HIMSS23: Epic analytics leader details efforts to bring GPT-4 into healthcare and growing work in genomics, life sciences

CHICAGO—Health IT giant Epic made a big splash Monday at the HIMSS 2023 conference when it announced plans to integrate generative AI into its electronic health record software.

Epic is expanding its collaboration with Microsoft with a focus on combining the Azure OpenAI Service with Epic’s EHR software. The aim is to increase the productivity of physicians as well as back-office professionals.

Initially, Epic is focused on using OpenAI's GPT-4 AI language model to help draft message responses from healthcare workers to patients and for use in analyzing medical records while looking for trends.

"Those are just two of many use cases that we're looking at. There's a lot of general applicability to this," Phil Lindemann, Epic's vice president of data and analytics, said during an interview on the HIMSS23 exhibit hall floor. "The area where we focused first was, 'How can we make physicians more efficient?'"

During the COVID-19 pandemic, as the use of virtual care boomed, doctors also saw a spike in digital messages from patients ranging from requests for medication refills to more complicated medical questions, Lindemann noted.

"Physicians and their teams have to be able to process those messages, and a lot of them are routine but some of them are more complex and physicians have to consume that. We are using generative AI within EHR 'in-basket' messaging to create draft responses," Lindemann said.

The AI-enabled solution will help triage the severity of patient messages as well, he noted.

UC San Diego Health and UW Health in Madison, Wisconsin, are among the first organizations to begin using the solution, and Stanford Health Care is expected to add the functionality soon.

The second use case for OpenAI's GPT-4 brings natural language queries and interactive data analysis to SlicerDicer, Epic's self-service reporting tool. That functionality enables clinical leaders to "explore data in a conversational and intuitive way" to find trends.

"Physicians might want to ask questions about the efficacy of a drug or they want to understand readmission rates. It's a way they can do on-demand reporting without having to call up IT. With GPT, we can talk like a normal conversation and it knows how to translate it into tech speak," he said. "Those are the two use cases that we see as really valuable to make a physician's day more efficient by helping them with that high volume of patient messages and then the ability to analyze large sets of data."

Some researchers are concerned about how generative AI and GPT-4 will impact diagnosis and clinical decision-making. So far, most healthcare companies are focused on administrative functions and medical documentation.

Lindemann said Epic's work with generative AI is guided by its principles of responsible AI. "One aspect of that methodology is 'human in the loop.' As we roll this out, every message that's going to be generated, a human is going to review it. When we're looking through a query, where it's translating the language, someone is evaluating it against the actual code. So there has to be this transparency early on or someone is reviewing every use case, which is slow, but it's the right thing to do to be responsible with AI," he said.

As large language models are used in healthcare, it's important to have guardrails and build trust and transparency.

"We have to make sure that whenever there is an opportunity for transparency, we show that to the clinician, to the billing office, even to the scheduler. And there need to be metrics to measure success and to say AI got the answer right X percent of the time. We don't just say let it loose and let it figure it out. There has to be constant monitoring," Lindemann noted.

He cautions that as healthcare organizations begin to pilot and test generative AI for different use cases, clinical leaders need to be diligent about how medical data are being used.

"It's up to you to manage your patients' privacy and make sure your data doesn't go into someone else's hands where it's just being sold off. That's an important thing as we're in this gold rush of generative AI is to understand how the data will be used and how it's kept safe," he said.

He noted that generative AI is at "peak hype" at the current moment.

"I do think the way that generative AI is going to change how we all are working and interacting is substantial. When this dies down, we are going to see some enhancements that historically would have taken 10 years and now they're going to happen in a year. That's what I'm excited about. This is a significant change," he said.

Epic is historically known for its work in electronic medical records software, but the company also has been expanding its efforts in genomics and life sciences.

Last fall, the company rolled out a genomic test ordering tool within its EHR with New Orleans-based Ochsner Health as the first user.

Often genetic results are received in PDF format and then scanned into the EHR. Many health systems are working with their EHR vendor partners to make ordering genetic tests and receiving results easier for clinicians. 

"Genomics is really just the next evolution of what we think of as a medical record," Lindemann said. "There are two things that we've enabled our clients to do. The first thing is, you need a place to store your genomic data so we had to build the database and the locations that store genetic material in the EHR right alongside their problems and their orders and their social determinants data. We had to make a genomically-enabled EHR first. And that's many years in the making, and that's something that our customers are implementing."

There are 40 health systems now live with that service and an additional 15 organizations are installing it, he said.

Epic also developed a network to integrate with genomics labs and biotechs like Tempus to enable physicians to seamlessly order genomics tests and view results within the EHR. Physicians can now order tests and access discrete biomarkers within the patient’s electronic health record to offer personalized cancer treatment options, as one example.

This gives physicians the ability to practice precision medicine directly in the EMR.

Epic also launched a life sciences program last fall aimed at facilitating clinical trial matchmaking

"We are working on integrating clinical care and research care into a single integrated experience. We've created software that allows clinical trials to run alongside clinical care in Epic," Lindemann said.

Providers currently rely on Epic to oversee more than 100,000 studies with 4.7 million patients, according to the company. 

The company is stepping up its work to match participating providers with clinical trial opportunities, offer providers customized Cosmos data queries, make trials more accessible by mitigating staffing barriers, increase efficiency by minimizing repetitive tasks, connect stakeholders through one system and support clinicians with data on when patients might qualify for a trial, according to the company.

"What the life sciences platform does is it helps to run clinical trials at scale. It allows someone to go in and build all of the elements for a clinical trial from an IT perspective once and pushed as a downloadable package to the sites that want to run the trial," he said. "Our sites can run more trials with less of an IT footprint and quickly decide if they want to move forward with a trial. That also means that they can more quickly find opportunities for their own patients."