MD Anderson CIO to hospitals: You're not 'behind' on AI

Alexis Kayser

By Alexis Kayser

Healthcare Editor

See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.

This is a preview of the September 24 edition of Access Health—Tap here to get this newsletter delivered straight to your inbox on Thursday mornings.

Good morning. Newsweek’s AI Health Summit was yesterday, which means I’m currently sorting through a full day’s worth of conversations, notes and insights to bring you the most useful takeaways. I’ll be back to regularly scheduled programming next week.

Pulse Check

Executive perspectives on key industry issues

Dan Shoenthal is the chief innovation officer at MD Anderson Cancer Center, based in Houston.

...

I sat down with Dan Shoenthal, chief innovation officer at MD Anderson Cancer Center, in Chicago on September 15 to talk about the challenge health care leaders are confronting as AI advances faster than the people, workflows and institutions expected to put it to work. Read on for a portion of our interview.

Editor's Note: Responses have been lightly edited for length and clarity.

We’ve been hearing about the promise of AI for several years now. What’s top of mind for you as health care moves from the hype to actually trying to realize that impact?

I think for the last probably two or three years, we’ve been hearing nothing but the hype of artificial intelligence. Every conversation begins and ends with AI, it seems, but I don’t know that we’ve actually seen the impact from the hype.

There are a lot of reasons for that. I think, as providers, we’ve been going through a maturation process of: How do you even adopt this technology, especially considering how rapidly it’s changing? We’re used to technology that we implement and it’s stable for five to 10 years outside of upgrades, not technology where yesterday’s version is nothing like today’s version.

There’s a maturation process that we needed to go through as providers in how you embrace technology like that, but then also, how do you help guide the change for all of our teams? Thinking about our nursing teams, they’ve been practicing the same way for years, and now all of a sudden we’re going to be introducing AI. How can they trust it? How can they trust the tests that they used to do themselves? Obviously, with the importance of their positions and taking care of patients, now we’re asking them to trust technology more to do some of those tasks.

I think it’s a learning curve all around for us as providers. And on the technology vendor side: How do they help us understand the value? Sometimes, I don’t think they understand the value of their tools, to be honest. There’s this collaboration that’s also needed with this technology. We’ve always needed collaboration, but it has to be deeper now with the pace that things are moving and the change that’s caused by that pace.

What are vendors missing when they communicate that value to health systems? What do you need to see to know that value is actually there?

The time to realize it is longer than what I think they realize. They’re running at an even faster pace than we are, just trying to keep up with all of the potential disruptors in their industry.

I take it back to ambient technology because it’s the example everyone knows. If you look at when Abridge and others came, they went from maybe two or three vendors to 20 in months. It’s more and more of a crowded market. Their value proposition and the impact that they have is constantly evolving because they’re having to constantly keep up.

One thing is just taking the time to realize that value. It’s not immediate. The other is, honestly, a lot of the early promises were around staff-time reduction and optimizing staff time. Well, what’s the impact of that, and how do you quantify that? It’s not as if we’re laying off staff. I might be optimized today, but they’re paying me tomorrow.

What’s the value of those new things that I’m actually able to do with the time? That’s something I think a lot of the technology companies miss the mark on, being able to tell that story. And I can’t blame them, because it’s hard to tell the story.

You have to be collaborating with the provider to actually deeply understand: What are those new tasks that I wasn’t able to ever get to before that I’m now getting to? I think it’s just a longer story than past technology was in terms of what that impact is.

How are you dealing with that uncertainty while still moving forward at the pace the technology demands?

A couple things. We’ve come at it from both sides of the coin. One is setting up a strong governance foundation. I think it’s critical. For the innovation guy to talk about governance can sometimes be weird because we like to just do things, but having that strong governance brings everyone to the table.

At the end of the day, we still are driving for technology that makes an impact. Part of the governance structure we have is our conversations around that: What are the most impactful tools? We can’t do them all. There’s so much technology coming at us. We cannot do all of it. So, how do you actually identify the most impactful technology in collaboration with technology teams and business teams?

The other side of that, though, is we have to allow for experimentation. We can’t be looking for perfect for the reasons I said before. Some of this technology, again, I go back to even the ambient technology, there were a lot of time savings with providers early on, but that time was going to something else. It wasn’t really true time savings.

The traditional ROI that people think about is financial, and we have to think about it differently. But to understand that, you have to experiment. To understand how we need to evolve a role, you can write it down on paper and plan all you want, but until you actually get your hands on the technology and experiment with it, you’re never going to know.

For us, I think it’s finding that right balance of governance to drive impact with experimentation.

What does that AI governance structure look like internally?

There are two layers of governance. We have a work group that we’ve established. It’s a cross-functional work group that’s inclusive of technology teams, cyber teams, business leaders, ethics, clinical teams, leaders in legal, contracting and compliance.

Weekly, they have a set of requests that they review. They’re doing more of what I would call your traditional risk analysis.

Upstream of that, we’ve started down a journey where we have business teams collaborating with the technology teams to identify where the technology can make the most impact, ensuring that we’re aligning our technology decisions with the business strategy.

Requests will go through that process first and then hit that AI governance group. It’s layered, but it’s all layered with the hope of making the most impact that we can.

Is that governance structure still evolving?

Oh, sure. I think everybody’s is.

What are you hearing from clinicians? Change management comes up constantly in conversations about AI, particularly the difficulty of changing someone’s role or workflow. How much variation are you seeing in clinicians’ willingness to use these tools?

It’s been a blend.

I’ve used the example too much already, but I’m going to keep using it: ambient technology. Some of our providers have said, “If you remove this, I’m leaving the institution.” No, they’re not really truthfully serious, but they love it.

Others have said, “You know what? I’m efficient, but give me this other tool.” I think they all want to try it, but they have a different perspective on the things that will make the most impact for them. That’s just some of the variability in how we practice with different diseases in oncology.

There’s a passion. I think the one consistent thing is: Please move faster. Please do more. Please let us do more. But then, when you start to bring them into cost discussions and things like that, they start to realize it comes with a cost.

How do you approach experimentation in such a high-stakes clinical environment? Are there considerations unique to oncology that you might not see in a more traditional outpatient setting?

When somebody gets diagnosed with cancer, they become much more deeply involved in their care. If you look at even basic things like our MyChart usage, it’s astronomically higher than your typical ambulatory care.

Patients really get involved in their care, so they’re going to ask more questions. They’re also going to come armed. They’re going to have taken what they know of their medical records, run it through one of the frontier models, OpenAI, Claude, whatever it might be, Gemini. They’re going to have done research and work in a way that we haven’t experienced before.

I think the change is probably higher for us due to the patient’s deeper engagement in their care. They’ll ask a lot of questions about AI. They ask a lot of questions about data. They ask a lot of questions about the usage of the tech and what technology is being used as part of their care. But they also come with their own research that they’ve already done.

What kinds of questions are you hearing most often from patients about AI and how technology is being used in their care?

It’s a little bit of a mix. Some want to see it because they experience it in their daily life. Some are concerned with: How is my data being used? How is my information being used? What tools did you use? How are you still involved in that decision that was made?

They’re coming more and more informed.

Some of these issues have always existed with technology. There’s always been a risk of data with technology companies, and I just don’t think the normal patient was as informed as they are now. With everything that’s going on around us, they hear more about it and they experience it themselves.

Do you have standard guidance for clinicians when patients ask those kinds of questions?

Institutionally, we protect our patients’ data. We do not allow any usage of the data for training of models. Direct patient data is not used or transferred to the company in any way where they can take it and train on it or do whatever they may do.

For us, it is one of those areas of patient safety, to be honest with you, that people might not naturally think of. Keeping their information safe is very important to us.

We do have some standard messaging on the tools that we will provide. Not out of the box, because every scenario is a little bit different in terms of the use case and the exact language, but that’s part of the workstream review and our governance reviews. We have a standard for how our patients’ data can be used, and we hold that very firm.

At the end of the day, a patient still may decide, “I’m not comfortable with it.” That has happened, and that’s okay. I think that’s going to be another new, evolving challenge. People are used to just using the technology they have. Well, this one requires more consenting, more transparency. How we adapt workflow on the fly sometimes is also going to be part of that change that we’ll have to work through with each use case.

You mentioned patients running some of their own health information through consumer AI models. How is that changing the clinician experience?

Boy, all of this is good and bad, I think. The trick in that is it really depends on the information that the patient has, because some of it is their interpretation of their data, not their actual data. Sometimes it’s going in with part of the story.

The one thing that, to me, is positive, not having to sit there and be a clinician to answer these questions, is it opens up the dialogue between the patient and provider. It creates more of a dialogue than what you would traditionally see, where I’m a patient and I’m sitting there consuming everything you’re telling me.

There’s a little bit more of a back-and-forth because I might be a little bit more informed. I might not be accurate, and so there has to be that back and forth. But I think it has an opportunity to actually create a better experience because there’s more engagement. It’s more of a two-way conversation versus what might feel like a one-way conversation sometimes.

I was at Epic’s UGM last month, and there was a lot of discussion about new MyChart tools, including AI agents that can communicate directly with patients. Are you currently using or exploring tools that put AI directly in front of patients?

We’re in the process of deploying that. We have started with some of the, not the latest features in Epic, but some of the past features in Epic. We are also looking to work with Hippocratic AI as well.

I think for us, it goes back to agents being a whole other thing that we have to learn. Technology has always been seen as this IT responsibility to manage. Now, our operational teams have to look at that agent as another one of their workforce.

How do they give the performance review to this automated tool? How are they listening, spot-checking on calls? That agent is going to hallucinate, just like the human agent. By the way, humans hallucinate, too. So how do you manage the human? You have to replicate that with these agents.

As we’re deploying this technology, that’s another thing that we’re thinking through: Where do we put them on our org chart, and who does their performance reviews? It’s not just the technology teams.

We are exploring with them, and we’re excited because I think, for us, we see the potential not, again, as replacing people, but creating touch points with patients that we otherwise might not have been able to do because we just can’t hire enough people.

Especially, again, when somebody has cancer, they’re going to have a lot of questions, as they should, and we should be there for them. It gives us an opportunity to actually increase that connectivity.

What emerging technology are you most excited about right now?

I think if you look at technology over the past 10 or 20 years, it goes back to that connection, the human connection.

The technology that we’ve deployed over the last 10 or 20 years has actually hurt that human connection, if we’re just being honest. Electronic medical records, all the data-capture tools we have, are very critical. They create new data insights that we never would have had. But it tethers people more to machines and not necessarily the engagement that they have.

They’re worried about the alert that’s going off, the documentation requirements for quality standards, for billing. They’re tethered more and more and more to machines.

I think we’re starting to see for the first time that the technology is being reductive in nature in terms of how dependent we will be on computers. [Dr.] Lavonia [Thomas, Chief Informatics and Innovation Officer at MD Anderson] and I joke all the time. I can see a world soon where the nurse pushing around the cart isn’t a thing.

Why do we need that? There’s going to be automation that will be able to bring supplies and medication. We’ll be able to do documentation on the fly. Information that they need as part of the care is going to be on their phone. Why do we need to have all of these devices, again, that we tether people to?

Why don’t we allow them to focus on what matters most, and that’s their interaction, their engagement and taking care of the patient?

For me, it’s the realization that we’re going to see very soon technology actually creating human connection, not taking away from it.

Have you started to see that effect in any of your pilots or clinical units?

We have with the ambient technology. We’ve seen an uptick in patient satisfaction. Patients even talk about it. Even the providers have mentioned how they can maybe see one or two more patients, but more importantly, with the patients that they’re seeing, they’re much more engaged.

I think there’s ancillary benefit, since we’re at a revenue cycle conference, to some of the tools that we’ve seen in the back office focused on CDI. If those prompts are managed in real time, again, there’s less that they have to do in follow-up afterwards. There are fewer nudges that they’re getting, and it allows the providers to focus.

They might not even realize this technology is having an impact on them all the time, but even some of the back-office technology really allows our providers to, again, focus on what matters most.

How are you approaching vendor management and the question of platforms? Are you using different vendors for different needs, from the back office to the clinical setting?

I think it’s hard right now to standardize because who’s going to be the winner? I don’t think any of us know.

Obviously, Epic has a very large footprint within the institution. But for all of the tools surrounding Epic, even if you think about the open models that are out there, Claude, Anthropic, OpenAI, who’s going to even be the winner in that? We don’t know. The same is true with the localized vendors.

To me, when we’re doing this selection, we do have to keep in mind a little bit: What is the adaptability from this to another tool?

You mentioned UGM. Epic is coming out with tools all the time. I mentioned Hippocratic AI. Well, they’re building technology that’s very similar.

How do we make sure that technology keeps ahead with a large vendor? Obviously, if one of our large platforms does it and does it well, that’s who we’re going to go with first. There’s no need to add more and more technology.

It is hard to determine who that actual winner is going to be right now in any of these areas. Revenue cycle can be one of them.

How do you plan for that uncertainty in contract discussions or when you’re trying to strategize five or 10 years out?

For us, from the planning aspect of it, our focus is really heavily on the workflow and the impact that we’re trying to make, and not the technology first.

I think if you go at it with the technology-first mindset, you are likely to have a lot of starts and stops.

Our focus and our belief, to be honest with you, is a lot of this technology requires the change management before you implement the technology. There’s no reason to put automation tools on top of bad workflows. Automating bad things is not really what we want to do.

A lot of our focus is actually that adaptation first. Then the technology, to be honest with you, becomes a little bit more obvious. Still a little bit of a leap of faith, but it becomes a little bit more obvious because it’s about: How do you fit that adapted workflow that we’ve created and that adapted working model?

And do you have a shared vision, you being the vendor and us? Do you have a shared vision on what that end experience should be?

What do you think health care isn’t talking about enough when it comes to AI and emerging technology? What would you like to see become a bigger part of the conversation?

A lot of providers feel like they’re behind. I’ve heard this a lot. I’ve seen even some articles from your area on this as well. A lot of providers feel like they’re behind.

I don’t know why. We’re all in the same place.

Instead of feeling this need to play catch-up with one another, it would be an interesting discussion about: How do we share what good looks like? How do we help each other, inform each other?

We’re not a competitive industry. One health care institution is not competitive to others. We all have the same aim, which is bettering human life. To me, it’s about how do you share those insights and share what has worked and what hasn’t versus: Boy, I feel like I’m behind. I’m not doing enough. I need to move faster. I need to do this. I need to turn on this tool. This organization across town turned this on.

We’re all in the same place. Sharing the lessons learned along the way is another element.

I think the final piece is there’s something around the patient perspective that hasn’t been talked about enough. What’s their expectation for what this new world is going to be like?

I think it’s a lot about what providers think it’s going to be like. I have to be honest with you, sometimes, who cares what we think? It matters what the patient thinks. They’re the ones that are coming for the care.

I don’t know that I’ve seen enough on what’s actually the patient’s perspective of this whirlwind of technology and emerging data.

This is a preview of the September 24 edition of Access Health—Tap here to get this newsletter delivered straight to your inbox on Thursday mornings.

View 0 Comments

Request Reprint & Licensing

Submit Correction

View Editorial & AI Guidelines

Read More on Access Health Newsletter

Pulse Check

Executive perspectives on key industry issues