Closing the Last Mile: AI's Role in Healthcare Transformation

[00:00:08] We are on HIT Like a Girl and I'm Demi Radeva. And I'm Kim Perry, the Chief Growth Officer of Intelligent. Kim, let's start with a quick intro. Can you share a bit about your background and your role in healthcare? Absolutely. So my background, I like to say, I functionally grew up in sales and account management. I spent most of my career at a Fortune 500 company, learning from what an established sales organization looks like, then moved into management consulting. And in management consulting, I've been in the healthcare practice, so I've been focused on the payers, the health systems, and really helping them with their business-driven transformations.

[00:00:38] enabled by technology. And then I moved to early stage health tech to help them through their growth and scaling period. So my participation is now as the Chief Growth Officer with Intelligent. We have been formally in market for about two years, was an R&D company for about seven years, building the maturity of the product. And we have now had our coming out party for the last couple of years. And it's been certainly a fascinating time for clinical AI in this market. We're here at Vive discussing what's coming with the new administration. From your perspective, what are the

[00:01:08] biggest policy shifts or industry changes on the horizon? There's a lot of speculation of what's to come and it's still unknown. So we're all keeping our heads down and doing what we, what's right in front of us. How is that affecting your day-to-day? It doesn't really affect my day-to-day. I just think there's a ton of activity. And I like to say upon reflection over the last three years,

[00:01:30] 2023 was really a year of education, at least for AI and healthcare. So people were still understanding what the technology is, what it could do, what the use cases we might be pointing at, understanding what potential ROIs. So it was very much a year of education. 2024 was, I like to say, the year of enlightenment. People were playing with more readily available technology. So chat GPT was introduced, other LLMs were introduced.

[00:01:56] So people were using these tools for the first time. So a lot of science fair projects, a lot of playing, but there was also enlightenment of what works and what doesn't. I do think what came out of last year was an understanding that these are powerful tools and technologies to implement, but there is going to be that last 20%. So they got 80% of the way there, but they're not solutions that are ready to scale. So how do we close that last 20%, especially in healthcare where accuracy in the data is so important?

[00:02:27] Clinical documentation remains a major burden for providers and unstructured data is a huge challenge. I'm curious, how is AI, particularly NLP, helping to structure and extract meaningful insights from the data? Documentation continues to be a challenge. Our providers are trained to use pros. They're trained to use language to communicate, and it's not easy to put their communication into discrete fields and dropdowns within the EMR.

[00:02:54] So they still need the ability to have this pros, the unstructured text. And so the burden around the documentation has been alleviated quite a bit with the ambient listening technologies. So the ability to listen to the conversation and taking the note taking off of the plate, but the reality is it's still an unstructured note going into the EMR and it's not usable downstream.

[00:03:16] And so that I think is the next generation of enhancing clinical documentation is how do we now structure the note at the point of capture so it can be usable downstream versus using manual intervention to extract the insights from the data, whether it be for billing purposes or point of care or payer use cases. There's plenty of use for what's in the unstructured note, but it starts with documentation.

[00:03:41] So how do we capture the structure at the point of documentation is really, I think, the next evolution. Do you think there are any policy or regulatory hurdles that are preventing broader adoption of this technology, or is it the technology itself? I don't think the regulatory, they're not hurdles for adoption from the regulatory side. I think it's really the technology and the accuracy of the technology and can it be usable at scale.

[00:04:08] So that I think is the barrier to adoption right now. I'm going to reflect back on that last 20%. We can get 80% of the way there. That's still not good enough to be implemented and adopted at scale. We have to get closer to that 100%. How does AI-powered NLP contribute to solving the challenge of interoperability? When you are sharing clinical data, it is still in PDFs, faxes, image-based documents.

[00:04:36] And those are challenging to work with. That is not clean data at all. And so where NLP comes in and some of the other tooling and technology that we've created is taking those dirty data challenges and being able to extract not only the text from those documents, but highly structuring the text to be able to be used for a lot of different use cases. So NLP is structuring that unstructured text to be able to be used by computers, readable by computers.

[00:05:05] What is the common misconception about NLP and AI-assisted clinical documentation that you think needs to be debunked? The common misconception around NLP as it relates to clinical documentation is that it's easy. That there are a lot of tools that can be applied to generic NLP tools can be applied to clinical language. And that's not true. Clinical language is a different language. It's not English. And traditional approaches to NLP haven't worked.

[00:05:34] And so taking a unique approach to NLP, which for us is coding too. I'm not going to necessarily go into that. But yeah, so I think using traditional NLP approaches applied to clinical language isn't working. It's not going to get the level of features and accuracy that is needed in healthcare by using the traditional approaches or traditional tooling out there. So you're going to need a healthcare-specific NLP engine, a clinical grade.

[00:06:00] Does that mean providers are actually involved in the process of building the NLP and or coders and or, yeah, how do we make it better? The approach that Intelligent took was a right brain, left brain approach. So our CEO is a physician. So he understands the language of medicine. Our CTO is a researcher, right? He is a professor in computer science with a specialty in NLP, understands the most modern approaches to solving language, but doesn't understand the language.

[00:06:28] So you needed to really blend those two, the clinical lens as well as the technology lens to really get after the challenge, which is understanding the medical language. I'm curious. Many health systems hesitate to adopt AI due to concerns of accuracy, bias, and even regulatory uncertainty. I love what you guys are doing as in blending the two skill sets.

[00:06:51] And so I'm curious, are there additional things that can be done around accuracy, bias that can help advance the technology adoption? Adoption, certainly with physicians or providers, they need to trust the technology. And the way they gain the trust is to have transparency on where they found the information. And so what we've done, again, is provide the transparency, provide the explainability. So there is proof in the answers that we're presenting in front of a provider of where it came from.

[00:07:20] So they can point back to the source of truth versus some of the generative tooling is certainly not explainable and oftentimes hallucinates and makes things up. So I think to overcome the challenge of adoption is you have to have the transparency. You have to have the proof. Looking ahead, I'm curious, what's the next big frontier for AI in healthcare? The next frontier is I will continue to go back to closing that last mile, closing the gap. To be usable in healthcare, it has to be trained from healthcare data, focused on healthcare.

[00:07:48] You can't use general tooling for the healthcare ecosystem. So I believe the next frontier into clinical AIs and driving towards the adoption is really finding those healthcare-specific tools that can close that last mile and drive the adoption. So move beyond the awakening into the execution and adoption. I'm curious, can you paint the picture for us on once we are beyond the adoption? Like, what does the world look like? Hopefully more efficient. Yeah.

[00:08:18] More cost-effective. This industry certainly needs to transform. It is not in a position where we can currently sustain the existing healthcare ecosystem the way we're going. So we need to use tools like this to advance and transform. So hopefully in the next five years, we are going to see a much more efficient world. We don't have enough physicians or clinicians to give us care. So we need to take all of the administration burden as much as possible off of their plates so we can get back to giving care.

[00:08:44] When I think about giving care, I'm curious, is the audience just providers or does the audience also include caregivers? Is it who is actually going to be utilizing this type of technology? I would say the entire healthcare ecosystem that uses clinical data is participating. So you have the payers that are using clinical data to optimize their processes. They're used to using claims data only. Now with interoperability and the ability to get clinical data, their worlds can change as well. They can optimize. They can be better risk adjusters.

[00:09:13] They can be better with prior authorizations and more timely with the prior authorizations, the payment integrity use cases. So payers are certainly going to contribute. Definitely on the health system side and the providers. But there's also research. There's a lot of personas that can use this technology. And then life sciences. Life sciences and the biotech, they're very eager to use clinical data to advance drug discovery, move through clinical trials much, much faster. And so if you can unlock the data, the next frontier is in front of us.

[00:09:42] But it starts with data. And unfortunately today, 80% of it's locked up in unstructured data. And how do we liberate it? The technology is the issue. It's not the policies necessarily. So I was like, is there any policies that are prohibiting payers from using this data today or life science from using this data today? But it sounds, again, it's more on the technology side. That's the barrier today. Is the technology advanced enough to be used? Yeah. I certainly can appreciate the need for transparency.

[00:10:12] Is the transparency the latest kind of version of the transparency policies? Is that enough? Time will tell. I don't know if I have an opinion. Yeah. Okay. And then where can our audience find and follow your work? www.intelligent.com. Yeah. And then I'm on LinkedIn. Amazing. Thanks so much, Kim. Yes. Thank you. Thanks for listening. You can learn more about us or this guest by going to our website or visiting us on any of the socials with the handle HitLikeAGirlPod. Thanks again. See you soon.

[00:10:41] Again, thank you so much for listening to the Hit Like a Girl podcast. I am truly grateful for you and I'm wondering if you could do me a quick favor. Would you be willing to follow or subscribe to this podcast or maybe leave us a rating or review? Or if you're feeling extra generous, would you share this episode on your Instagram stories or with a friend? All those things help us podcasters out so much. I'm the show's host, Joy Rios, and I'll see you next time.