At a time when artificial intelligence tools like ChatGPT can answer increasingly complex questions, the question arises: How long will it be before AI can provide trustworthy medical advice?

Probably not as long as you think, according to Epic Systems executives who spoke Tuesday at the healthcare records giant’s annual user group meeting at the company’s Verona campus.

The event attracts tens of thousands of the company’s customers each year who want to learn about the company’s latest developments and learn from colleagues in the information technology and healthcare industries. This year, more than 44,000 people from all 50 states and another 16 countries attended in person or online, the company said.

In typical Epic style, the meeting always has a fantastical theme. This year it was Storytime, and CEO and founder Judy Faulkner took the stage as “Lady Swan,” her version of Mother Goose.

For many, the content of the morning speech was was perhaps almost as unbelievable as those children’s stories. The speakers described tools already available and in the works that would fit perfectly into the science fiction novels that inspired the massive Deep Space Auditorium in which they spoke.

Some of them are already known. AI helps nurses write messages to patients And allows doctors to summarize patient visits without making a single note.

There is also the huge Cosmos database, which now contains the anonymized data of 270 million patients. This allows researchers to conduct studies in record time and doctors to help their patients in new ways.

Through the The Look-Alikes feature allows providers who care for patients with rare or mysterious diseases to search for similar patients elsewhere and connect with the physicians who have cared for those patients to learn from their experiences.

Another tool currently being tested would allow doctors to evaluate treatment options based on the outcomes of other patients in the database. Best treatment options for my patient, It is designed to give providers “another tool in their toolbox,” says Rahul Shah, the project’s lead developer.

“Best Care Choices looks at your patient and finds hundreds or thousands that are very similar,” Faulkner said. “It shows you the different treatments those patients have received and how effective each one was. Then you and the patient can discuss the results and choose the best one.”

The AI ​​frontier

Seth Howard, Epic’s vice president of research and development, told the crowd that Epic is in the midst of an “explosion of development” as past efforts “break out and serve as the foundation for the next wave of innovation.” The morning presentations, he said, would highlight only about 5% of the company’s current projects, which include around 100 different applications for generative AI.

Currently in the works are features that use AI to shorten or simplify patient records, automatically provide information about a patient’s medical history or test results when a provider composes a message to the patient, answer patient questions with chatbots, or summarize a previous shift for the next caregiver to take charge.

Perhaps what attracted the most attention, however, was a futuristic feature that Epic said it revealed much earlier than usual. With a cellphone in hand, Seth Hain, senior vice president of research and development, demonstrated a possible future feature of MyChart, the company’s ubiquitous patient portal.

In the demonstration, an AI voice contacts Hain to see how he’s doing after recent wrist surgery. The voice asks him to rate his pain, then tells him to hold his hand up to the camera and bend it back as far as possible. It calculates his current range of motion and creates a graph that compares Hain to other patients.

“According to Cosmos data, your recovery is faster than patients like you. Keep up the good work,” the voice says. When Hain asks if he can play pickleball again, however, it advises him to wait.

At a meeting with reporters after the presentation, Hain was asked how long it might take for such a tool to become a reality.

“It’s still very, very, very early to say how and where the broader medical community will adopt these kinds of things, but it’s doable,” Hain said, adding that the demonstration was intended to show “where we think our products can go … and to stimulate discussion about it.”

One reporter went further and asked whether this type of technology would be available in, say, three or ten years.

“Three years seems like a long time right now. It’s moving really, really fast,” said Hain, explaining that a number of factors are accelerating these changes. Not only is technology advancing rapidly, but many tools are also becoming cheaper and more accurate. This is encouraging more and more vendors to adopt these tools.

Additionally, patients are already asking questions about their health outcomes to general AI tools like ChatGPT and “regularly getting meaningful answers,” Hain said. “But there is a real risk in doing so. To be clear: I advise against it.”

But as medical professionals retire and health systems struggle to hire and retain enough staff, Hain sees an opportunity for Epic to answer these kinds of questions while “establishing a set of guardrails and managing expectations.”

“There’s still a long way to go to help people do these things successfully, (but) the technology is evolving rapidly.”