How Oracle Developed Its Clinical Digital Assistant

Ryan McFarland, M.D., of Hudson Physicians in St. Croix County, Wis., was one of the first Oracle Health customers to deploy the new Oracle Clinical Digital Assistant, which combines generative AI and multimodal voice and screen-driven assistance. Healthcare Innovation recently spoke to McFarland and Jigar Patel, M.D, senior director of healthcare solutions at Oracle, about the potential for this solution. 

The Oracle Clinical Digital Assistant is available for integration in ambulatory clinics that use the Oracle Health EHR. Oracle says the solution "transforms the doctor-patient experience by combining clinical automation, conversation-based note generation, and proposed clinical follow-ups directly at the point of care.”

Healthcare Innovation: Dr. Patel, what were some of the issues you were hearing about from customers that the Oracle Clinical Digital Assistant solution might help address?

Patel: One main area is decreasing how much time it takes to create a note. A physician's note today is reiterating and formulating what the conversation with the patient was, and that is time-consuming. Our data shows that, depending on the clinician, it can be between 30% and 50% of the time they spend in the medical record. So it is a large portion of what they do inside the medical record, and it's ripe for automation.

Additionally, how do I potentially automate and make it easier to find information? For almost my entire career, people have said to me: Why doesn't it work like my iPhone, and I now get to say it does work like that, because it is on an iPhone, and in particular, the multi-modality of the solution: touch, type, swipe and voice. We’re even helping to automate orders. That's a capability we're starting to work on and continuing to evolve and again, it can be derived from the conversation. 

If you break apart what a physician essentially does in the EHR, it's three main things. It's review the chart, do their documentation, and write their orders. So when you break it down into those very basic building blocks, we're trying to address a little bit of all of those things, so that when we look at the full time spent in the EHR, we can shave off in all of those areas.

HCI: When I go in and talk to my primary care doctor, he's typing most of the time that we're talking, and he looks up occasionally and makes eye contact and then goes back down to typing again. Does this have the potential to allow doctors to do less of that during the conversation? 

Patel: 100%. We’ve heard from numerous clinicians in our beta programs that that is absolutely true, that they are able to get to go back to focusing on the patient. When I took some engineers to visit some of our clients and shadow some clinic visits, the first thing they noticed, and the first thing they commented back to me was that every single one of them had a moment in one of their visits where the doctor took the computer, pushed it aside, and turned to the patient and talked directly to them, because they knew for that moment it was very, very important to remove the technology and have a human conversation, so bringing that back is part of the goal.

HCI: We write about some startup companies in this generative AI space — companies like Abridge. Did Oracle have to look at the market and decide if it wanted to build this in house or partner with one of these other companies  developing something new in this space?

Patel: Yes, absolutely. Oracle did its market research. This was before I joined the group, so I don't know all of the machinations that went into it, but they also asked: what tools do we already have in the bag? We are Oracle Health, but the larger Oracle is a database and cloud company at its core. And in our cloud, we have aIl these services available to our customers that we can reuse. We had Oracle Digital Assistant, which is a chatbot that engages with a person to really understand and help them with some issue. It’s usually retail-related or something like that, right? But Oracle Digital Assistant also has speech language generative AI capabilities, and it's multi-modal, it's voice, touch, type, swipe. So it had a lot of the bones that we needed. We had experience and we had knowledge in this space. It was really honing it and taking it to a healthcare use case, and to a provider use case, that would be something that was unique about that.’

HCI: So perhaps Cerner by itself would have made a different decision…

Patel: When I joined the product management team and started to learn a lot more about all these tools in the toolbox, I was completely flabbergasted, because — to your point — as Cerner, we didn't have those. We had to partner with somebody and assess what it’s going to cost. With Oracle, the barrier is a lot lower, because we do have these tools, and we know we can engineer them, and we have a cloud that is accelerating from an adoption and utilization perspective. 

Hudson Physicians is an independently owned, multi-specialty group with 60 providers. It is primarily primary care with family medicine, pediatrics, OB/GYN and other specialists as well as an urgent care. It has been a Cerner customer for more than 10 years and was the first beta client and started testing the Clinical Digital Assistant product in October 2023.

HCI: Dr. McFarland, Oracle Health says that deploying the solution is helping providers save from 20% to 40% in documentation time. Have your providers seen that kind of impact? 

McFarland: Yes, at least that. We had the small group of us who initially started using it. We had scribes before. The initial group of us, it was going from scribes to this, but the vast majority of our providers didn't have scribes, and once we rolled it out to them, it was a huge improvement for them in terms of documentation completion. We had a handful of docs who were routinely two to six weeks behind on documentation and they are now at most one to two days behind, if at all. We went from people having “pajama time” to over 90% now who are not finishing documentation at home.

HCI: Are you working on deploying other new features? 

McFarland: Currently we are working on the order entry. I'm part of a couple of beta testers working on the Clinical Digital Assistant activating the problem list and putting in referrals for us, which is nice, because I'm not hunting and pecking during the visit or forgetting to do it. I am looking very forward to their plan with the full order entry where it’s proposing the meds, the lab orders and everything else to me, because that will save me computer time. Everything this is doing just makes more face time available with the patients and keeping me more engaged in that and less in the computer. 

HCI: Do you think it works especially well in primary care or family medicine, or does it work just as well for the for the specialists in your group? 

McFarland: It works just as well for the specialists as for primary care. Our surgeon is 60. After he started using it, he said he'll never work another day without it. We have a doc who had previously retired but worked part time in our urgent care, and we got him to use it, and he said, “Hmm, that’s the best note that I've written in my entire career!”

The documentation is better from a billing standpoint. I think the order entry, the AI search capabilities, the time savings from a chart review standpoint will be huge. As excited as I've been for what this has done, I think I might even be more excited about what's next, from both an efficiency, time-saving standpoint, but also a patient care standpoint, I think will be much better. There's going to be less missed data, imaging and lab results as we go forward. 

HCI: Dr. Patel, could you describe how you worked with these beta customers to incorporate their feedback into the solution? 

Patel: Early on, we had a client development program, before we even went to beta. We met with five clients every month, and we brought to them initial designs, initial questions from a practice perspective. We validated workflows, user interfaces and a bunch of things with those five clients even before we got to a working piece of software. 

The software was getting built behind the scenes. We were meeting with customers, continuing to improve that, gather feedback on what we were doing. Our initial beta customers really provided a huge amount of feedback. We ended up with 13 total beta customers. Early on they would say things like “This is not organized like I would write a note,” or “It missed these details, and it added these details but I don't need those details.”

By the time we got to number 13, the feedback was generally positive in that we had sorted out a good way to organize the information for clinicians. We had figured out how to tell the story of the patient and format it in a way that the clinicians were comfortable with. Those are all things that we learned over time. We also learned that some clinicians are very particular about how their note looks and what's in their note, and they have a style to it that they prefer. So there's a style and a content and a structure that is not the same from provider to provider, and our beta phases help to confirm all of that.