students’ emotional connection and investment in the patient and their care. Sherman adds that it takes a lot of social intelligence and ability to talk people through difficult situations and get them the help they need — especially in situations where there are interrelated medical and social issues that a patient may be willing, or unwilling, to deal with.
“I want them to go through an emotional experience. There are things you can't learn in books, and we want to increase their empathy, bedside rapport, and resilience in the face of a patient who might lash out at them,” Sherman says. “One of the important skills we can teach our students is sitting at the patient's bedside and being with them in their illness.”