Subscribe to The Podcast by KevinMD. Watch on YouTube. Catch up on old episodes!

We sit down with Carlo Mahfouz, a technology executive and author, to explore the transformative potential of artificial intelligence (AI) in health care communication. Carlo delves into AI’s historical roots, its recent acceleration in adoption, and the unique challenges facing its integration into health care. We discuss how AI, particularly large language models like ChatGPT, can humanize the interface between humans and machines, revolutionizing health care practices.

Carlo Mahfouz is a technology executive and author of Reality Check: In Pursuit of the Right Questions.

He discusses the KevinMD article, “How AI can save lives by simplifying communication.”

Our presenting sponsor is Nuance, a Microsoft company.

Together, Microsoft and Nuance are leveraging their rich digital technology and advanced AI capabilities to tackle some of health care’s biggest challenges. AI-driven technology promises to revolutionize patient and provider experiences with clinical documentation that writes itself.

The Nuance Dragon Ambient eXperience, or DAX for short, is a voice-enabled solution that automatically captures patient encounters securely and accurately at the point of care. DAX Copilot combines proven conversational and ambient AI with the most advanced generative AI in a mobile application that integrates directly with your existing workflows.

Physicians who use DAX have reported a 50 percent decrease in documentation time and a 70 percent reduction in feelings of burnout, and 85 percent of patients say their physician is more personable and conversational.

Discover AI-powered clinical documentation that writes itself. Visit https://nuance.com/daxinaction to see a 12-minute DAX Copilot demo.

VISIT SPONSOR → https://nuance.com/daxinaction

SUBSCRIBE TO THE PODCAST → https://kevinmd.com/podcast

RECOMMENDED BY KEVINMD → https://kevinmd.com/recommended

GET CME FOR THIS EPISODE → https://earnc.me/T4NmIq

Powered by CMEfy.

Transcript

Kevin Pho: Hi, and welcome to the show. Subscribe at KevinMD.com/podcast, and get CME for this episode by clicking on the CME link in the show notes. Today we welcome back Carlo Mahfouz. He’s a technology executive and the author of the book Reality Check: In Pursuit of the Right Questions. Today we’re going to talk about how AI can save lives by simplifying communication. Carlo, welcome back to the show.

Carlo Mahfouz: Thank you so much, excited to be here, Kevin.

Kevin Pho: So Carlo’s been on before. Go to KevinMD.com/podcast to search his name and see his prior episode and story. Today we’ll jump right into the most recent KevinMD article, “How AI can save lives by simplifying communication.” So Carlo, for those who didn’t get a chance to read this article, tell us what this one is about.

Carlo Mahfouz: Absolutely. In this one I’m honing right in on the trendiest topic of today, which is artificial intelligence. And this time I’m taking you through a different lens on what artificial intelligence can do, and how it can lend itself to being a useful technology, especially for health care professionals.

Specifically, I’m not talking about intelligence. I think there has been too much hype about what it could replace, the avenues around knowledge, especially diagnosis, all of those things. I think those might happen, and they are important. But what I would like to talk about today is what it will offer as an interface, as a way of facilitating the communication layer. I think that is a very different lens from what has been discussed so far.

Kevin Pho: All right, so I’m interested in hearing that perspective from your lens as a technology executive. And you’re right, we have a lot of physicians talking about how AI can replace some of the tasks that they’re doing. So I’m interested in hearing your view of that intersection between AI and health care.

Carlo Mahfouz: Awesome. So in this instance specifically, we were talking about the interface. Typically, in a lot of the tasks that happen in a hospital, or in education and health care, there are too many layers of communication that end up happening, and too many stakeholders in that ecosystem as a whole.

One of the key advantages of artificial intelligence, and specifically the new large language models that have been introduced, especially with the prevalence of OpenAI, has been this way of accessing data much more seamlessly. In this context you no longer just ask a question and get a search result, or ask a question and then need multiple steps to get to where you need to be. The learning model, for the most part, synthesizes most of the data and gives you an answer. Of course there are refinements needed, and it is not necessarily 100 percent accurate, but there is value in that.

Because what we can now do is this. There are a lot of tasks that are mostly translation tasks, where all that happens is data being relayed from one stakeholder to another. That could be removed entirely. It could be abstracted out and eliminated as a friction point. So you maintain the integrity of the data, or of the communication layer in between, while at the same time removing it from the process. That simplifies the process and reduces the bottlenecks that exist in that ecosystem. Because the more touch points we have, the more the data or the communication flows from A to B and B to C, the more points of failure you introduce.

Kevin Pho: So give us an example, a real life example or scenario, where you see AI facilitating that interface between patient and the health care system.

Carlo Mahfouz: Sure, the simplest one. I recently went in for a checkup visit, and one of the first things the physician did was bring a laptop with him. Before he actually engaged in the conversation, he turned it on simply to transcribe the visit and produce a summary of that interaction.

At a very basic level, that is one of the first steps. Relaying that information previously would have required someone to document it next to you, or to take notes and then provide them to you. All of that has now been packaged into: hit record, get the transcription and the summary, and you are able to digest it after the fact.

Another good example would be when we are diagnosing someone, when we are seeing a new patient. The amount of knowledge necessary for you to come prepared is substantial, and it could benefit from signaling which areas might have deficiencies, or which areas might be overlooked. I am not saying that artificial intelligence will collect all of the data and tell you what to do. Instead, it can aggregate based on the history of the patient and the history of your knowledge and education, and say, there are these things that might be interesting for you. It is a nudge, presented in a way that is very easy to digest. You need perhaps two or three minutes, you can go over it, and then you are much better prepared for that type of engagement.

So these are small changes, nothing instrumentally huge, but I would say they can improve the overall experience.

Kevin Pho: So there have been so many technological advances when it comes to health care, and a lot of them come and go. Very few of them actually change the way we practice and change the way patients interact with the health care system. But I’m feeling that AI is a little bit different. So I’m interested in hearing your perspective as a technology executive. Tell me, is AI different from prior technological advances, and if so, how is it different?

Carlo Mahfouz: Great question, and I am going to come back to the interface, because I think it is a really meaningful one.

In my industry we work in simulation, and I can tell you that one of the biggest technology advances was when the mannequin for simulation was introduced. That innovation had no friction for people to use, because what they ended up using was something familiar, whether that was presenting CPR or something similar.

I think that is the challenge with a lot of technology today, especially if we talk about electronic health records or anything of that sort. When your interface is a touchscreen and a keyboard and a set of menus and multiple layers to get to where you need to be, you are introducing a whole new way of working, which is already taking your attention away from what you need to do, which is basically save someone’s life.

AI is different, because it uses what we already know very well, which is language. These language models are so powerful that they can talk, even if they hallucinate sometimes, and I think we can iron that out at some point. But they communicate in a much more seamless manner, and as a consequence the interaction with the technology becomes more seamless. It becomes easier. Instead of capturing information by manual entry or by some other means, you just have to speak it.

That is the beauty of the transcription example. The exchange itself did not have to change. It sits off to the side, not disrupting the current workflow of the physician and the patient, and it actually supports them. That is the value. It is not an added layer in the chain.

I think that is a really major difference in the technology, because what we have introduced before, over and over again, has been another layer, another thing you need to do on top of everything else you already had to do. Now we do not have to do that. It could just be recording a video, interpreting the video, checking things and making sure that this is the right dose of medication that has been administered. Or it could be simply the transcription, so that the next physician coming in can easily read it and assess it.

I received the electronic health records recently, and I had comments on what was happening for an incident. I can guarantee you, if that had been captured and I had received a synthesis of it, and I do this for my business meetings all the time, I would get a much better understanding of what is happening. It does not get it all right, just to be clear, I am not saying it is magic, but it is still a much better improvement.

And that is the difference. It is no longer a new thing we are adding. It is something sitting off to the side, supporting, which is as seamless as possible, and there is great value in that.

Kevin Pho: Do you feel like we’re moving too fast? I think things like ChatGPT, it’s about a year and a half old maybe, and it’s come such a long way from when it first came out. Do you feel like we’re getting too exuberant about the applications of AI, especially as it relates to health care? Are we getting too far ahead of ourselves?

Carlo Mahfouz: Honestly, I feel we are going slowly, and I will tell you why. ChatGPT was introduced with all its hype, but a lot of the technology has been around for at least eight years or more. What it did was create a new access format that more people were able to approach, and of course more data has been ingested, so the outcome was much cleaner and much more accessible.

I think that for health care to adopt it, it actually needs to leap even further before adoption takes place, because the level of error that is acceptable is very low. You cannot have a high rate of error. One of the examples I put in the article is RAG, retrieval augmented generation, which means that when ChatGPT answers you, it answers grounded in a knowledge base. That is very critical, because I do not want it creating information out of thin air.

So if anything, I think it needs to move faster for adoption to really start materializing in a good way. But again, the changes that we can make today are small and incremental, and they are not necessarily revolutionary. How can we optimize some of these small pieces in the chain that do not require much and where the risk is minimal? That is the key part, where the risk is minimal, where the risk is not too high. Because the higher the risk, the lower the adoption. There is a relationship between the two.

Kevin Pho: Do you see any downsides or risks about how far AI is going to be assimilated within our health care system?

Carlo Mahfouz: Downsides, I think there are a lot of downsides. Again, I am very optimistic about it, but I am not saying there will be none.

In any technology transition, what ends up happening is that we are in between, and you have malicious actors taking advantage of that transition and of the lack of understanding that comes with it, because the technology is still new. I hope that keeping the technology open for more people to access will alleviate some of that, because we will see innovation and security improve much more quickly. I think that is a model that will support it.

There is also the risk that it is going to make an error and that we are going to shut everything down because it made an error. And it is going to make an error, I can guarantee you that. We see this with automated cars. We have seen it for a while now with automated cars, and if you want we can draw a parallel. One accident in a self-driving car and the reaction is, let’s shut everything down, this is not useful. But in reality, if you look at it, the accidents caused by people actually driving, and the craziness of that, are much higher as a percentage. There is a cultural shift involved, a shift in understanding and accepting that, and that is going to take time.

Of course there are also identity issues, and AI taking on some personalization. And there is always going to be the gap of knowledge, where someone knows more about how to leverage that tool and is able to abuse it. But the reality is that this is always a risk, and it is a matter of us being more ready and prepared for it, rather than just saying it is not going to happen.

Kevin Pho: So I think that illustrates some of the different mindsets, right, in the technology field. Sometimes in technology you’re able to release things in version 1.0 and use some of those mistakes and errors in future versions to help perfect that tool, whereas in health care, if you make an error, people get harmed or people can die. So I think that is just an interesting juxtaposition between those two ethos, about how fast technology or technologists want to go, but how sometimes health care by its very nature can prevent that. What do you think about that?

Carlo Mahfouz: Again, the risk is high because we are talking about lives here, and that automatically changes the dynamic completely. That is why I said initially, do not leverage it for things that are highly critical. But you still need to learn and understand the technology’s capabilities, and that can be done in areas where the risk is lower, or where there is not as high a risk, whether that is on the education side, or in training, or in simulation.

You need to get to know the technology before you can actually use it and make sure that it does not harm people, and for that to happen it does not need to happen in the high-risk space. So, where are the areas that are less risky? Where are the areas where errors could happen and could be contained? Because we still need to learn the technology, we still need to learn what it is capable of.

I would put the difference this way. Between someone who is completely unexposed to artificial intelligence and what it can do, and someone who is exposed, the only difference is exposure. But do it in the areas where it is less risky. Even if it is in things you do just to improve your own efficiency on administrative work, which has nothing to do with the continuum of care or the quality of care, nothing along those lines, start there. And that is fine, because as soon as you start using it there, you begin to see some of the value. Because now you understand it, and physicians, who are the ones inside that ecosphere, can start helping the technologist understand how best to adopt it.

We need that communication layer, because it is not an either or, and of course it is not simple. But there are ways to optimize for getting started, and to understand where and how and when, more than anything else. That is what could move the needle forward. And in my vision it definitely will move the needle forward. I do hope that we start now, rather than saying, let’s wait until it’s ready. It is never going to be ready. Realistically there is no point at which we are going to say it is ready. It is more about how we familiarize ourselves with it, and the mindset changes, and that does not need to happen only in the context of health care.

Kevin Pho: We’re talking to Carlo Mahfouz. He’s a technology executive and the author of the book Reality Check: In Pursuit of the Right Questions. Today we’re talking about how AI can save lives by simplifying communication. Carlo, we’ll end with some of your take-home messages to the KevinMD audience.

Carlo Mahfouz: Probably the take-home message for me is this. In health care we look to standardize so much. We look to create a lot of homogenization, to reduce risk and create a risk averse environment.

I do believe that artificial intelligence, in streamlining these connections, will allow us not just to leverage standardization but to leverage integration. And there is power in integration, where we leave room for the innovation that is necessary to help save lives, while at the same time streamlining the connectivity, the connectedness, and the communication layers, whether from one stakeholder to another or from one system to another. I think there is so much power in what we can accomplish there.

Kevin Pho: Carlo, thanks again for sharing your perspective and insight, and thanks again for coming back on the show.

Carlo Mahfouz: Thank you so much.