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

Health care executive Dave Wessinger discusses their article “How to adopt AI in health care responsibly.” Dave explains why organizations must move beyond the hype of artificial intelligence to focus on measurable goals like driving growth and improving quality. The conversation highlights specific high-value applications in documentation and patient intake while emphasizing the critical need for a clinician in the loop approach to ensure patient safety. By prioritizing administrative automation over clinical decision-making initially, leaders can build trust and avoid the risks associated with imperfect technology. This episode explores how to shape AI as a practical tool that enhances rather than replaces human judgment.

Partner with me on the KevinMD platform. With over three million monthly readers and half a million social media followers, I give you direct access to the doctors and patients who matter most. Whether you need a sponsored article, email campaign, video interview, or a spot right here on the podcast, I offer the trusted space your brand deserves to be heard. Let’s work together to tell your story.

PARTNER WITH KEVINMD → https://kevinmd.com/influencer

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

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

Transcript

Kevin Pho: Hi, and welcome to the show. Subscribe at KevinMD.com/podcast. Today we welcome Dave Wessinger. He is a health care executive. Today’s KevinMD article is “How to adopt AI in health care responsibly.” Dave, welcome to the show.

Dave Wessinger: Thanks for having me.

Kevin Pho: All right, let’s start by briefly sharing your story. Then we will jump right into your KevinMD article.

Dave Wessinger: OK. I will give you the short story for sure. I started almost 30 years ago now when I was just a young man with hair. I was an IT guy working in a nursing home of all things. I recognized that there is an opportunity to improve care delivery and provide clinicians with solutions that were better than what was currently available in the on-premise world, if people remember that.

It was delivering software over the internet at the time, which we now call cloud. That was a big differentiator at the time. We happened to leverage the technology, and we were the first ones that came to market with that technology. We have been doing that work ever since. We work closely with customers delivering really effective system of record type solutions in the senior care space.

Kevin Pho: And of course, AI, which we are going to talk about today, is really rocking the health care markets. So many solutions and startups are out there. For those who did not get a chance to read your article, tell us what it is about.

Dave Wessinger: There are a few things that I was trying to share with the community. This is the way we think because we have been through some things. We have some experiences. We do not all have the opportunity to connect on those, so why don’t we share some of the things to save other people some time?

I think the one thing that resonates in that article for those who have not read it is that there is a lot more to AI than the acronym. As you think about responsible AI, people do not really give that enough time, but there are many different ways to do it. You have to be really thoughtful. If you are not doing it yourself, you need to ask: Where is the data going? What does the access look like? What is security? How do you train the model? What is the efficacy?

All of those things will have an impact on your ability to be successful with it. I just wanted to share experiences and let people know that it is a journey. It takes time. Do not chase AI. Chase the value if that is a thing, and do not get caught up in it. It has to land in real material value. Do not be scared of it. Lean into it. If you feel like you are behind, you are like everybody else.

Kevin Pho: Tell us the type of specific AI-centric products that we are talking about now and that you are most familiar with.

Dave Wessinger: Where I think AI really lands after you spend some time versus in a lab is where we can truly augment the clinician in the loop and be helpful. We all recognize that in this industry, the clinical and administrative burden is very heavy. We want to make it a great place to work again.

As I think about where you have opportunities, I will take one specifically in a product area that I am excited about. For those close to health care, you recognize litigation is a big issue. Litigation often is related to how well you document. Almost oftentimes, it does not matter what you actually did. It is how well you documented what you did, and do you have a trail to defend yourself? It is a sad situation, but that is the reality.

If you use AI to move through charts, look for opportunities where either there are gaps in documentation or things missing near real-time to help the clinician more effectively respond to what might have been a fall but was not captured that way. They did not inform the physician, did not get the rails up, and did not communicate with the family. You actually deliver really good care. At the same time, you are protecting yourself from potential litigation, which makes your business more sustainable.

AI is perfect for that. That is what we have currently released to market in our business. It is well received in that it is very difficult to go through every chart for all of your patients or residents and understand what is happening and make sure you have closed off those gaps in real time to deliver great quality care. It is something that was just really hard to do. That is a perfect target for AI.

Kevin Pho: Now you are talking specifically about the nursing home space, which of course you are familiar with. Give us a scenario before and after. What is that documentation like now or before AI? Give us an example of how AI can really fill in those gaps just to illustrate your point.

Dave Wessinger: There are a couple of ones. I might even jump to another one regarding resident status. You probably recognize that in situations there are times where you are with a patient and as you reconnect with them, being a physician, a CNA, or an RN, you want to understand what is the current status. If you want to do that across the shift you happen to be on, that involves combing through a lot of information or shift notes and can be challenging.

AI can surface that very quickly and give you the quick “dot-jot” format of what has happened so I can quickly get up to speed with that individual in a moment that otherwise would have taken hours. You are way more informed to provide the right level of care and the right interaction with an individual given the information that you did not have to scour for that is appropriate and available in the moment.

We leverage that same capability for the reason for transfer. Oftentimes when somebody shows up in the ER, it is very difficult to figure out why they are there. What we think is helpful is a 250-page CCD, but the reality is you want to turn that into actionable information that somebody can ingest. They need to know how to treat, manage, and work with the individual. That is where AI has landed really well.

Kevin Pho: So you are saying that AI can take a 250-page document of a patient’s chart, synthesize it, and summarize it. So whenever they have that transfer of care in the emergency department, for instance, it will just surface the most actionable essential information.

Dave Wessinger: Exactly right.

Kevin Pho: How accurate do you think AI is currently in its iteration in terms of doing things that we are talking about?

Dave Wessinger: That is a really good question. I think that is the trick in getting it right. What we have seen as an example is the efficacy in models where we are using natural language processing of clinical data versus standardized documentation that is codified. There is a significant difference in our ability to demonstrate or share information that is helpful versus information that might hallucinate or might not be accurate.

What we have been pushing for is to make it as easy as possible at documentation. The more structure we have around that, the more effectively they are going to benefit from the actual intelligence that comes out of the AI model. So what goes in really matters, and the poorer it goes in, the less effective the models are.

Most models have an opportunity to land well. I think the challenge, quite frankly, is that in health care you cannot just be close enough. One medication error is not acceptable. The bar is much higher in health care.

I would say there are a number of use cases that will land, but it does take time for them to mature. There are some things that the model cannot do without getting better information to it.

Kevin Pho: So we are speaking in mid-December. This episode probably won’t be published until mid-January. So with the current models, which I know iterate and evolve so quickly, if I dump 200 pages in and whatever comes out, how confident should I be today in terms of what AI can synthesize from that 200-page chart?

Dave Wessinger: It is a really good question. I would tell you as a technician that you should be extremely confident in what comes in relative to what you could do as a human. I think you would be very confident. I think what is interesting though is we have to spend some time and iterate through it and build trust ourselves as clinicians.

I say that because we have seen situations where we have been able to demonstrate a model that is 99.9 percent effective in terms of what the outcome would be to the point where it requires clinical verification. It is so common and they trust it so much that they often do not even look anymore. In fact, they are causing more errors than they did before because it is so fast and so accurate. But that 0.5 percent is a big problem.

It is a very tricky question to talk about. I think we are going to get most of the way there. I think we are going to operate as well or better than humans in the stuff that they just cannot do as well, which is ingesting information and summarizing it. But there are going to be some things lost along the way as we create some of those efficiencies. It is going to take a little bit of time to get perfectly accurate.

I think that if you chase perfect, you are probably not going to get there. Month over month and year over year, the models will get smarter. They will learn from how humans interpret that, and they will get better. If your model is learning from human behavior, you will be better off than with a static model.

Kevin Pho: Now as AI infiltrates workflows in pretty much every area, we always talk about the new definition of the human in the loop. So in the scenarios that you are mentioning, where is the physician, clinician, or human in the loop as this whole process iterates as AI evolves?

Dave Wessinger: It is very clear that as you make decisions, you can surface information that they then need to improve and take action. It is not a medical device. It is not there to tell you what to do. It is not there to treat patients. It is there to provide information in a faster way than a human can ingest so that you can do more with less.

I take a great example regarding intake. You have patients coming to you and you want to understand if you can be effective with that patient. Generally speaking, if they have the payer mix and they really understand their financial profile and their clinical capability, they can make a decision fairly quickly. Oftentimes that is human-based and takes time. There are avoidable days that get piled up in the system cost.

Ninety-five percent of that is automated. “Yes, I can take the patient,” or “No, I can’t.” Somewhere in the middle is where the human is in the loop automating a process that is largely outside of a clinical process, which I think is easier than creating that automation in a clinical process. You do need a clinician in the loop; that is just not optional today.

I actually think the clinician, interestingly enough, is AI-proof in terms of the future. The administrative folks think that is going to move a little faster because there is less risk. You are starting to see that automation.

So the human in the loop is in a gray zone. Can we take this patient? Do we have enough information to know our contracts well enough to say we will get paid for a bariatric bed? Or will we get some of these exceptions and not find ourselves in the red on this patient? Or will we be able to manage, deliver the proper care, and do well as a business?

That is where you have a human deciding between the gray. More and more, it could be 10 percent today, and over time you will see it go to five and then two and then one. What happens is, and what I am really excited about is, all of these folks that AI is hitting will move to higher value work.

The last person I talked to that loves entering data into a system? Well, I cannot remember when the last time that was, quite frankly. That would be the first time. Getting to a different level is just better for all of us. We have seen that through history.

Kevin Pho: Now for those two scenarios that you mentioned, decreasing litigation because AI fills in administrative gaps and perhaps synthesizing information for transfer of care, do we have outcomes data in terms of really reducing litigation or improving patient outcomes, or are we too early in the process?

Dave Wessinger: I would love to tell you I have a great answer for you on that. It is crystal clear that we see significant advantages. We have identified that in many cases, 90 percent of the incidents are not documented effectively. Simply surfacing that information, taking an approach, resolving those, and working towards them will invariably have a meaningful impact.

But that impact won’t be felt for years because litigation takes a long time. Two years from now, I will come back on your podcast and talk about the actual impact. But we all see the value there in terms of the impact they can have to their business and quite frankly the patient outcomes, which is why we are all doing this.

It will take a little time for it to hit, but the expectation is they should be able to meaningfully reduce that risk to their business. When I think about AI, Kevin, I really think about whether you can affect the bottom line. Typically in health care for the last 25 years, we really haven’t. We have done a lot of work as systems of records to digitize information, but we haven’t really materially changed how we work with it.

AI is going to meaningfully change that over time. When you think about those workflows or the opportunities, we now have an ability to say we can actually do better at the FTEs we have in certain areas. You can put them into different jobs. No one likes to talk about reducing the FTEs. Certainly on the clinical side we have staffing shortages, but we want to be more effective.

I think there is opportunity on the efficiency side, on the compliance and quality side, and even on our revenue side. It is not about anything more than understanding the care you are providing and making sure that you are coding it appropriately to get paid what you deserve. That is kind of it. I think there are opportunities in all of those areas that we are trending towards with AI in terms of where can we be helpful, largely on the administrative side and working clinically carefully, thoughtfully, and responsibly.

Kevin Pho: So let’s zoom out a little bit. For those listeners listening to you on this podcast and they are inundated with all these AI-related solutions, tell us the type of questions they should be asking to determine whether it is right for them and perhaps anticipate some of the potential risks that these tools may bring.

Dave Wessinger: I will use a great analogy as a thought-provoking analogy here. I was buying a putter the other day and there was a putter on the shelf that was an “AI putter.” I thought, “Interesting. Are you really going to figure out how to make me putt better on this physical thing that doesn’t change?” I think you have to be careful of the marketing aspect of AI.

I will move forward and say the things I would look for are: What are you delivering to my company in terms of value? Every board, every organization, and everybody I talk to is all about AI first. It is more about moving on that and moving it fast. I would be opposed to that. What is the value you are going to bring to the business?

Once you get past the value side of it, what does it really mean? Is it affordable? And then the next part is the due diligence. Talk to me about your company. Where does the data go? What is your pledge to responsible AI? How do you think about building it? Where is your infrastructure? Where does my data go?

Another analogy is that we all have Gmail accounts and we know what it is like when that leaks out. The next thing you know, you have an ad for the thing you just sent an email about. That is happening. Data is king. As that gets into potentially the wrong hands, you give out accounts and let them come in. What they have access to, what they do, and where that data goes is really important. It could do more harm than good.

You really have to look under the hood. There are lots of ways to evaluate AI, but you need to get educated on not just the value it brings, but how is it delivering that value? Is there anything under the hood that you should be concerned about? You kind of need to be educated there and get some help. There are lots of vendors coming to the table with some solution. It is really important to just understand their infrastructure. Is it in the country?

The little things like that seem small, but who they use as a partner and where it goes all really matter. It is a little bit of everything, but be curious. Focus on the value, not AI. That is interesting, but that is not what you are really doing it for. It is a technology that allows us to deliver more value.

I would say this: I believe in the next two years, we as general software people will deliver more real value to customers, especially in health care, than we have in the last 25 years.

Kevin Pho: We are talking to Dave Wessinger. He is a health care executive. Today’s KevinMD article is “How to adopt AI in health care responsibly.” Dave, let’s end with some take-home messages that you want to leave with the KevinMD audience.

Dave Wessinger: I think, be curious. That is number one. Do not be afraid. It is easy to kind of jump into it or kind of hold off, but I think you should engage and learn. You do not have a lot of time to evaluate 50 products. We are all really busy. That is clear.

Word of mouth is really helpful. Sometimes going slower is faster. The ones in the bleeding edge bleed. The ones that are on the next frontier tend to do better. You have to get up to speed on your diligence, look for where it is actually working, ask the right questions, and be OK to try some things.

It is not about a full commitment. If there is a way for you to learn and try a pilot, it is not a terrible idea. Sometimes those are kind of hard. But if it is going to deliver that kind of value, you could probably find it fairly quickly. If it does, then that is the kind of thing you can lean into a little bit more.

So just do not be overwhelmed by it. Take the time to learn. No questions are dumb questions. We are all learning. We are all on that journey. In a month from now, you look back and think that question is kind of silly, but it is just the pace at which we are moving. There is no pattern. There is no booklet. Those are coming, but we are all learning at the same time.

I would say just be curious. Look for the opportunities and specifically look for the areas that will have the biggest impact on your business. We have tried 70 AI vendors in our business, and probably two are delivering value. It is a lot of money out the door. You do not have that opportunity. I would say pick the things that will add value, validate it, spend some time on it, and do not be afraid of it.

Kevin Pho: Dave, thank you so much for sharing your perspective and insight. Thanks again for coming on the show.

Dave Wessinger: Thank you.