Background

Encounters prior to major surgery afford a unique opportunity for patient-provider communication, particularly regarding goals of care (GoC). Unfortunately, discussions involving surgeons, primary care providers, and anesthesiologists are often deferred or left incomplete due to providers’ discomfort and inexperience in navigating these sensitive conversations. Simulations have shown promise as solutions to provide in-creased exposure to trainees of these challenging discussions. In addition, large language models such as ChatGPT, have recently shown the ability to be used play the roles of characters in a variety of settings. The work aims to address this issue by using a language model (ChatGPT) to simulate GoC conversations, providing a training tool for general surgery and anesthesia resident physicians to improve their communication skills and enhance the alignment of surgical outcomes with patients’ values and preferences.

An Example - Mr Al-Farsi

Here, we show how a synthetic patient can interact with a student as a chatbot, or visualized with an avatar.

The synthetic patient can be modified to allow for trainees to interact with the patient at different stages of their illness, for example after recovery.

Developing additional patients is simple, allowing for a wide diversity in cohorts.