Building Trust in Clinical AI–What Hospital Leaders Need to Know About Evidence‑Based Decision Support [PODCAST]

← Back to All Podcasts

Building Trust in Clinical AI–What Hospital Leaders Need to Know About Evidence‑Based Decision Support

In this episode, Dr. Claudine Lott, Physician Executive for Commercial Transformation and Implementation at Elsevier, discusses building trust and clinical AI, what hospital leaders need to know about evidence-based decision support.

Highlights of this episode include:

• What ClinicalKey AI is
• How AI enhanced clinical decision support tools can help organizations improve both clinical efficiency and financial performance
• How AI-powered tools can support clinicians in real time to reduce errors, avoid denials, and strengthen the overall revenue cycle
• ROI opportunities for health systems adopting AI-powered clinical intelligence
• How AI-powered tools remain evidence-based, transparent, and aligned with clinical best practices
• The most common misconceptions hospital leaders have about implementing AI and clinical workflows

Kelly Wisness: Hi, this is Kelly Wisness. Welcome back to the award-winning Hospital Finance Podcast. We’re pleased to welcome Dr. Claudine Lott. She is a board-certified family medicine physician who is passionate about developing and implementing tech-based clinical solutions that improve both patient outcomes and provider experience. As physician executive for commercial transformation and implementation at Elsevier, she supports the development and deployment of their reference products for healthcare providers, including ClinicalKey AI. Dr. Lott received her medical degree from the University of Massachusetts Medical School and completed her residency at White Memorial Medical Center. She served as a primary care physician at the federally qualified Santa Cruz Community Health Center, where she was promoted to site medical director. She then joined Healthcare Startup Crossover Health, where she contributed to the development and expansion of their virtual care model, as well as the creation and deployment of their Patient Engagement Technology Platform. Since joining Elsevier in July 2023, Dr. Lott works cross-functionally to support key initiatives, including customer implementations, product development, and change management.

In this episode, we’re discussing building trust and clinical AI, what hospital leaders need to know about evidence-based decision support. Welcome, and thank you for joining us, Claudine.

Dr. Claudine Lott: Thanks for having me on. Appreciate it.

Kelly: Well, it’s great to have you. And let’s go ahead and jump in. So, what is ClinicalKey AI, and how are its new capabilities designed to reduce clinician burden and improve documentation accuracy?

Claudine: So ClinicalKey AI is Elsevier’s flagship generative AI tool that’s designed for clinician use to quickly surface the latest evidence at the point of care to support clinical decision-making. And just to take a step back and provide some context, so here at Elsevier, we’re an almost 150-year-old publishing company. So, for almost 150 years, our role has been as a provider of scientific information and clinical evidence that clinicians can use in their decision-making and in their patient care. And as we’ve moved into more and more clinical solutions, that’s always been kind of our guiding North Star. And so with generative AI coming on the scene, we’ve really thought about, okay, how do we use this emerging technology in our role as a provider of clinical evidence, scientific information to really further that goal of getting clinicians what they need to make decisions and take the best possible care of patients as quickly, accurately, and effectively as possible. And rather than just sort of slapping generative AI on everything because that’s sort of the new thing to do, how do we really leverage this new tool to solve that problem? So ClinicalKey AI is a conversational search tool. The clinician’s able to ask a question in natural language, almost like they might ask a colleague. And then the system goes and searches a curated set of content that we’ve given to it. So that includes much of our Elsevier clinical content, but also some non-Elsevier sources as well, and searches for information and then surfaces that for the clinician. So, it’s not replacing their clinical knowledge or decision-making, but it’s really supporting them by getting the information that they need and we’ve been developing and iterating on this tool for several years now, constantly thinking about how do we make it better and more suited to this clinician use case. So constantly thinking about how we expand our handpicked content sources, thinking about making sure that we always have traceability so clinicians can see where the information is coming from, citation verification, and always thinking about technology upgrades. So, things like privacy, security, and supporting HIPAA compliant use.

Kelly: Wow, that ClinicalKey AI technology sounds really fascinating. So how can AI enhanced clinical decision support tools help organizations improve both clinical efficiency and financial performance?

Claudine: So clinically, the biggest win is what we might call speed to evidence. So, we’re in a situation now where patients are increasingly more complex. Medical knowledge is expanding exponentially. And so, getting that information that is really tailored to the clinical situation as quickly as possible is going to enhance clinical efficiency so that AI enhanced decision support can really surface the most relevant trusted information. In seconds, really supporting those consistent decisions under time pressure and given all those other complexities. In terms of how that clinical efficiency translates into financial performance, I think this is something that we’re going to see continuing to evolve as more and more organizations are integrating these types of tools. So certainly, it makes sense that improving clinical efficiency, improving the quality of care is going to translate into financial performance, but sometimes that ROI can be a little bit difficult to quantify. So, I think that we’re going to see those benchmarks continuing to evolve as more and more institutions are implementing these tools.

Kelly: Yeah, and I love what you said at the beginning, that speed to evidence. I love that. So, what should hospital healthcare system operation leaders look for in the first six to 12 months to know an AI tool is truly delivering clinical value?

Claudine: Yeah, I think this is a great question and something that a lot of both vendors and organizational leaders are really thinking about. Because again, these tools are still new. We’re still seeing how they affect healthcare and how they affect the clinical workflows. And so, we’re still really figuring out how we quantify this sort of clinical value. So, thinking about sort of what can you look for at that 6- or 12-month point to know if your tool is delivering that clinical value. For things like time-saving, improvement of quality of care, those things can be hard to really quantify. And also some of the benefits of generative AI tools, as we mentioned, is that helping clinicians provide faster and better care, it leads to a better experience for those clinicians, for that care team, really addressing that sort of fourth leg of the quadruple aim. But again, that’s something that can be hard to quantify. So, in thinking about, okay, what are some of the metrics that we can sort of look at those sort of checkpoints to see the value that these tools are providing? Certainly, usage metrics are one aspect in terms of just seeing how many providers are using the product, how often are they using it. But that’s only sort of one aspect of it.

Given that there’s more of a– there may be more of a qualitative improvement, some customers and organizations that we’ve seen have chosen to use surveys. So, for example, we had one customer who was utilizing ClinicalKey AI, who did a survey to ask their users after they had trialed it for a given period of time to rate the improvement in their ability to conduct patient care, their confidence in their clinical decision making, and their time saved. And so, they were able to, through that sort of surveying of the users, to sort of quantify the improvements they were seeing in all those areas in that way. And this is also a place where having a clinical champion really involved in the implementation process and in the adoption of these tools can help because checking in with those champions can really connect you to understand, again, some of those sort of improvements in experience that can be a little difficult to quantify. And I think it also comes down to for organizational leaders thinking about what is the problem that the generative AI tool was implemented to solve. So, as I kind of mentioned before, we don’t just want to throw a tool at clinicians just to give them something AI because AI is sort of new and exciting now. We really want to think about, “Okay, what problem are we solving with this?” And from there, then at those checkpoints, I think you have a place to go back and say, “Okay, here’s the problem we were trying to solve. What progress have we made on that?” And use that to kind of quantify the value?

Kelly: Yeah, I know it is difficult to quantify that value there, at least for now. So, it sounds like you guys are making progress with that. So how can AI-powered tools support clinicians in real time to reduce errors, avoid denials, and strengthen the overall revenue cycle?

Claudine: In real time, AI-powered decision support can reduce errors by helping clinicians quickly sort of cross-check their decisions against trusted evidence, or by getting them information that they need to make that decision quickly, especially in an environment that’s high-pressure and time-constrained. In thinking about aspects of revenue cycle management like coding integrity, managing denials, having that grounding in clinical evidence is so vital. Having that documentation that’s based in clear and trusted evidence that’s traceable is really going to provide that sort of grounding and foundation for the decisions that are being made and then the documentation that’s going into that. And that’s going to really support those aspects of the revenue cycle.

Kelly: Yeah, thank you. That makes a lot of sense. So where do you see the strongest ROI opportunities for health systems adopting AI-powered clinical intelligence?

Claudine: So, I think there’s three sort of big ROI opportunities that I see. So first of all, speed, as we’ve discussed, just making decisions more quickly frees up more time for patient care, can help with reduction of administrative burden, and just really free up clinician time. So that just speed is a huge part of it. And then I think the second part is the accuracy and that strong evidence base that I mentioned. So, making sure that decisions are based in strong clinical evidence and that that is really documented in a well-supported way, that’s going to not only support patient care but also those aspects of revenue cycle management that we mentioned. And then I think another opportunity is thinking about standardization. So, there’s definitely an art to the practice of healthcare. So, we may still see some variation in the way that different clinicians might approach the same problem. And having an evidence-based tool has the potential to support more sort of consistent practice patterns across settings. So, I think that standardization and ability to make sure that all care team members have access to evidence on which to base their decisions is another significant opportunity.

Kelly: Sure. Sounds like there are quite a few really strong opportunities there that you shared with us. So how is Elsevier ensuring that AI-powered tools remain evidence-based, transparent, and aligned with clinical best practices?

Claudine: So, as we mentioned, we really anchor clinical key AI in peer-reviewed, copyright-cleared medical evidence. So that includes full-text journal articles as well as journal abstracts, clinical practice guidelines from different organizations, full-text medical textbooks. And we’re constantly thinking about curating that content set, what we need to add, what we want to expand on, how we want to adjust it to make sure that it’s really providing what clinicians need. And we also keep the content current. So, we have a content pipeline that updates every 24 hours. So, the outputs are really reflecting the latest evidence and guidelines as much as possible.

And we’ve really tried to build in that transparency, that traceability, so that the clinician can really see down to the paragraph where that information is coming from. So, they can have that trust. They know that the citation is not being hallucinated or made up by the AI. They can have that trust in where the information is going from, and they can also do a deeper dive if there’s a topic that they want to explore further. So, it really gives them that ability as well. And we use a clinician in the loop approach with a rigorous evaluation framework to continually test the system, follow up on feedback that we get with users, and really just make sure that we’re maintaining and constantly improving the quality of the insights we’re providing.

Kelly: Well, it sounds like that trust is very important to your team there, and that’s appreciated. And you all take that responsibility very seriously.

Claudine: Definitely.

Kelly: Yeah. So, Claudine, from a physician executive’s perspective, what are the most common misconceptions hospital leaders have about implementing AI and clinical workflows? And what advice would you give them as they evaluate solutions?

Claudine: So, there’s three main points about successful adoption of clinical generative AI tools that we’ve seen from our teams and customers, as well as what we’ve been hearing from others in the industry. So, I think these are a great starting point for organizational leaders who are considering implementing a generative AI tool. So first and foremost, as we mentioned before, really knowing the problem that you’re solving with the generative AI tool. So, if you have a generative AI tool, but it’s not solving a problem for the clinician, it’s not making their experience and their care better in some way, nobody’s going to want to adopt that. Nobody’s going to take the time out of their schedule to learn and integrate something new. So really knowing the problem that you’re solving and making sure that you have a tool that fits that.

And so, for us at Elsevier, as I mentioned, we’re seeing this problem of increasing patient complexity, increasing medical knowledge beyond what anyone can sort of memorize. And so, thinking about, okay in our role as a provider of clinical content, how do we use this technology to really solve that problem? So that’s the first part. The second aspect is making sure that the tool is accessible and easy to use, that it’s really embedded in the workflow. Because even if you have a tool that does solve a problem for the clinician, if you’re going to implement something that they have to leave their workflow to try to utilize, that’s not something that they’re going to want to adopt. And certainly, if you’re implementing something because you want to increase their speed and efficiency, if it’s an inefficient process, that’s not going to be helpful at all. So, for us, that consideration goes into things like making sure that our product is integrated into the EHR, having an API option, and basically just making sure that the tool is really in the workflow where the clinician is making that decision.

And the last point is really coming back to this point about trust, because I think some of the misperceptions about clinical AI tools themselves are really related to a lack of understanding of how these tools work. So not understanding that a standalone general use large language model is going to answer clinical questions just based on its training. It’s not actually going to be going out and searching. Whereas a tool that pairs LLM capabilities with retrieval is going to actually be searching and surfacing information in that way. Knowing that some general use tools are drawing from perhaps the whole internet or from sources that are unclear as opposed to a tool that is really clear about where the content is coming from. Risks of using a tool that the privacy protections are not clear. So, all of this sort of lack of understanding contributes to lack of trust, and that’s going to make sure that, again, this is not something that is going to be widely adopted.

And I think that here this is a place where organizational leaders need to think about support from both internal clinical champions and strong vendor partnerships. Because those internal clinical champions, as I mentioned, they’re going to have that deep clinical expertise of the workflow. They’re going to know those problems that the clinicians are facing. And they’re able to be a voice to their peers to say, okay, here’s how this tool works, here’s why it’s trustworthy, and here’s how it’s going to solve the problems that you’re facing. And that’s going to really lead to more successful adoption.

Similarly, having a partnership with a vendor that’s trustworthy, that you’re able to work with them, you’re able to provide feedback and get support for your implementation and your adoption efforts are also very important. And we at Elsevier, as a vendor, really do try to be partners to our customers in that way in supporting them and helping them understand our tools, how they work, how they can benefit them. So those are kind of the three main points that I think are really helpful in thinking about implementing generative AI tools in the clinical setting.

Kelly: Right. It sounds like having those champions and partners are really key to success there. So, Claudine, looking ahead, how do you see AI shaping the future of hospital operations and financial sustainability? And what role will Elsevier play in supporting that transformation?

Claudine: Well, it seems like AI is here to stay, right? So, I think we’re going to, in the future, see AI continuing to lead to changes in really every aspect of healthcare. In terms of clinical decision support, I think we’re going to see these clinical generative AI tools increasingly becoming like a standard layer inside these clinical workflows, so helping clinicians find information, make those quicker decisions, supporting their documentation, but really just with an increasing integration and seamlessness as these tools become more integrated and more widely used. And I think Elsevier’s role is going to be to continue to build on what we’ve been doing all along. So again, constantly thinking about how do we deliver responsible AI solutions that clinicians can trust grounded in that evidence, not replacing their clinical judgment, but really getting them the information that they need in our role as this provider of trusted clinical content and just continuing to think about usability, what features are needed, what content is needed, and how do we continuously think about supporting the clinician with this new technology.

Kelly: Thank you, Claudine, for sharing your insights with us on building trust in clinical AI, what hospital leaders need to know about evidence-based decision support. And if a listener wants to learn more or contact you to discuss this topic further, how best can they do that?

Claudine: They can definitely connect with me on LinkedIn.

Kelly: Awesome. I will do that as well, and thank you all for joining us for this episode of the Hospital Finance Podcast. Until next time…

[music] This concludes today’s episode of The Hospital Finance Podcast. For show notes and additional resources to help you protect and enhance revenue at your hospital, visit besler.holdings/podcasts. The Hospital Finance Podcast is a production of Besler Holdings.

If you have a topic that you’d like us to discuss on The Hospital Finance Podcast or if you’d like to be a guest, drop us a line at contact@besler.holdings.

---

---