Episode Overview
LISTEN TO THE EPISODE:
Daniel Marino:
Welcome to value-based care insights. I am your host, Daniel Marino. You can't listen to the news, or watch anything on TV, or read anything these days without hearing about artificial intelligence and the impact of AI in all of our lives. And we're starting to see that right with using Chat GPT, and incorporating AI into our phones, and so forth. In healthcare, in a lot of the work that we all do in the healthcare community. There's a lot of conversation, a lot of excitement around where artificial intelligence is going. But there's a lot of myths that are out there. There's a lot of what I would call vaporware that is out there from technology firms, and it's really difficult to kind of guide through what is what is real and how will artificial intelligence affect what we're doing in our daily work. How will affect our patients? Versus those things that are just let's say vaporware, or things that are on the horizon, or things that we just need to get excited about. Not to mention a lot of the challenges, or some of the things that are confusing with artificial intelligence.
Well, I'm really excited today to have a colleague of mine, a good friend who I have known for many, many years. He is the chief information and data officer of an organization called Brand Engagement Network. Ben, for short. He's been there for about a year and a half. leads the whole artificial intelligence, growth and a lot of the data build and so forth for the organization. Prior to that, he was the chief data officer for about 4 years or so for ascension health. Rick Howard. Rick, welcome to the program.
Rick Howard:
Well, thank you, Dan. Glad to be here, and thank you for inviting me.
Daniel Marino:
So, Rick, let's dive into this. Give me your thoughts on where you see AI now, and how do you see it impacting healthcare? And in particular, how do you see it impacting our providers?
Rick Howard:
Well, thank you, Dan. I actually see AI in healthcare offering numerous applications that can significantly enhance both the efficiency of how doctors operate, and quite frankly, the outcomes and the satisfaction against patient care or improving patient care. Let me go into a couple of those, and maybe we can go into some more details later. First, AI serves as a valuable tool to alleviate the current staffing shortages in healthcare. We all know that clinical professionals are difficult to find out there. There's this extreme shortage of nurses. There's a shortage of doctors quite frankly, and Covid kind of helped that along.
Daniel Marino:
Sure.
Rick Howard:
Even though we had a pretty significant shortage before Covid, Covid actually accelerated some of that those challenges. But by automating some of the repetitive tasks and providing personalized patient engagement through digital AI. It enables the healthcare professionals to focus more on more on those complex challenges that they have with those patients.
Daniel Marino:
Well and to your point. It probably allows I would assume a lot of our healthcare resources right, our human resources become much more efficient, be able to do more with less. Do you see AI, in some cases taking the place of some of our providers?
Rick Howard:
I don't really see AI taking the place of providers. I see AI as being a significant tool to complement our providers. Healthcare is going to be personal, no matter which way you go.
Daniel Marino:
Right.
Rick Howard:
The you know, the transaction of a clinical professional and a patient coming together to address needs is always going to be there. I don't really see AI going through the FDA process to be diagnostic. I do see AI as being complementary.
Daniel Marino:
Yeah, kind of as a as a tool to our physicians. As to our providers. I, I would agree with that. What when you when you look at a lot of the vendors that are out there, many, probably every single digital analytics vendor or technology vendor, they have some level of artificial intelligence that they are incorporating into their products. Where do you see some of the biggest advancements of AI occurring in healthcare? Maybe, as it exists now.
Rick Howard:
Well, yeah, I do see a number of advancements there. And you're right. A lot of organizations like to use the new buzzword, which is AI, or I guess the buzz algorithm. In this case. That doesn't necessarily mean that they have learning algorithms which AI is based on the fact that the algorithms are always learning as you apply data to those algorithms, and they get much more specific against the data sets that you present to them.
Rick Howard:
So but I do see a lot of advancements. One of the areas, for example, is revenue cycle.
Daniel Marino:
Yeah.
Rick Howard:
If built correctly on the revenue cycle side, and that, I say, if but for on purpose. Those algorithms can possess the capability to learn and to adapt. And what I mean by that is, when you train those algorithms on best practices and a comprehensive data set, it has the ability to go retrieve the right data to support billing. What that then translates to is potentially minimizing or eliminating denied claims. But if claims do get denied expediting the claims adjudication process.
Daniel Marino:
Yeah, I agree with you. I think the impact of AI on revenue cycle could be quite large. There's so many, all of us right. And I'm a old revenue cycle guy. So run a billing shop. Denial management has been like the bane of my existence. Right? And if you have some type of technology or technology that can come in that helps you think about the challenges with the claim before it goes out, or as the claim is the denied claim is coming in help you to manage it quicker. Boy. You know time is money, right? That's where I think we're going to see a lot of value.
Rick Howard:
Yeah, I agree. And again, even as claims come back denied. If there's a sufficient reason for the denial, the algorithm can learn that reason. And if it learns it, then it can, then it can basically build its model to where it doesn't repeat the challenge on the next claim. So that's the beauty of AI, it's got that machine capability to do those activities and to take on those challenges that most humans won't have at a much more efficient rate than a human.
Daniel Marino:
Right kind of as a, you know, a regular learning process. How about value-based care? You know, there's a lot of, the whole premise around value-based care is to look at quality, to create efficiencies, right? To bring down the total cost of care, to create some proactive care modeling with how we take care of patients. From your perspective, how do you see AI really impacting, or maybe advancing value-based care.
Rick Howard:
I actually think value-based care is one of the immediate opportunities for AI for a number of reasons. But let me go into some of those. I think it's AI can support care management by providing a empathetic data driven, but yet human-like interaction with patients that enables a bidirectional exchange of information. A conversation, if you will, with that digital agent. And we can train our AI agents on very specific data sets like specific chronic condition data sets. Where it only responds to that chronic condition. Or if we want to do multiple cohorts of conditions, we can do that as well. The point being is, we can train a digital agent to have a conversation with a patient and have a very level set exchange of that information with the patient that's empathetic. I think the other component to this is did through that empathy that a digital agent can provide. It disarms the patient. Let me give you an example here. A digital agent is not going to be judgmental.
Daniel Marino:
Right.
Rick Howard:
It's always going to be empathetic. It's going. It's going to create a comfortable environment for that patient. And here's the real benefit to having a comfortable environment. Many patients are reluctant to be fully disclosed with respect to the things that they do that are conversation points for a care manager and that patient in managing those diseases. They don't want to say they've been having a high saturated fat, a high salt diet, because they're afraid of the result. If we can disarm the patient to have a more free exchange of information through the conversation, we get an important valuable data set from that conversation to help train that algorithm to understand how we manage that patient better on a go forward basis.
Daniel Marino:
Right. So what it sounds like to me is you can use the agent then, to look at what some of the past healthcare characteristics are of that patient, maybe incorporate it with what the care pathway should be, or the clinical protocol should be. But then have that outreach, and that conversation with the patient in such a way, where you're making the patient feel relaxed. Maybe you're asking more proactive type questions. You're using this really as a complementary tool or service, if you will, to kind of drive the right level of outcome.
Rick Howard:
Absolutely. And from my days as cheap data option. You know, this data is the lifeblood of any industry. If I've got more complete accurate information coming from the patient, so that I can add that to the very distinct and clinical information that I already have, you can imagine how important that data set combined is in trying to manage that patient which is the very premise behind value-based care.
Daniel Marino:
Well, and I'll tell you, care managers in a lot of the work that we've done. We develop a lot of programs for care managers and help care. Managers and organizations become more efficient, right? To do more with less. One of the biggest challenges that care managers have is aggregating the right clinical data, or all the clinical data from all the sources that are out there. And aggregating that data is the biggest let's say, the amount of time. And they need to do it right before they even have that conversation with the patient. If artificial intelligence can help to aggregate that and place it into some type of a format that allows for an efficient conversation. I think that's where you're not only going to drive a lot of clinical outcomes with a patient. But you're going to have tremendous amount of efficiencies with the care managers.
Rick Howard:
Well. And, Dan, you bring up an important topic right there. Clinical data, at least, the last study I have seen clinical data is doubling every 45 to 60 days.
Daniel Marino:
Wow! Isn't that amazing?
Rick Howard:
That is a lot of information to stay on top of as a human, whereas a machine can ingest that data, adjust its algorithms, adjust its models in seconds or minutes. And the and then be precise with the new data set. In addition to the historical data set that it was trained on so that it knows the sequence of the information to provide. So again, I think you're making an important point here is that there is just too much data out there, and it is doubling too frequently to stay in front of it all.
Daniel Marino:
If you're just tuning in, I am Daniel Marino, and you're listening to value-based care insights. I am here with Rick Howard. Rick is the Chief Information Data Officer of brand engagement network, Ben. Having a fascinating discussion on AI in healthcare. Before we continue our discussion, I do want to offer out to any of our listeners. At Lumina we have what we call our AI readiness assessment. A lot of organizations aren't really sure where they are, where they stand, and being able to incorporate AI into their organization, and in particular into the culture of their organization. We have a simple 6 question assessment. If anybody is interested to help you think about where your organization is, and some of the things that might be important as you integrate that within your organization. Anyone interested, please feel free to reach out at dmarino@luminahp.com. Rick, kind of getting back to our discussion here. One of the comments or the challenges I hear, maybe criticisms, I hear a lot of times is with the use of AI. It's how we manage the bias of AI, right? And as you, you know, your comment, I think, was really interesting as to the level of information that continues to double in size right every 30 to 45 days. How do we know that AI and the information coming out of AI is correct? How do we manage the bias?
Rick Howard:
Well, it's all in how you're training your AI. And quite frankly, if you're training the AI with A, you know with, without considering any of the bias appropriate associated information. Then you're training the AI on the specific data itself. Now, that brings up a different bias point. If we have data that has bias built in, for example, 30 year old clinic trials. As we both know, we have bias built in. There's not much AI organization can do because those clinical trials, if successful, are FDA approved and you have to follow the FDA guidelines associated with that with that particular set of information. You cannot stray from that. You can support updates to some of that that gets approved and gets sent out. But short of the bias built into the physical data. It's all in how you train the model. And if you're training the model, for example, specific to a chronic condition, you're training it on that chronic condition, regardless of any of the bias elements associated with the individual population, that that might be serving.
Daniel Marino:
Well, and I think you have to go back to what are those? What are those leading sources? Right? What are those recognized industry. You know. I don't know main sources that have been recognized by the community as being the preferred or the best. Let's say outcomes or data source, or what have you. Within, Ben, though maybe we can talk about it in terms of how you're solving for this. Within, Ben, as you're building some of your capabilities around artificial intelligence. How are you managing through the bias?
Rick Howard:
The one thing Ben is doing that a lot others are not doing is, for example, we have our own LLM. Everybody's heard about ChatGPT, or Copilot, which are really popular applications, but they depend on the open Internet to retrieve information. We have our own proprietary LLM. It is built off of a, you know, a llama 2 model or other open source type LLM Models. But the point I'm making here is the data that everything is trained on comes from our customer. I'll give you an example, Dan. We recently did a drug adherence model for a very popular type, 2 diabetes drug that everybody's talking about today. The only data set that is used in that model is the FDA approved data that was released against that particular drug. And we went to the National Institute of Health to pull down type 2 diabetes, data. A trusted source that everybody turns to that model can only speak to the information that it has been trained on which are those 2 particular sources. It cannot stray outside. If you ask the model a question outside of it, it will tell you it doesn't have that information. It's here to provide information on type 2 diabetes, or that drug and redirect you back in that area. So that's 1 of the reasons Ben has taken the approach it has, is we're trying to eliminate hallucinations or incorrect information, if you will. As well as what could be industry inappropriate results you can get from going to the open Internet.
Daniel Marino:
Wow, that's interesting. So really, the way that you're building this, it is just it's really almost restricted right? To that nationally recognized data sources. Just to ensure that the data that is sort of at the foundation or at the source of creating your intelligence is really based off that nationally recognized information.
Rick Howard:
That's correct. We call it ring fence. So we're basically fencing the data that the model is built on so that it doesn't stray outside of the data set. And, Dan, you know, I come from the provider world.
Daniel Marino:
Oh, yeah, sure.
Rick Howard:
My passion is supporting a patient or supporting a clinician that is supporting a patient. So I am very passionate about making sure that we are diligent in following this model of only using customer provided or trusted information to build our model so that we can eliminate all of the challenges that are associated with the open Internet. Tthat helps patient care, that helps physicians, that ensures that we have an honest conversation that is accurate with that patient.
Daniel Marino:
Do you think that the information that we're creating in AI, and let's say, the capabilities that AI can provide in healthcare. Do you think it could almost be too good? And and let me give you an example of this. I was at a I was at a presentation not too long ago, where a digital analytics AI vendor was talking about the ability of the solution to create advanced diagnoses. And this particular vendor was touting the accuracy of the diagnosis, and basically said he thought his solution did a better job of diagnosing a particular patient or a condition more accurately and quicker than the physicians. And of course the physicians in the room are all scratching their heads and saying, there's no way that this could happen. But it led me to think. I mean, do you feel like, maybe physicians or providers could maybe over rely on the technology? And if so, that that does create some challenges.
Rick Howard:
Yeah, I would agree. In fact, there was a study. It's been about a year, maybe a year and a half ago that UC San Diego did, where it compared the responses of the LLM with the responses of the providers, and quite frankly, the LLM won when we look at that data.
Daniel Marino:
Yeah, that's fascinating, isn't it?
Rick Howard:
It is. And what that tells me is that the providers are handicapped by the fact that the algorithm is being armed with the most recent information around the healthcare data that's out there that we just talked about is doubling so frequently, where the provider, unless they're spending an awful lot of time reviewing all of that material which I think would be nearly impossible, given their schedules. They don't have the benefit of being armed with that most recent data. So accuracy is a part is to my perspective, is a part of the data that you have and the understanding of that data, and how you are interpreting that data to make the right diagnosis.
Daniel Marino:
Well, and I and I feel like, it is so many AI has just so many opportunities and so many advanced capabilities that can support physicians and providers in the care that they're giving to patients. I personally don't feel like it'll ever take the place of it, and I think if physicians, as they, as AI advances, as it becomes more incorporated into the care, model and clinical decision making and so forth. I think the real value of AI is physicians using these capabilities as an advanced tool, right? One of many right one of many tools in their toolbox that they can use to get a more accurate diagnosis, maybe to create a more advanced treatment plan, to provide that level of advanced outcomes to patients, because at the end of the day. That's really what we're working towards.
Rick Howard:
Yeah, I would agree. In fact, we did a little test, if you want to call it that, with a leading university here in the US. And the head of the department gave us a textbook on high-risk pregnancies, and said, I want to test what you guys can do with this. So we ingested the textbook, we ingested about 8 or 10 clinically peer reviewed papers that brought that textbook to current, and we did all this in 4 hours, by the way. So we ingested the entire textbook. We released the model to him through the conversation that he was having with our digital agent. His response was, this textbook could pass the current tests.
Daniel Marino:
Wow!
Rick Howard:
It's only because the model had recent information. The raw model was able to ingest it quickly and interpret it quickly, to be able to have that conversation with that with that physician, to satisfy that him that we were able to ingest and help a provider with more recent information than what they could probably get unless they want to do a lot of reading. So.
Daniel Marino:
Wow! That's that is that's fascinating. So, Rick, when we look at AI right now in healthcare. You know, it's advancing quick, right? And I think there's 2 pieces of it. It's the AI advancement supporting the administrative side as well as AI supporting the clinical side. But there's a lot of what I would call vaporware that is out there right? Everybody's kind of touting that they have the next best thing in artificial intelligence. Any advice for our listeners or our provider community? How can they manage through what is considered vaporware? Maybe in development versus what's actually there, right? And how they can use the actual components of AI right now as a way of maybe dipping their toe in the water, so to speak, related to artificial intelligence. Any thoughts?
Rick Howard:
Yeah, I think there are a number of ways here, and this is some of the things we would have done at Ascension, you test. You test the solution. You create pilots. You put it in a situation where you're going to evaluate its performance against the pilot. And then you do a slow but measured rollout of the technology, assuming it passes your criteria during the pilot for that rollout. I think that's the only way healthcare providers or insurance organizations, or even pharmacies and pharmaceutical companies can get comfortable with the new technology engaging their patients is to pilot this.
Daniel Marino:
Yeah, you have to pilot it. Test it right?
Rick Howard:
Exactly test it. Make sure it's doing it. Test it, you know. Test it rigorously. But then I think what we have to do is, we have to be willing in healthcare to understand that we're not going to eliminate 100% of risk. You can't do it today without digital AI, or without any of these capabilities. So I think, then, adoption has to be open minded, and ready to move forward with those tools so that you can better serve your patient population that you're that you're reaching with these technology sets.
Daniel Marino:
Well, Rick, this has been a fascinating discussion. I'll tell you. This is something that I just, every time I read a little bit more about the impact of AI in healthcare and talk to folks such as yourself. I learn a lot. I just get really jazzed about what these opportunities are. Really want to thank you for coming on the program. If any of our listeners are interested in maybe connecting with you or learning a little bit more about this. Any information that you can share on how they can contact you?
Rick Howard:
Oh, absolutely. It's just Rick Howard at on LinkedIn, or you can reach me on my company email address, which is rick@beninc.ai.
Daniel Marino:
Great. Well, thanks Buddy, I really appreciate it. And oh, by the way, I didn't, I failed to recognize. I wanted to mention this as well. Congratulations on you being on the cover story or on the cover of CIO Review. That that's huge and just, you know, a great congrats to you.
Rick Howard:
Well, thank you, Dan, and thank you for having me on today. This has been a pleasure of mine to have this discussion.
Daniel Marino:
Well, I appreciate it, and would love to have you back again, and to our listeners. Thank you for tuning in. Really appreciate it, and until the next insight. I am Daniel Marino, bringing you 30 min of value to your day. Take care.