ChatGPT: a new milestone in medical AI, driving the next evolution of healthcare

Smart healthcare and AI have become major trends in medicine, and ChatGPT is a powerful tool for applying AI to improve patient education. Applied to smart healthcare, ChatGPT brings more convenient and more efficient approaches to medical education and clinical scenario training.

From morning news to business weeklies to social media, ChatGPT is everywhere—and one researcher has been quietly using it to improve healthcare.

Improving patient education with ChatGPT

James Benoit, a postdoctoral researcher at the University of Alberta, is passionate about using AI to improve health education and information delivery in the healthcare industry. He believes the ChatGPT chatbot can make patient education more personalized and better suited to patients' needs. As he puts it, "Tools like ChatGPT can improve health education and training, and in turn support more personalized medicine and patient-centered care."

Using ChatGPT to find cheaper, more efficient training methods

Clinical vignettes are simulated clinical scenarios that play a key role in medical education and research. For example, a vignette might describe a couple discussing their child's allergic reaction: "My daughter's skin is itchy and covered in bumps. After taking the medicine she broke out in a rash all over, and she's also having trouble breathing and her voice has gone hoarse."

Useful as they are, clinical vignettes are not easy to produce. Most are painstakingly crafted by hand, and medical training is rarely cheap—similar questions and multiple-choice items also have to be designed, putting the cost of each vignette at roughly US$1,500–2,500. Attempts have been made to generate vignettes automatically, but they tend to offer less flexibility across different medical scenarios.

ChatGPT generates clinical vignettes effectively

In his recent research, James investigated whether ChatGPT could effectively generate clinical vignettes—rewriting them from different medical perspectives through the chat interface, and even producing diagnoses from different sets of vignettes.

With ChatGPT and simple prompts, James easily generated ten clinical vignettes for common pediatric conditions—of remarkably high quality, and without spending a cent.

Can ChatGPT produce more accurate diagnoses than healthcare professionals?

One part of James's experiment tested whether the ChatGPT chatbot could generate diagnoses from existing clinical vignettes.

Across 45 clinical vignettes, ChatGPT identified the correct condition with 75.6% accuracy—comparable to physicians' 72.1%.

What's next for James?

James next wants to use ChatGPT-generated clinical vignettes and Oracle Cloud services to advance personalized medicine. He hopes to develop machine learning algorithms that could produce an accurate diagnosis from a child's own description of their symptoms as relayed by parents.

ChatGPT produces clinical vignettes quickly and cheaply, generating vast amounts of vignette data for testing and training these new machine learning algorithms. Oracle Cloud can provide the compute engines to power these algorithms at scale.

What can ChatGPT bring to healthcare?

We don't yet know how much revolutionary innovation ChatGPT will bring, and news of its negative applications often dominates the headlines. What matters is that researchers like James are harnessing the power of ChatGPT to personalize medicine and improve how it is taught and delivered.

The content of this article has been licensed by Pharma Statistical Insights to PatientForce™ (Media-WIND Communications Co., Ltd. / Unified Business No.: 27903403). Its intellectual property rights are jointly owned by PatientForce™ and Pharma Statistical Insights in accordance with the law. If you believe this article may infringe any rights, please contact the relevant department of our company by email and we will respond as soon as possible.