We’ve only seen a few times where something has dramatically changed the EHR product roadmap. The first time this happened was when the HITECH act and associated $36 billion of stimulus money was included to stimulate adoption of EHR software. In order to get access to that stimulus money, you had to use a certified EHR. If you were an EHR company at this time, you quickly realized that you weren’t going to make any sales if you weren’t certified. As such, the EHR product roadmap changed to ensure that you could meet the EHR certification and meaningful use requirements. In fact, one could argue that it hijacked much of the EHR development roadmap for 5-10 years, but that’s a topic for another day.
When I recently attended the eClinicalWorks Enterprise Summit, I saw something pretty similar when it came to how AI, large language models, and automation had dramatically changed the eClinicalWorks product roadmap. You may remember that about 7 months ago I wrote about the big announcements made at the eClinicalWorks user conference. While they offered some impressive features and enhancements for their users, there was almost no mention of AI and large language models. 7 months later on stage, the CEO of eClinicalWorks, Girish Kumar Navani, did an hour or more demo of all the new AI, large language model, and automation features they’d built into their software in just the past few months.
It was quite astounding to see what they’d done in such a short period of time even as compared to what Colin wrote about a few weeks ago. However, what was more interesting was that unlike meaningful use and EHR certification where EHR vendors basically had to code to the test, eClinicalWorks could take these amazing new technologies and customize them specifically for their users’ needs. Plus, these new technologies were sprinkled throughout the entire EHR. That means that everyone in a practice including the patient are going to benefit.
Let’s dive into some of what eClinicalWorks shared so you can see a bit more of how they’re approaching these technologies now. Although, it’s worth noting that there only a few months in. This is definitely just the start. Plus, Navani did offer some important warnings about the technology too.
To kick off the event, the MC suggested we prepare to have our jaws hit the floor. Plus, he jokingly said that if our jaws weren’t closer to the floor by the end, then we may want to get checked for lock jaw. I personally felt like they lived up to this expectation. There were some specific jaw dropping demos and the demos in aggregate were jaw dropping as a whole.
Great comparison by Girish Kumar Navani, CEO at @eClinicalWorks of his experience when the iPhone was introduced to what’s happening now with AI.
Although, he notes that the smartphone took years to really make an impact. AI has started making the impact in weeks. pic.twitter.com/azsGGPcJdB
— John Lynn (@techguy) May 23, 2023
Navani was spot on in his reference to AI feeling similar to an iPhone like moment. However, the iPhone took years to fully mature. The fact that AI has developed this quickly over the last few months is astounding to watch. We’ve seen this with the growth of technology like ChatGPT which has blown adoption numbers out of the water.
While the new AI options feels like an overnight success, Navani did reference that Turing was doing some of the first AI learning models in World War II as is depicted in the move The Imitation Game (which you should definitely watch if you haven’t seen it). Plus, many of us have things like Roomba robots cleaning our houses which is part of the evolution of these models and we all remember how well Watson did on Jeopardy. While Watson has largely been a disappointment in healthcare, it was still a demo and building block for where we’re at today.
These technologies aren’t without their limitations. For example, ChatGPT is great at filling in the blanks of sentences, but Navani noted that it often doesn’t know when to stop.
ChatGPT has been trained like you have been trained to fill in the blank on sentences.
One challenge with it is it doesn’t know when to stop. @eClinicalWorks
— John Lynn (@techguy) May 23, 2023
One other amazing thing that is happening is the confluence of multiple technologies, data sources, etc. In eClinicalWorks they did a demo of how it was magical to combine their search feature called Prisma together with large language models.
What happens when search and large language models come together?
I call it magic.
Would you like to see it inside your EHR and how an AI assistant will impact every part of your office? @eClinicalWorks
— John Lynn (@techguy) May 23, 2023
Plus, this becomes even more powerful when you include data from outside the eClinicalWorks EHR. One of the most jaw dropping demos was where the eClinicalWorks AI assistant took all the data from every source including outside labs and external EHR software and created a readable summary for the doctor. Instead of having to sift through hundreds of pages, they had a nice summary of the most important things. Since eClinicalWorks is leveraging Microsoft Azure for this, they can tune it to always include every diagnosis, for exmaple, which is quite different than ChatGPT which may ignore diagnosis that it thinks aren’t worth mentioning.
When you combine an AI assistant with the powerful health data search in Prisma that looks across data from inside eCW and other outside EHR data, the benefits compound.@eClinicalWorks
— John Lynn (@techguy) May 23, 2023
Speaking of using Microsoft Azure for these new features, Navani noted that their large language models are not public like ChatGPT and so all of the patients’ PHI will stay on private servers that are governed by BAA agreements to ensure the privacy of the data. This is key in healthcare.
In the world of AI and privacy, @eClinicalWorks is training a private model that’s covered by BAA agreements. You can’t use public options like ChatGPT without sharing PHI with public sources.
— John Lynn (@techguy) May 23, 2023
As I mentioned, the most impressive thing I saw was how eClinicalWorks had worked to incorporate these new AI, Large Language Model, automation features throughout their EHR. Check out this list of just some of the integrations they showed us at the event.
Other examples when combining AI with search and data from multiple sources, image recognition, and voice:
-Summarize the patients entire health record in a readable format
-Take a picture of an insurance card and see if the insurance is correct
-Image identification on Faxes to…— John Lynn (@techguy) May 23, 2023
One other feature that was pretty unique was something they call Playlists. For those of us in the tech world, we’d call them essentially macros where you can record a series of steps and then with one click it can do the series of steps for you in the EHR. The idea took inspiration from the gaming world. I know my sons have some click tracking automations they use to generate dollars/coins/points in the game they’re playing. It makes sense to apply them to the repeatable processes in an EHR. I’ll be interested to see what users think of these playlists that allow them to do repeatable tasks with just one click rather than 10.
Wherever you have to go through multiple screens, you can use an RPA – Robotic Process Automation (@eClinicalWorks calls them Playlists).
— John Lynn (@techguy) May 23, 2023
Navani did illustrate how quickly this technology is evolving when he shared the following:
I often get asked our 5 year roadmap. I’m starting to feel like ChatGPT where I only know the next best word.
-Girish Kumar Navani from @eClinicalWorks
— John Lynn (@techguy) May 23, 2023
He’s right that it’s hard to imagine now what large language models and AI will look like 5 years from now. We’re just at the start of what’s going to be possible, but I was impressed by eClinicalWorks creativity in how they are leveraging this new technology for their users. They also mentioned that they created something on the backend for their developers as well. It’s something that no user will see, but it saves days and days of developers time creating test scripts and evaluating new code that’s being created. Plus, it’s customized to the eClinicalWorks environment. That’s really powerful!
Needless to say, I was incredibly impressed by what they showed us at the conference. It really did give a glimpse of how AI is going to impact everything. Although, as Colin noted in his previous article, eClinicalWorks is implementing all of this in co-pilot mode instead of pilot mode. In other words, the AI us augmenting the humans rather than replacing them. That’s a really wise approach as we all get used to a new way of approaching technology in healthcare.
eClinicalWorks is a sponsor of Healthcare Scene
