Editor’s Note: This is the fourth in a series of articles from top voices in healthcare IT discussing the use of artificial intelligence in healthcare. To read the first part, about Dr. John Halamka from the Mayo Clinic: click here. To read the second interview, with Dr. Aalpen Patel at Geisinger, click here. To read the third, with Helen Waters of Meditech, click here.
Sumit Rana, executive vice president of research and development at electronic health records company Epic, works at one of the centers of artificial intelligence efforts in healthcare IT. His days are filled with creating ways in which AI can help doctors and nurses today and in the future.
Epic uses AI and ambient listening technology in an effort to improve patient-provider interactions. It allows physicians to generate progress notes based on a conversation between patient and provider in the exam room. This allows doctors to immediately create a draft answer to a patient’s question. It shows healthcare providers what’s new with a patient since they last saw the patient. And it’s working on AI that helps staff with medical coding.
Healthcare IT news sat down with Rana for a deep dive into all these applications of artificial intelligence and more.
Q. Epic uses AI and ambient listening technology with the aim of improving interactions between patients and healthcare providers. Some healthcare IT leaders at healthcare organizations are concerned. Are these technologies reliable today?
A. I call myself a skeptical optimist. It’s always good to ask questions. I think context matters here. Often, AI tools, especially generative AI, can be used to help people be more efficient. And in these contexts, for example when drafting a possible response to a message from a patient, it is very important that the AI always keeps the doctor informed. These tools are very useful.
I have a few rules of thumb. First, does the AI tool fully automate the process, or is there a human in the loop making the final decision? Be wary of the former. A much higher bar must be set. Second, from a user experience perspective, is it always clear that the software clearly indicates if and when AI is involved and how this happens?
And then I guess the third part is: does the application provide enough audit and monitoring data to track both AI usage and associated outcomes and decisions in the workflow?
Q. Specifically to Epic: Your EHR now uses AI to generate progress notes from conversations between patients and providers in the exam rooms. Explain how this works and what results you see.
A. So the doctor asks the patient at the beginning of the visit if he/she can record the visit to help with documentation, and most patients are fine with it. And then they turn that on through their mobile phone application.
If necessary, they can instruct the AI at any time to stop listening or start over, depending on whether there is anything the patient is uncomfortable with. The AI listens to that environmental conversation. And once that conversation ends, a few moments later, the note is updated with a draft of what was discussed, and then a clinician can look at that note and make changes similar to what they might do in a dictation workflow. and then they finalize their documentation.
This is now used on over 30 sites and I just want to share some results from a few of the sites. So one site reported an average savings of five and a half hours per week. Another study looked at the time physicians spent after clinical hours and they saw a 76% reduction in the time physicians spent after clinical hours, which I think is quite substantial.
And then a third site said that more than 60% of their users reported an improvement in documentation quality. I think this is another good aspect to look at in terms of how well the notes are written.
Q. Your healthcare provider’s users can use AI to generate a draft answer to a patient’s question. How does this work and what results do you see?
A. The way this works is that when the caregiver opens a message from a patient, there is a draft waiting for them and they only have two choices. They can start with the draft and then make changes to it as they see fit, or they can start with a blank text box, as it was before. You press reply and then type a message.
A response is never automatically sent, so start with the draft or start with blank text. And in terms of results, we have a number of sites trying to do deeper qualitative research on this, but I’ll provide some early feedback.
What we heard first from providers is that providers like the AI can pre-fetch data in the patient record and synthesize it for them. They have reported spending less time collecting data themselves. They also report that it sometimes helps overcome writer’s block, and that it creates more descriptive and likely more empathetic responses to patients.
Finally, we had to teach the AI to write in the doctor’s voice. For example, you may have a doctor who says, “Dear Bill.” Another might say, “Hi, Bill.” A third might just say “Bill,” and a fourth might not say your name at all. And if the writing style does not match what the patient is used to from that provider, it can feel very unnatural. So we have to teach the AI to do that too.
Q. It’s interesting that you say AI can sometimes be more empathetic.
A. There was an early example of a patient who had been on holiday on another continent, I think in Europe, and had sent a message. And not only did the AI compose a response, but one of the things the response contained was something along the lines of, and I’m paraphrasing, “I’m sorry you’re going through this, I hope your vacation was good.” ”
And the doctor said, “Boy, I don’t know if I would have written that in my busy time.” But it actually strengthens that human connection, and it’s a good thing to write.
So I think it’s not so much that people aren’t otherwise empathetic, but that when you have a busy day and you’re seeing patient after patient, it’s sometimes harder to write those kinds of messages. And if the AI can help, that’s a good thing, and it strengthens that relationship.
Q. How does AI in your EHR show a provider what’s new in a patient since they were last seen?
A. There were many options to do this in the system even before generative AI. But I’ll talk about what we’re doing with generative AI.
So our users are starting to explore a new feature we’ve added, summarizing previous comments in the graph. And the most important thing here is that the summary must be specific to the context of the physician and the healthcare environment.
For example, if you are a triage in the emergency room, the way you summarize and what you choose will be different than if you do a physical after five years. And summarizing actually means throwing away facts. And so, what facts should we preserve? Which facts to throw out matters based on that context.
The other big lesson we learned is that it is super important that the provider can see the citations and see not only what the summary is, but also what facts in the medical record were used to generate this summary.
And this is crucial from our perspective to limit the risk of hallucinations that I mentioned earlier and to increase confidence in the technology.
Q. How does Epic use AI to assist with medical coding? What are the results so far?
A. The AI will list potential procedure and diagnosis codes based on clinical documentation, and then a coder can easily add them to the charging session without having to start from scratch for each visit and each proposed code.
It also highlights supporting documentation in the map without having to go back to hunting and gathering. As for results, it is not yet available for production use. We’re getting ready to release it.
Our initial internal testing has shown that generative AI is quite effective at translating free-text clinical documentation into discrete codes for billing. This will make the coding workflow more efficient.
More broadly, it’s one of the things we’re doing to reduce administrative overhead for organizations, and we expect to automate more and more of these types of administrative tasks in the future.
Editor’s note: Click here to watch the video of this interview, which includes bonus content not included in this story.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.