AI and machine learning offer great promise for healthcare. Patient diagnosis, care delivery, claims processing, clinical decision support, population health, and security–all can be significantly improved. The adoption journey has already started, but the top reported challenges include lack of business and physician buy-in, adoption of new processes, staffing, and ethical issues. What would it take to harness most value out of these disruptive technologies and change health and lives for the better?

Contributor

• • Ron Double, Chief Information Officer, Parkview Health

Download Podcast
Apple Podcast, Google Podcast, Spotify, Pandora, iHeartRadio, SoundCloud, TuneIn, and Stitcher. Find other syndication channels here or search CIO Talk Network podcast on any other app.
Explore More

• • How will AI Transform Patient Experience?
  • Healthcare in a Cloud?
  • Ensuring Digital and Virtual Healthcare Security
  • Should you continue to invest in AI and ML?
  • Better Pharma through Digital
  • EmblemHealth + ACPNY: A Healthcare Payvider (Payer + Provider) Patient Empowerment Journey
  • Individual Revolution in Healthcare: Beyond the Hype
  • Streamlining Healthcare Data Retention and Integration
  • Patient Centric Health Care Through IT!
  • IT Challenges of Hospitality Industry

Transcript

Sanjog Aul [00:00:24]:

Hello, and welcome to this segment on CTN. To learn more, please visit ciotalknetwork.com. Our topic for today is Adopting AI And Machine Learning In Health Care. Our guest for today’s show is Ron Double, who’s the CIO with Parkview Health. Hey, Ron. How are you?

Ron Double [00:00:41]:

I’m great. How are you today?

Sanjog Aul [00:00:43]:

Very good, sir. Very good. So the reason we wanted to talk about this topic is we have covered AI and machine learning on multiple shows in the past, but that was about whether it has value for different industries, health care. The short answer is yes; there is value, but the devil could be in the execution, and that’s where we have to see whether AI and machine learning, when we start looking at it from an adoption standpoint, how people will end up using it, how the value will get created. What has been the journey so far? What have been our adoption challenges so far, and what does it take for health care? How does it get ready to embrace and invite AI and machine learning into the enterprise? So that said, the first question, Ron, is we are looking at the potential that AI and machine learning offers, which is in the form of significant improvements it can bring about, but do you think based on the effort being made so far within the industry or perhaps in your organization, what do you see has been the state of adoption and the actual outcome of adopting AI and machine learning?

Ron Double [00:02:10]:

It’s a great question. As I thought this through and thought how this impacts us: AI and machine learning can be used throughout the complete care continuum. I think we have great opportunity in health care to leverage AI and machine learning. We’ve been maturing over the last several years and my philosophy here at Parkview has been that we should gather every bit of information we can possibly gather and hold it in our stores for the time when we really need it for AI. We have really started moving across the continuum from diagnostic analytics, which is where we really started, where you’re using this to do more of the diagnostic work, to now predictive analytics, and we’re really doing a lot with predictive analytics, moving on to prescriptive analytics, and AI and machine learning are along that continuum. So we’re slowly as an industry and as a health system moving towards that world of AI and machine learning. I think the first thing you have to do is walk before you run. We’re working our way through that and really taking some big steps. We’ve been using predictive analytics for a while here in predicting sepsis risk, risk of 30-day readmission, things like that. We’ve been using the information to begin to do some predictive analytics. I think the next step is how we use it to start being prescriptive in how we should respond and how we should act. I think that’s a little scarier area for many people in health care.

Sanjog Aul [00:03:56]:

So what you just mentioned, in terms of where you’re going with it—prescriptive and other things you may have envisioned—that if you bring AI, we’ll start happening, but this doesn’t, from an execution standpoint of inviting it in, adopting it, and starting out with this whole journey, happen overnight. It’s not that one day you make a decision, the second day you start deploying it, because there needs to be buy-in from the business, and they would have to give you some funding because there are bigger changes required before you slap on a tool.

Ron Double [00:04:38]:

Right. I think the thing we have to understand is what problem we’re trying to solve. The business comes up with the problem, and AI and machine learning are really just the way for us to solve the problem, to look for new patterns. As an organization and as an industry as a whole, we have a lot of problems to solve. The way you start getting adoption of AI is by allowing people to define some defined problems, and then you take that and you begin to look at how you can use AI to solve that problem. If you start getting a few big wins in solving those problems, people are much more willing to adopt the technology and want to use it and leverage it. That’s what we’ve seen from the beginnings of what I would call business intelligence, which is really that first step in the process of doing some predictive analytics. Once we gave caregivers one prediction, they liked it and they wanted more, and so there becomes that thirst for more of this information. So you’re right, we have to walk through the process over some time, and yes, it requires the organization to be willing to change and to supply us with the resources we need to produce this information. That’s where the domain expertise comes in that we will hopefully talk about somewhere in our conversation today.

Sanjog Aul [00:06:13]:

So I recently had a conversation with three health care leaders separately and also as a unit. We spoke about the need for getting a better handle on data. Health care, among other industries, has had a chronic issue with respect to data being harnessed not only from your internal organization, because health care cannot be delivered in isolation. The data is floating around among the different entities which come together to deliver health care. So when you talk about AI and machine learning, it cannot truly be just limited to what you generate within your four walls. It has to also include the ones generated by your partners, but when we talk about data integration and data sharing, health care has had a very long battle, which is still being fought. Do you think your adoption of AI would be complete before you have had better wins in the data battle that you’re fighting?

Ron Double [00:07:23]:

Yeah. From my perspective, the use of AI will never end. It’ll continue to grow and keep changing and modifying over years, but regarding data and what we need from data, the first thing we have to realize is that health care traditionally has been silos of information, and those silos have been held as competitive advantage. It used to be that the physician had a record on the patient, and that was kept in their office; it wasn’t the patient’s information. So the first thing we have to realize as an enterprise or industry is that the information we have really belongs to the patient, and we’ve got to break down those silos. Information is often seen as a competitive advantage, but I would argue the information is the patient’s and that shouldn’t be the competitive advantage. How we use the information to enable the organization and how we do things is the competitive advantage. When we start thinking about these silos, the first step is they have to be broken down. The second thing is interoperability. Although health care has made major strides in interoperability, we still have a long way to go. The reason is our protocols for interoperability are not interpreted and applied consistently across the enterprise and the industry. For example, HL7 transactions are a way we exchange transactions between systems, and yet organizations interpret how those transactions work differently. We really have to get more standardization in how we exchange information. We’re on the right path, but we’ve got a long way to go. Another big component is we don’t have consistency in the enterprise or industry about how and where things are documented and the process for doing that. Everybody does it a little differently. So it becomes really hard to normalize the data and make sure that the data lines up correctly. Although we’ve made progress in codifying a lot of data, there’s still a long way to go. A lot of the work in how we’re going to be able to do this is to get more standardization across the data, and it will be a challenge for us to move effectively through AI until we get that standardization in place.

Sanjog Aul [00:10:16]:

So given that you acknowledge and agree that there are data inconsistencies, and I understand that machine learning could be more of that. It’s not truly inside the scope of that data inconsistency related battles that you’re fighting, but AI is. Do you think you can start your journey by getting some insights from the AI effort, but they’ll be incomplete at best because you still don’t have an aggregated version of data? How could your insights be complete or comprehensive even if you generate them using AI?

Ron Double [00:11:04]:

Well, I think we have a vast amount of data already. One thing we have to really think about as we’re starting this journey is that a lot of the information we need is directional. It doesn’t need to be perfect. We are able to use the data we have to begin to identify patterns and get some direction to what is happening. Even if it’s not down to the specific degree, at least we’re making identification of patterns and able to use that to give us some direction. That’s more than we’ve been able to do in the past; it’s more than the human brain could do by itself. That’s where we’re getting some value already out of AI. By collecting data from all these other sources, if we can get socioeconomic data and data from payers and data from outside the organization, it improves the accuracy of our predictions. I think the direction of our predictions is probably pretty accurate. Each piece of data we start to get in or each new source of data just adds to the accuracy and will build over time to a better product.

Sanjog Aul [00:12:34]:

So that said, what we wanted to understand or confirm from you is: are the systems, the processes, the data ready? Would you give a green check or green signal to what you have going in your organization to say, yes, I’m ready to welcome AI? Because when we’re talking adoption, readiness is the first question to be asked.

Ron Double [00:13:03]:

That’s a really tough question, and I’ll try to answer in a couple ways. Are we technologically ready to begin this journey and make some major steps in that direction? Yes. The more data and the more information I can glean and the better I can start to do predictions and look across that data and find patterns, we’re ready. Every health system needs that because there are lots of patterns of problems we don’t even know exist, and AI will help us identify those so we can begin to define better problem statements and fix them. Culturally is the challenge. Culture is what is really the holdback on AI adoption. Organizations that are willing to make change, take a little risk, and test hypotheses will move forward. Those steeped in traditional ways of doing things and unwilling to adjust are going to struggle to adopt AI. We are a pretty change-adaptive organization, and we’re getting pressure from our executive leadership and our board to be a data-driven organization using AI to identify patterns and problems and then help us solve those problems. So I think we are as an organization ready. The health care industry as a whole is going to struggle because it is pretty steeped in tradition.

Sanjog Aul [00:14:45]:

So let’s look at the leadership within the larger health care organization and also at the physician community because they all have to work together for health care to happen. What is the level of buy-in that you’re able to get from both of these constituents to go and invest significantly? At least enough that you can start seeing some value. You can do experimentation and proofs of concept all day long, but something you can truly deploy and start seeing material value coming out requires buy-in, support, and funding. You have these two different sets of stakeholders. What’s the current state, and how are we tackling it? Please stay tuned, listeners. We’ll be right

Speaker 0 [00:15:49]:

back. Predict your company’s future by creating it. Is your workforce able to connect, exchange ideas, and share brilliance simply and securely? Create tomorrow, today. Empower your people to innovate anytime and anywhere with secured BlackBerry Enterprise Mobility Management and document sharing solutions. To learn more, visit blackberry.com/enterprise. Your growing business needs a highly productive workforce, effectively communicating and collaborating without exposing corporate data to cyber attacks. Are you looking to balance security and workforce productivity? Move beyond short term measures and securely scale your business with BlackBerry Enterprise Mobility Management Solutions. To learn more, please visit blackberry.com/enterprise.

Speaker 0 [00:16:49]:

Patient centered care requires a connected enterprise. Are you ready? If you’re looking to scale your health care IT efforts, visit redmane.com/health today. Whether it’s to connect data from multiple partner solutions or developing software for unique needs, RedMane can help. To find out how RedMane can help your company deliver on the patient centered care promise, visit redmane.com/health or call (773) 693-3919. Visit today.

Speaker 0 [00:17:26]:

You are listening to CTN, CIO Talk Network with Sanjog Aul. To learn more about our program, please visit ciotalknetwork.com. Now back to the show.

Sanjog Aul [00:17:38]:

Welcome, listeners. So, Ron, when we look at the buy-in and we know that we have entities, we’ve got internal stakeholders, which could be your health care executive management, and then you could have the physician community. Of course, there are many others, but perhaps these are the ones who play a key role in what will end up happening leveraging these newer disruptive technologies. So when you look at the buy-in, if you were to define the current state, where are we overall? It could be what’s happening in your company, your organization, or the industry overall.

Ron Double [00:18:12]:

Great. I’m very fortunate. I’m in an organization where we have buy-in from the executive leadership and from the board that this is our future, that we will use information and what we can glean from that information through artificial intelligence and machine learning. How we can leverage that technology is very important to our future. Given that, the leadership has committed not only financial resources but human resources to us. In fact, if you look at my IT budget as a whole, our biggest growth areas are two areas and one of those is our AI, BI/AI, our machine learning. It’s really moving us through that continuum on the intelligence side of the house. The other one would be information security, which we obviously have to continue to grow in this risk world. With that said, resources in our organization have been committed. From a provider perspective, there are so many problems you could solve and so many things out there in health care patterns that you can’t put them all on the providers at once. It’s just too much change. The providers are willing, if you can prove a few big wins and show them that using this technology helps improve outcomes and that patients are better served, they begin to adopt it. We’re doing some work right now in cardiology and we’re able to identify some variations in care and also identify what those outcomes look like. You can see that just because we’re doing more or the cost is higher does not mean the care is better. You start to identify where you can improve care, and once you show a few wins, people start to buy in and want more of those opportunities. These happen throughout the whole care continuum where we can identify variation in cost, variation in how treatment is provided, and outcomes differ depending on the method. If we identify those, adjust, and get people to follow a more standard protocol, it tends to help. Then we see providers wanting to engage and look for more opportunities to improve because they want to be the best they can possibly be, and now we have data to help them be the best. I think adoption is slower on the provider side, but once you demonstrate a couple big wins, they come along.

Sanjog Aul [00:21:19]:

So I’m sure you’ve heard this as well that AI and machine learning, I’d say more the AI, is positioned to replace or displace health care workers. That includes physicians. That could be because it is knowledge-based. If you could combine knowledge of 1,000,000 physicians and get that to give you a recommendation on what to do in a given situation, that could be very powerful. As we know, I don’t want to name the providers, but there are some providers who are betting their paycheck on it. So what do you say about the adoption-related pushback or passive or active resistance? That could put them out of a job. Is there job loss or business loss?

Ron Double [00:22:11]:

I find that an interesting question because as I think about it and as I talk to the providers I work closely with, what drives that idea is fear. An example: I read recently that if you’re going to become a radiologist, you might want to think about your future because machine learning is going to replace radiologists. I think there are cases where we’re beginning to see that AI and machine learning in some conditions can identify those as well as radiologists, but across the continuum, we’re a long way from being able to look at all aspects of human physiology, comorbidities, along with that diagnostic image, and really understand the patient. We’ve got a long way to go, so I really wouldn’t be worried about that right now. Talking to our providers, one of the things we should be looking at is how this technology can complement what they do. It can take the mundane work off of them and put them on more specialized work. That’s the message I try to get with our providers: this takes some of the mundane work away, and the more specialized work—the work you really like to do—is complementary because it gives you the clue as to where to look, and then you can investigate further. I struggle with the idea that we’re going to displace physicians, providers, radiologists in the foreseeable future.

Sanjog Aul [00:24:10]:

So now based on wherever we are, another area to look at is experimentation. Suppose the buying happens. While I spoke about proof of concept, would you go full-blown with AI and machine learning? You mentioned we will go slow, but do you have what I call a path to adoption or complete adoption? Do you have some low-hanging fruits that you recommend? Or are you going through areas where, say, this, this, and this area are the ones we will experiment with and then eventually deploy first? Depending on the outcome, we will go to the second area and the third area. Is there a playbook in short that someone should have?

Ron Double [00:24:59]:

I think there is a playbook and it goes back to the idea that there’s so much we could be solving. As an organization, we’re looking for clinical variation and cost to see if… We’ve chosen some DRGs and are using them to identify certain conditions or disease states we want to look at and then we’re looking at the variation across those. We’re starting small. For example, we do cardiology, we look at a few DRGs, we see the variation, we see patterns, and we try to identify which problem to attack. Then we develop a problem statement and we use AI to solve that problem. If we’re successful, it breeds another problem or another question and we keep going and it starts to expand slowly over time by answering one question which generates another. As an organization, we’re lined up across service lines and we’ve decided to take the approach at the service line level and look for clinical care and cost variation at a service line and a few DRGs, and do that concurrently with another service line. Begin to look for some wins. Our roadmap is to start with a few things in each service line and then expand over time. As we start to fix problems, identify patterns and fix problems, we’ll keep moving through the continuum and it will grow over time. I wouldn’t call it experimental; I would say take a few steps in the right direction and expand over time.

Sanjog Aul [00:27:10]:

So in terms of the policies and the processes, we’ll take a break before we get to this question. I wanted to at least share it with the audience. The policies and procedures we have—and we know that AI is not just peripheral—cannot be put peripherally. To get the most value, digital processes that bring insights have to be injected back into processes. There are ethical issues and other policy issues which will kick in because you’re using AI, and errors, especially in health care where lives are at stake, will impact policies. As part of that playbook, would there be specific candidate policies and procedures which one should start looking at to see to the level of adoption that we are kicking off with? What changes do we make and how do we continue to tweak them until you’re done with full-blown adoption? Let’s discuss that when we come back. Please stay tuned listeners.

Speaker 0 [00:28:29]:

Predict your company’s future by creating it. Is your workforce able to connect, exchange ideas, and share brilliance simply and securely? Create tomorrow, today. Empower your people to innovate anytime and anywhere with secured BlackBerry Enterprise Mobility Management and document sharing solutions. To learn more, visit blackberry.com/enterprise. Your growing business needs a highly productive workforce, effectively communicating and collaborating without exposing corporate data to cyber attacks. Are you looking to balance security and workforce productivity? Move beyond short term measures and securely scale your business with BlackBerry Enterprise Mobility Management Solutions. To learn more, please visit blackberry.com/enterprise.

Speaker 0 [00:29:28]:

Patient centered care requires a connected enterprise. Are you ready? If you’re looking to scale your health care IT efforts, visit redmane.com/health today. Whether it’s to connect data from multiple partner solutions or developing software for unique needs, RedMane can help. To find out how RedMane can help your company deliver on the patient centered care promise, visit redmane.com/health or call (773) 693-3919. Visit today.

Speaker 0 [00:30:05]:

You are listening to CTN, CIO Talk Network with Sanjog Aul. To learn more about our program, please visit ciotalknetwork.com. Now back to the show.

Sanjog Aul [00:30:19]:

Welcome back. So, Ron, when we talk about the policies and processes which we will need to change in a health organization to effectively adopt AI and machine learning, what’s the playbook chapter going to look like for that? To what degree would we start shifting things around because we are embracing AI and machine learning into the healthcare process?

Ron Double [00:30:53]:

When you think about policies and processes, our challenge with policy is that traditionally health care has had silos of information and those silos are basically unto themselves. They’re blocked and sharing and exchanging information has been challenged. So the first thing when we think about policies as an organization or industry is we really have to reevaluate our privacy policies. I’m an advocate for privacy, but at the same time there’s a need to share information effectively between entities in order to identify where we can improve. We have to look at privacy policies and how they need to be adjusted. Data use policies are another challenge. What payers, providers, and health care institutions can use and how they can use that information can sometimes be a challenge to getting good information across the continuum. That’s another policy that needs to be looked at. Then there’s the whole ethical component we need to talk about. As you start to prepare for AI and the use of AI, go back and revisit those main policies: privacy and data use, and think about all the different data you’re going to be using, all the different sources of that data, and review and try to normalize those policies across the continuum. As we have stood up a clinical integration network, which requires a lot of data and a lot of data use, that’s been one of our biggest challenges—making sure we can provide the right data in a secure manner and protect the privacy of the patient in that work. Aligning those policies first is the first step in the playbook and then you’ll start adjusting the process of how you consume data and what you can do with that data. I’m not sure if I answered your question, but that’s how I see it.

Sanjog Aul [00:33:38]:

Sure. Understood. So when you are looking at the receptivity of this whole idea of bringing AI into the organization, I’m sure the original business case was cost efficiencies and automation wherever possible. People have also looked at introducing it on the clinical side. Would you say we have not gone full-blown yet, and maybe there is time before physicians will be replaced by an algorithm? Or are we setting the foundation as part of adoption of these two to eventually go there? Or are we going to primarily be making this an efficiency play?

Ron Double [00:34:25]:

I would argue it’s both sides of the house. We are using AI to identify cost and efficiency gains; that’s easier to move, but we are also looking at it from a clinical care perspective. I’m not going to share specifics, but we have done some work and I called our clinical variation model where we are beginning to look across clinical care, looking at outcomes and what was done to get us to that outcome and identify if there are certain protocols we could follow that would improve outcomes overall. We’re beginning that work of looking at patient outcomes and what was done along their journey and where we see the best outcomes, and trying to use this information to adjust clinical practice. It’s happening on both sides of the house: clinical practice is getting adjusted, as well as administrative overhead components.

Sanjog Aul [00:35:39]:

If you were to look at the collaboration that we need among the different parties, as I mentioned, health care is all about working with different collaborative parties to make it happen. When we look at AI, yes, data could be one central foundation everyone will need to support, but what else? What else is required if you were to draw out the requirements for what’s needed among all the different value chain partners in health care? What else is needed from the rest of the entities who are working together to make healthcare happen?

Ron Double [00:36:23]:

I think socioeconomic factors and what’s happening in our communities—we need that information to complete the story and we don’t necessarily have all of it. We’re looking at payers to provide information on claims because we have a view from a provider perspective into the work done within our organization, and we also have a view into things that come to us through Health Information Exchange, but payers probably have a better view of everything across the patient: the complete story of care or the complete story of charges or claims. They give us clues into other things done outside our organization. As we look at this, it’s really a collaboration between providers, payers, health information exchanges, laboratories—it’s a collaboration between all those entities to get the complete story and complete information around the patient.

Sanjog Aul [00:37:47]:

So we just spoke about the payers and the providers, the insurance companies and the health care. Would you think physicians are more at the receiving end for the most part, or are they expected to step up? If they are to step up, the question is true even for the payers: what’s their incentive to do so? If they have not done so far, the last thing on their mind is to support the service provider toying with new technology. Why would they invest their time and energy to change unless they get value out of it?

Ron Double [00:38:25]:

I’m not sure I understand the question, but I’ll try.

Sanjog Aul [00:38:29]:

No. I’ll repeat it for you. What I’m saying is what’s the incentive? One is what they should be doing for adoption, right, to support the AI, but then what is the incentive for them to change? There should be a reason why they would invest their time and energy to change anything that they have unless they get value out of it. People will not just make this change for altruistic reasons.

Ron Double [00:38:55]:

Right. If we look at it from the provider side first, then the payer side. From a provider side, we are seeing that payers are beginning to incentivize providers to provide information to improve quality of care. The idea is we’ll give the provider a little incentive if they provide this information so we can look across the continuum and use that data. At the same time, by providing us with information, payers provide the provider with information we can use to begin to create—we’re doing a lot with our clinical integration network where we’re looking across the population and addressing gaps in care and addressing risks because now we have information that completes the story and we know more about that patient. By doing that, we’re improving quality, and if we can improve quality and lower the cost for the payer, it’s a win-win. So the incentive is to incentivize providers by offering them incentives for providing the data and improving quality, and at the same time payers are getting incentivized by us by providing us with data that we can use to improve quality of care which in turn lowers their cost. If we share data back and forth, quality improves and cost is lowered, which incentivizes both sides.

Sanjog Aul [00:40:46]:

What would you say about benchmarks? What’s the benchmark or the blueprint that we are trying to develop or the holy grail? What are we pursuing that, if we get to this level of adoption or collaboration among partners, then we have reached where we wanted to with respect to AI and machine learning adoption? What are we chasing? If you don’t have a picture, how do you develop goals and build a roadmap?

Ron Double [00:41:19]:

That’s an interesting question. I’m not sure there is an ever-satisfied mode of where we’re trying to get to. I think it’s constantly evolving and improving. Until we can get to the point where we feel we have the complete story, we don’t have all the data. We’re constantly striving to gain access to more information around the patient to give them the complete story of care. I also think there’s a patient component because we want to engage patients in what their ideal state is—what they define as the care they should be getting based on what we’re doing with AI. What role do they play?

Sanjog Aul [00:42:29]:

Now, if you are looking at the different ways—we’ll take a quick break now—what we should be talking about is the people, because people will make or break any shift or transformation. While AI and machine learning are not transformational technologies by themselves, they could become the basis of how you deliver health care both clinically and administratively. Are we ready yet? Or are we looking at some cool cloud technology which will allow us to take care of this and we can do what we are doing today? Please stay tuned, listeners. We’ll be right back and explore.

Speaker 0 [00:43:22]:

Predict your company’s future by creating it. Is your workforce able to connect, exchange ideas, and share brilliance simply and securely? Create tomorrow, today. Empower your people to innovate anytime and anywhere with secured BlackBerry Enterprise Mobility Management and Document Sharing Solutions. To learn more, visit blackberry.com/enterprise. Your growing business needs a highly productive work force, effectively communicating and collaborating without exposing corporate data to cyber attacks. Are you looking to balance security and workforce productivity? Move beyond short term measures and securely scale your business with BlackBerry Enterprise Mobility Management Solutions. To learn more, please visit blackberry.com/enterprise.

Speaker 0 [00:44:21]:

Patient centered care requires a connected enterprise. Are you ready? If you’re looking to scale your health care IT efforts, visit redmane.com/health today. Whether it’s to connect data from multiple partner solutions or developing software for unique needs, RedMane can help. To find out how RedMane can help your company deliver on the patient centered care promise, visit redmane.com/health or call (773) 693-3919. Visit today.

Speaker 0 [00:44:58]:

You are listening to CTN, CIO Talk Network with Sanjog Aul. To learn more about our program, please visit ciotalknetwork.com. Now back to the show.

Sanjog Aul [00:45:12]:

Welcome back, listeners. So, Ron, let’s talk people. People are critical and are the very basis of whatever is going to happen. AI and machine learning on the surface look like technologies you can subscribe to by plugging into a utility model or cloud and get that data churned, but is that enough or is there some machinery required? Include people from the technology side who report to your department, but also the people who will end up using it to deliver care.

Ron Double [00:45:48]:

That’s a great question. Let’s start with the patient. The patient is key to our success in what we do with AI. Everything we have about the patient is the patient’s data, it’s not ours, and we should share openly with the patient and let them help us decide what we can use in this process. That’s a whole show in itself, talking about patient engagement in their care. From a provider’s perspective, you have to have collaboration with the providers. The key is looking at this as complementing their work, not displacing it, and how do we get them to use this to complement what they’re already doing? Medicine is both an art and a science; there’s the factual science and there’s the things AI will identify and define some prescriptive approaches. Part of the art of health care is understanding the patient, understanding compliance, and understanding financial or social impacts and addressing that in an artistic way beyond pure science. So it’s a mix of art and science and this is the complementary science part of care. Then think about resources to manage all this data. We struggle, especially in a rural Midwest area, to recruit talent that understands how to manage, manipulate, and utilize this data. We’ve had to take the approach of having a great team of data analysts and data architects, but the domains are so diverse between specialties that it becomes impossible for them to be good in all areas. So we’ve moved to creating tools that are self-service, and we guide people to leverage these tools and look for their own patterns and problems and then identify how to address them. Recruiting higher-level data scientists is next to impossible in the rural Midwest, so you define a problem and find the best domain expert in the industry and contract them to solve it and then move to the next problem. It’s a combination of people involved in this process and each has their own role. We try to work across the continuum and understand each perspective and how to engage them.

Sanjog Aul [00:49:15]:

So when you said that the people problem is daunting, would you say there is a solution in sight or will you just promise to work hard to solve it?

Ron Double [00:49:28]:

A solution would be too simple. It’s a continual evolution, continually adjusting how we engage people as more information and better tools become available. A few years ago, we would have solved this by adding hundreds of IT resources to generate reports for people when they asked. Now we look at it as a self-service approach and we’ve changed our approach: move to self-service, get domain experts in departments to lead the charge, and we’re there to support them. It’s a change and it will continue to evolve as tools, ease of access, and accuracy of information evolve. It changes what kind of resources you need and will be a continual evolution.

Sanjog Aul [00:50:35]:

One last question about leadership. Who should be leading this effort? Who should own it? On the surface it might look like a technology adoption, but it’s a business problem. Who as an individual or which group should own it? What type of leadership would you need for success? It is unlike other technology adoption; this stands to disrupt how we imagine and deliver health care.

Ron Double [00:51:17]:

I don’t think of this as a disruptive technology; I think of it as a solution to organizational needs. Leading the aggregation of data and getting tools in place can be led from an IT or IS perspective, but the actual use has to be led by operational leaders of service lines or entities. In our case, domain leaders should drive what they need, what problems they need to solve, and use domain experts to identify problems using the data. Leadership is outside the technology itself.

Sanjog Aul [00:52:12]:

On behalf of the show and our listeners, thanks so much, Ron, for taking the time to share how you and your organization are attempting to tackle AI and machine learning adoption challenges, and how health care organizations can come together with value chain partners and different parts of society to make sure this is adoptable and makes a marked improvement in how health care is imagined and delivered. Thanks so much.

Ron Double [00:52:42]:

Thank you for inviting me.

Sanjog Aul [00:52:45]:

Thank you. Listeners, hope you enjoyed and learned a few nuggets here. Like us on Facebook. Search for CT and CIO Talk Network, and be sure to follow us on Twitter. Join our LinkedIn community, and please go ahead and rate us and also download our podcast here on iHeart, Spotify, iTunes, and Stitcher. You name it. We are on a lot of different networks. Please go ahead and enjoy. Thank you again for listening to the segment on CTN. This is Sanjog. Your talk show host till next week. Take care, and God bless.