A recent survey of CIOs and other health IT leaders found that two-thirds of respondents still develop AI strategies, while 20% “is being restricted or fragmented. Only 2% of IT leaders considered the provision of AI in their e-health records as “mature.”
CEO of Qventus, the AI-led care operations company that conducted the survey, and Joseph Sanford, Maryland, Head of Medical Affairs and Chief Clinical Information Officer and Director of the Institute for Digital Health Innovation (IDHI) at the University of Arkansas University of Medical Sciences (UAM), Healthcare Innovation delve deeper into the findings of the survey.
Here are some of the highlights of the research:
•When asked about key ways to assess ROI for AI investments, margin improvements ranked top (26%), followed by cost reductions (24%), staff productivity and clinician satisfaction (16%).
•54% identified increased operational efficiencies and reduced costs as the strategic goals that have the greatest impact on care operations.
•More than half of survey respondents indicated that there is a formal AI Governance Committee that collectively oversees and manages the development and deployment of AI tools.
Healthcare Innovation: Dr. Sanford, do the AI preparation responses for this survey line up with your own experience at UAMS?
Sanford: I think the answer is pretty typical of the overall healthcare range. The first forks are academic and non-academic. Where are the state-of-the-art additional resources and non-commercial interests? Among them, you have an academic who connects the undergraduate engineering department. There, you have a real startup entrepreneurial mentality. And you have a development pipeline because you're interested in doing pilots like this. UAMS is a graduate organization. We have undergraduate programs, but they are very limited to nursing and other alliance health professionals. We have a partnership with the University of Arkansas Fayetteville University, primarily to train our graduates. Perhaps most notable in their engineering school. But we don't have the same pipeline. That's what we've been working to build.
The University of Arkansas is filled with passionate physicians, early adopters and technicians. Also, there are more traditional clinicians who are always focused on patients and don't necessarily have that interest, and I'm a little bit more hesitant to be one of those bleeding doctors. And there are people who don't want to do anything with it. It's not that they're not interested. They have other responsibilities. And in the commercial space and non-academic health systems, you have a totally different personnel.
HCI: So, if you are a CIO or CMIO of a small or medium-sized community health system, is there a much less way to resources and could limit your choices regarding AI strategies?
Garg: To some extent, I feel that all healthcare now feels like it's limited in resources and not so many options. Refunds and labor shortages are constantly declining, but you are right. Probably in a small place it's strong range. One of the unifying factors across most CIO conversations is that although I don't think it's AI for AI, I know it's a really great tool in our toolbox. It can be solved a lot, but it really focuses on the problems we are solving. It needs to be appropriate clinical, but it also affects staff. It also needs to affect it in terms of financial margins and quality. In a world of limited resources, we need to be more thoughtful about the issue. We are equally interested in the health system as well as the physician group.
HCI: In the survey, only 14% of respondents told Qventus that the current AI strategy is comprehensive and clearly defined, but two-thirds of them are still in the development stage and 20% are restricted or fragmented. That 14% number seems pretty low, but I don't think people are trying to figure out the strategy of the entire system because they haven't been working on it for so long.
Garg: Yes, first of all, even if you build a strategy, things are changing quickly enough that they don't feel completely done. In 3-6 months, you know that learning will probably change a significant amount. But I think it's fair to say that the degree of possible change means that people are still working on what it means. For many, the first step is to understand operational parts, financial parts, patient fragments, understanding the technology and filling the gaps. I think that's the stage where many institutions exist.
HCI: Only 11% of surveys found self-reports that they fully implemented responsible AI capabilities, including governance and risk management tools. Dr. Sanford, can you talk about that at UAMS? Has your governance approach evolved or changed over time as you become more refined?
Sanford: Yes, I have it. We are separated from deterministic models and tools that can be examined very specifically in stochastic spaces from the perspective of reproducibility. When you do it, you can't really make an individual one-off decision because it's not really suited to it. Therefore, you need to set up a set of guide philosophies and some guardrails. You can then run unit tests to ensure that they safely fail in the traditional sense of the concept and adhere to the organization's top priority principles.
In our situation, these top priority principles are patient autonomy, data privacy, transparency and trust. We are willing to engage patients with the various tools available to us and try new things.
We are transparent about how we use our data and when patients are talking to AI assistants and flesh and bloody humans. These are new concepts of our governance.
HCI: Mudit, listening to Dr. Sanford talking about governance approaches at UAMS, does that seem pretty similar to the other health systems you work for? Or are there many different approaches to governance?
Garg: I think that in general, what he explains from the perspective of changing governance is that everyone is working on it. People agree to core principles regarding the use of data and prepayment declarations of who you are talking to. But I would say there is a big spectrum of how people are adapting to new reality.
HCI: One thing I thought I didn't ask in the survey was interested in creating whether the organization would help Chief Health AI Officers, and whether you're looking at it a lot, and whether they fit into the executive team and reporting structure?
Garg: In some cases, the CIO or CMIOS plays this role, and in some cases there are AI or Chief Transformation Officer officers created to do so. I think there are so many changes that you see more operational involvement, people are getting closer to clinical care and workflow, technology and bringing them all together. And, like I said, in some cases, these are new roles, but in almost every case, they are people who change what was expected of them to bring all of this together more closely.
HCI: Does the research address the issues of how people think about working with EHR vendors on AI innovation, how they feel they can rely on EHR vendors, and how it will affect decisions about working with startups and third-party AI companies?
Sanford: First and foremost, it depends on who your EMR partner is. There is a lot of variation in that space, so what market segment will encourage their interest and who they focus on, will promote their adoption rate. I think there are always benefits to doing something completely native within the EMR. The expertise and focus perspective has its drawbacks. If you have a comprehensive EMR, your organization won't be able to concentrate anywhere at once, right? It's not possible. So I think a healthy ecosystem will allow organizations to make enlightened decisions about whether they want to try to create themselves and partners. Every organization needs the opportunity and ability to choose between these options. I think that's a general good thing for healthcare as a whole.
Garg: People see that there are options in EHR where things are very well integrated and that they have options to get deeper into certain areas in those areas and drive profits. The survey respondents said they viewed it as a truly important issue for the organization in terms of patient benefits, in terms of ROI and in terms of margins. If you want to make sure you are the best, and that you can't afford to wait, those are the places where you tend to look for the best varieties.
HCI: We also look more and more examples of healthcare systems, especially those that come with venture arms, as well as deploy and develop them along with AI vendors.
Garg: Absolutely, I think that the true success of AI is to understand the work that is very deeply done, and therefore working very closely with our customers is essential to success in our work. So being very close to our customers is the most important thing we can have in developing these new tools.
HCI: I mentioned how rapidly this is changing. Will we redo this research next year?
Garg: That's the intention. I am confident that there will be new challenges, problems and learning opportunities next year will present.
