Healthcare industry is expected to enjoy high-value business benefits from robot-assisted surgeries, virtual nursing assistants, and preliminary diagnosis enabled by AI i.e. Artificial Intelligence. What does it mean for the patient?

Contributors

• • Marc Probst, Chief Information Officer, Intermountain Healthcare
  • David Chou, Chief Information & Digital Officer, Children’s Mercy Hospital
  • Daniel J. Barchi, Chief Information Officer, NewYork-Presbyterian

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

• • Maintaining Health IT Toolset for Patient Centered Care
  • Vision Enabled MedTech for Improved Patient Care
  • Patient Experience Beyond Portals
  • How Social Media Monitoring Can Improve Patient Care
  • Patient Centric Health Care Through IT!
  • Healthcare Meets Hospitality
  • Bringing Smarts into Patient Care
  • Rethinking Health IT for Patient Centered Care
  • A Journey to a Better Patient Care Lifecycle: A Case Study in Mobility

Transcript 

Sanjog Aul [00:00:28]:

Hello, and welcome to this segment on CTN. To learn more, please visit ciotalknetwork.com. And the topic for today is How will AI Transform Patient Experience?   And our guests for today’s show are Mark Probst, who is the Chief Information Officer with Intermountain Healthcare. Hi, Mark. How are you?

Mark Probst [00:00:50]:

I’m very well. Thank you so much. Glad to be here.

Sanjog Aul [00:00:53]:

Great. Oh, glad to have you too. And David Chou, who’s the Chief Information & Digital Officer with Children’s Mercy Hospital. Hi, David. How’s life?

David Chou [00:01:02]:

Life is good. Thanks for having me.

Sanjog Aul [00:01:05]:

Oh, great to have you. And finally, Daniel J. Barchi, who is the Chief Information Officer with New York Presbyterian. Hi, Daniel. How are you doing?

Daniel J. Barchi [00:01:14]:

I’m great, Sanjog Aul. Thanks for having me on today.

Sanjog Aul [00:01:16]:

Great. I could not have been luckier to have these top executives from health care organizations who are willing to talk and give their ideas and thoughts and experience about how patient experience has been tackled so far and what can AI do. That said, Mark, I will start with you as a first question. We have always tried to be progressive in health care. Yes. Their technology has been seen in some cases adopted a little late, but for the most part, we are becoming very progressive and we are trying to do innovation. And we have always been working on the business side, in some cases on the clinical side. But then sometimes we see patient takes a back seat, the very entity who we are trying to support and or help solve problems for. So what exactly do you think are we after when we talk in context of patient referring to AI?

Mark Probst [00:02:13]:

Well, it’s a good question. If you think about innovation, health care is innovation. We talk about the practice of health care because we haven’t. There’s a lot of science, but we’re continually improving how we care for our patients and learning. And that’s kind of where AI comes into the discussion. Because we’ve learned through traditional methods, because that’s what we’ve had. We’ve learned through data analytics, which probably for the last 15, 20 years, we’ve focused on things like electronic data warehouses and how we use analytics to improve it. AI is a really exciting next step as we look at how we can learn and how we can improve our care processes, and therefore the patient experience. Best experience, right, is to get cured or taken care of. And AI is really a tremendous opportunity to take brand new data sets that may not be as coded as we’re used to and structured as we’ve been historically used to, maybe more in text blobs and those types of things, but through natural language processing and the great power of artificial intelligence, getting to conclusions and learnings more quickly. And we’ve had some success with that. One area we’ve focused on has been around our care process models. Intermountain historically has achieved pretty good results through guidelines very guided care these care process models. And typically, these would take a year, 18 months to develop a good care process model, which is essentially these are the best steps to take and if you run into such and such a circumstance along the care path, it gives you the structured approach again on how to continue to provide the very best care. These would take months and months to develop with lots of different clinicians. We actually worked with one AI vendor, brought them in, and we ran the same set of data, plus some augmented data that we had, through the AI engine to say, okay, how are we doing, and can we improve these care process models? And it developed exactly the same care process model for 1 modality of care. It did it in 3 days, and it found about 5 variables that really improved the outcomes of this care process. So really strong power through the analytics and the learning that we can do to improve our patients’ experience. We have other examples, but there are other folks that need to talk here. 

Sanjog Aul [00:04:43]:

So, David, if you were to look at AI as a new phenomenon or a disruptive technology, and, yes, as Mark said, use that to churn data better. You can use to get some insights, but we are only as good as the implementation or the action we take on them and the intent we have behind it. So do you think organizations, when they are looking at this whole improvement, is it happening more to improve the top line or savings on the bottom line, or is someone looking at the patients and saying, let’s go that route?

David Chou [00:05:23]:

So I like what Mark mentioned about learning. Right? AI is a tool that is constantly learning, relearning, and the goal is to be more efficient. So when we look at AI in health care, there’s always a lot of different areas that can be improved and be more efficient, whether that is improving the patient experience or it could help the organization make better decisions or it could even help drive savings. So I think all three play in a factor in terms of where AI can be extremely beneficial. Most important, just think about patient experience. Someone walking into the lobby of your hospital by using AI and some sort of facial recognition, you could identify this patient or even a family member. And then taking that information to create a better experience or a personalized experience that’s tailored towards them. Overall, I think that could be just 1 simple use case, but then we can also utilize AI to make intelligent decisions that can automate how we have a procedure or even help a physician make a better choice when it comes to selecting an implant that may be more cost effective, therefore reducing our expenses and improving the bottom line. So there’s a lot of ways to utilize AI, from an efficiency standpoint and lots of different use cases. So I would say AI plays a factor in every area of an organization, and they could solve a lot of various use cases that could help an organization become more efficient. 

Sanjog Aul [00:06:58]:

So, Daniel, when we are talking about AI in this context and we are saying, okay, we’ll do something with the patients, do you think the way they have expressed their desire to be treated better than what we are treating them today? Is it about being cool? Is it about being as if they were going to a hotel, somebody really treated them with that special attention? Or are they saying cut to the chase, give me a good healing, and let it be as, you know, smooth as possible, but stay close to what I came here for? What’s the ask?

Daniel J. Barchi [00:07:38]:

That’s a great question. I think that patients come to us because we’re a beacon of hope in their time of crisis, and it’s incumbent upon us to focus first and foremost on making them well. And, as health care providers, it’s our responsibility to take the best tools that we can find and use them to help make our patients better. And so it really isn’t about cool. It certainly should be as pleasant an experience as it can be for the patients. But at the end of the day, they’re looking for us to leverage the skills and talents of our physicians and nurses with whatever tools and analytics we can put in their hands to make them better. You know, health care technology, I’ll actually quote my friend Mark Probst here. Health care technology is really 80% people, about 15% process, and it’s really only about 5% technology. Artificial intelligence is getting a lot of focus these days and for good reason. It’s a potential game changer for what we do. But at the end of the day, it’s only part of that 5% that really enables us to have our people and our processes work better for the outcome of patients.

Sanjog Aul [00:08:51]:

So, Mark, coming back since Daniel just quoted you, if I were to come back to you and say that here, when we are trying to go towards these people and improve the patient experience using AI, but at the same time, we are also hearing, which is kinda unnerving for the medical practitioners themselves, that they would be rendered redundant because of AI, because some people, suppose they interpret X-ray. In fact, I have a friend who does that. And if AI is gonna come and say, okay. We’re gonna use it for a better job, but here are a few people going. Do you think in order to really apply it, you will need many, many health care entities to work together and some people feel that I’m gonna either lose business, I’m gonna lose my job. Do you think we’ll get to see the outcome the way we expect?

Mark Probst [00:09:41]:

Well, yeah, of course we’re gonna get to the outcome because of the benefits that the technology provides. Will people be redundant? And you brought up radiology. I absolutely believe that what we’re seeing with AI right now in the space of imaging is going to necessitate fewer radiologists in our industry in the long run. But we’re years from that. And frankly, we can’t find enough of them anyway. So I think it’s complementing a problem that exists today. It will do it faster. It will do it better. Do we ever fully replace humans in the care setting? I don’t think so because it’s about care, and I haven’t seen a tremendous amount of empathy from a computer. But a lot of the science underneath are care processes that can be automated, and yeah, it will change the dynamic of many of the people that work today. But think about drivers and autonomous cars. I mean, is that going to replace taxis? It is, but it’s going to shift the need for drivers. It’s going to shift the need for the people that are doing the work today to focus on different and maybe more important things. So I think it’s very complimentary. Is it threatening? Yes, it’s going to be threatening. But I think for the kind of people we want providing care and leading in health care, they’ll ultimately get excited about it. They’ll adopt it and they’ll help teach these systems to do the things that we need the systems to do. So it’s going to be a delicate balance. This is Daniel. It’s going to be better off.

Daniel J. Barchi [00:11:24]:

Sorry, Mark. I wanted to extend what Mark was saying there. Artificial intelligence is gonna help us do what we do, and there will be some jobs that are no longer necessary, but I think we should look at it not as a job eliminator, but as a job enhancer. Vinod Khosla is one of the founders of Sun Microsystems, that now is a private equity firm. He’s deeply interested in artificial intelligence, and he thinks that artificial intelligence won’t take away 80% of physicians’ jobs, but it will take away about 80% of the physician work. So the kind of things that we make physicians do today, asking for family medical history, looking up lab results, looking up an old lab result and charting between the lab result, old lab result and the new lab result to see whether the patient is getting better or worse. That’s 40, 50, 60% of what physicians do. And if we can have artificial intelligence do that kind of work as Mark described, then we can allow them to spend more time focused on the care of the patient and outcomes.

Mark Probst [00:12:31]:

Well, I think Daniel’s right on. I’m sorry to interrupt. This is Mark again. But we haven’t cured an awful lot of disease. There’s a lot of problems in healthcare that can be focused on, and if we can take these talented individuals and take away much of the things that Daniel was outlining and let them focus on new things and new ways to take care of people, wow. That’ll be incredibly powerful.

Sanjog Aul [00:12:56]:

So, David, when we look at AI as a technology, and it is, are we looking at it as a point solution, or are we thinking to use it pervasively starting from the patient’s bed to the boardroom? The reason I say that is because we did a show a couple of months back about how AI is tackling different areas. And the guest mentioned that even as a board member, a financial services company has invited an AI-based algorithm as a board member against humans. So that means we are touching almost every area. So do you think something like that could be happening in health care? So we’re talking patient experience here, of course, but at every place within health care, we could inject AI and reap the benefits. Is that where we are going, or is this just Star Trek that I’m talking about?

David Chou [00:13:54]:

No. I think, to echo what Daniel and Mark had mentioned, it’s a complement to being to a person’s job. So, yes, it’s human nature to be fearful. It’s almost the same analogy as when we talk about moving to a cloud-first model in IT. All of the engineers are extremely nervous at the start because it means a shift in how they operate now. They don’t have to rack and stack or touch a server. For them, that’s what they worked on their entire life. Now we’re moving to a cloud. They feel all of this is getting taken away. But now it’s AI, voice recognition. We’re doing so much of it right now on a personal level. It could be enhanced from a patient’s perspective. Why have a nurse call center when you may be able to have something like Amazon Echo in every room that a patient can speak to and get some intelligence out of it and make something make an action item occur. So it definitely needs to be pervasive and it will be. If you just look at how we live as consumers individually, let’s look at how much something like Siri has impacted your life now versus before. So definitely it will be pervasive throughout the various departments and throughout the entire industry. It’s just how fast can we adapt, how fast can we adapt to it, and are people willing to make that change? Technology can go as fast as people are willing to adopt it. 

Sanjog Aul [00:15:24]:

So, Daniel, would you say in order for us to invite AI into an organization like a health care investor, like yours or anyone else on this panel here today, would you say we will have to first at least do some cleanup or rethinking of the very underlying workflows and processes because we are not just saying, okay, churn my disk data faster, and that’s why it’s such a tool. But if you are trying to use it strategically, you got to go out there and kinda start not ripping it apart, or perhaps you do a revamp or a tweak. Where are we going to introduce risk? Because here, you have a very smooth or well-oiled machine that you’re running, hopefully, at each of the health care organizations where patients’ lives are at stake. And here you bring a new technology and say, I’m gonna try to do things differently, how would you ever get it adopted, whether you go step by step or go transformative? In both cases, we’re introducing risk.

Daniel J. Barchi [00:16:25]:

So first of all, I appreciate your very kind interpretation of the modern US health care system. And I think Mark and David would agree with me that while we’re very proud of the care that we delivered, nobody is characterizing what happens today as a smoothly oiled machine, and we’ve all strived to do better from the patient experience point of view. And, you know, if you look at the national standards, if you think back to the To Err Is Human report, there are many, many patients who have adverse outcomes and die nationally annually because of medical errors. So I wouldn’t characterize the modern US healthcare system as free of error and fully, smoothly running. So that actually gives us more opportunity to embrace technology and not feel like we have to wholly rip and replace. I think by layering new technologies, including artificial intelligence and machine learning onto what we do today, we can actually improve the systems that we have. And I can think of a couple of examples. We’re already using some AI algorithms to do a couple of things to predict readmission scores. So when we’re discharging a patient, we want to know if it’s likely they are going to have to be readmitted in the next 30 days for one reason or another. And if they are, it’s necessary to give them more follow-up care, perhaps more education, perhaps home visits to make sure that they don’t slip back after their discharge. So we use machine learning and intelligence to inform us as clinicians if that might happen. Similarly, when a patient’s admitted, we want to get a pneumonia predictor and know whether it’s likely that a patient is going to contract pneumonia and whether we need to admit them to an ICU or we can treat them in an ambulatory way and send them home. So we’re already using some of these tools and they’re simply layered on to what we do. I think the real frontier is when artificial intelligence is helping us on a day to day basis in patient care. A number of institutions nationally including ours are already starting to do this. So we’re working with artificial intelligence that’s thinking about what’s happening in our health system, looking at the millions of transactions that we’ve had over the past 3 or 4 years, and then starting to guide us in a way that is really helpful. So for instance, we all nationally focus on length of stay. We really don’t want to keep patients in the hospital longer than necessary. But unless everything really goes smoothly, it’s really hard to discharge a patient. So if it’s likely that a patient might be discharged, so we get an order for a lab done by 11, and we get the result done before 2 so we can determine that it is safe to send her home. If it’s 11:15 and that lab hasn’t been ordered and the physician who had ordered is really busy with other patients, it would be great if the artificial intelligence system wakes up and sends a nudge, essentially a message through one of our phones that all of our physicians carry and says, hey, Mrs. Smith has not had that test. It’s 11:15. Would you like to go ahead and order it? And have the physician say, oh, yeah. I wasn’t gonna get around to that until noon. Let’s go ahead and order it, and perhaps we can get her out today. It’s not replacing anybody’s job, but it’s small nudges like that where the intelligence system can help us do better what we do today.

Sanjog Aul [00:19:52]:

Very interesting. So, David, would you see any specific areas where you are thinking beyond just making these people productive or maybe give them an aid in terms of their decision making where AI can really show value? Because all of these examples that we said, they look incremental, and the question will come up is this: who’s gonna pay for it and why? If it doesn’t show us really something earth-shattering, why should you introduce this?

David Chou [00:20:21]:

Right. So, you know, patient experience is not just tied in it’s also tied to there’s some revenue impact as well. If you’re not doing well on your patient experience and the engagement piece of it, you may have a negative impact on your bottom line. So that’s how I would approach it in terms of why we need to provide the best experience. There’s both financial impact, but also from the patient’s perspective. So definitely, there’s a lot of different ways to look at this and make AI part of your operations. I think currently, it’s still relatively in the infancy stage. We’re still trying to figure out how do we roll AI in various fashions for any company. So we even think outside of patient experience. You know, people, I would love to think about or I’m starting to look at AI from a just a pure IP security practice. Can all this information predict whether we have exposures to ransomware, the viruses. So there’s a lot of different areas where AI is there to help us be more efficient. And the big impact is the ability to predict an outcome before and allows us to be more proactive, whether that’s patient experience, whether that is IT operation or health care operations. That is where AI is really going to impact us powerfully.

Sanjog Aul [00:21:44]:

So, Mark, if you’re going to look at any kind of rip or replace with respect to AI, which areas would those be where you will say, we are not ready to even invite AI into this specific segment or this specific area of our operations?

Mark Probst [00:22:02]:

Well, I don’t know if I’m gonna specifically answer that question, but I’ll get around it. I don’t see the relationship between rip and replace and AI. There may be opportunities, new tools that come out with AI that we want to move into and leverage. But I think as we’ve heard, there are a lot of augmentations that’s going to happen through AI. Let me give you two examples. One is more machine learning with AI, but what we’ve been trying to do for the last probably 18 months is teach a computer to see and to look at handwashing. So handwashing in our intensive care units has been a challenge forever, and we’ve all been dealing with all these funky devices to do it. We’ve been trying to teach cameras to watch our clinicians, and as they come in the room, do they wash their hands? After they’ve touched a catheter, do they wash their hands? After they’ve touched the patient, do they wash their hands? And we have been successful in teaching the computer to see through AI. Now this is through a partnership with one of our academic areas. But this has been pretty exciting. This isn’t rip and replace, this is brand new capability. It’s something that can be and really does have an impact on patients’ wellness and experience that’s been pretty darn exciting as we look at it at Intermountain Healthcare and some of the things that we’re doing. So I think that’s one really good example. It’s not rip and replace, but it’s brand new capability that the organization can leverage.

Sanjog Aul [00:23:36]:

Daniel, what do you think about from a patient’s and their family’s standpoint? Do you think they are ready to be diagnosed by a computer and or someone conducts surgery, which is a robot versus a human being? Are we there as a society yet? And if you are tackling or trying to tackle this, so eventually you want to retain the patient as well. How are you making the case, and who’s doing the convincing?

Daniel J. Barchi [00:24:07]:

So a fair question. I’ll approach it incrementally. I’ll start by saying today, most health systems in the United States are already using computer-based tools and medical records that when prescribing happens, compare the patient’s medical history, allergies, and past medical history to the drugs that they’ve been prescribed looking for drug-drug or drug-allergy interactions. And if you ask the patient, are you comfortable that a computer is checking your allergy and potential reactions against the formulary of perhaps 20,000 drugs? They would say, oh, yes. You know, it’s unreasonable a pharmacist or a doctor would know every one of those 20,000 drugs and know whether it’s gonna react with my allergies. So I think patients are comfortable with computers helping guide decisions. I’ll take a step closer to artificial intelligence and think about remote health care. We’ve gone live with the telemedicine program at New York Presbyterian we call NYP On Demand where patients who come into our emergency department, if they’re not in critical condition, if they’re quite frankly an ESI 4 or 5, they have a rash on their hand, they stubbed their toe, they have a head cold. Instead of putting them through the regular process that might take 2 to 2.5 hours to see a physician, we sit them in a small room and have a video visit with a physician. Now the physician is only about 300 yards away on a video camera, but it’s a video interaction instead of a hands-on interaction. So they’re essentially having a virtual ED visit right there in the ED. They’re in and out in about 30 minutes. We have a net promoter score of about 96%. Patients love it. Certainly, it’s quicker, but they also like the idea that they’re getting one-on-one focus time with physicians. Now that physician’s on a video screen and then they’re not touching the body of the patient, but it’s a very good experience and patients like it. The reason why I bring that up is I think that we’re going to get to a point where the physician becomes more and more remote. And at some point, it might be a virtual physician. And so I think that we’re going to get there in steps. I don’t think we’re ready for it today, and I think Mark and David will agree with me that artificial intelligence itself is not ready for it today. But I think patients are gonna get comfortable with it more and more over time, especially for non life-threatening conditions.

Sanjog Aul [00:26:28]:

So while we are all excited about it, talking on this show about how AI will change patient experience, how is the business of health care thinking? So, Mark, if you were to look from their standpoint, are they saying, oh, yeah, whatever. There’s a new tool or a new cool way of churning out data. See what you can do with it. Or they’re saying, this is gonna change how we will even operate or exist, so we gotta take a closer look. And then, everyone gets together in a room and paints a blueprint, if you will, to see, okay, this is where we ought to be going, and then the related strategies and tactics and projects are falling out. Is that what’s happening?

Mark Probst [00:27:09]:

Yeah. We are absolutely getting in a room, and we’re absolutely talking about not just AI, but all these other technologies. Daniel went through some really interesting ones around virtual care that he just talked about. And how can we change our operations using these technologies to lower the overall cost and increase the quality of the care that we’re providing. And we’re getting paid differently and more and more that payment is shifting to a value-based care where we’re getting paid to keep people healthy and we make a lot more money if we do keep them healthy than if we have them in our facilities. So it is absolutely part of our business and operational strategy on how we leverage these technologies. What Daniel was saying I really agree with. The one thing I would say is I do think we’re closer with AI in some of our patient experience and patient care. A mother who has 4 kids, this is her 3rd child. She’s seen otitis media and earache about 100,000 times because these kids get them all the time. For her to be able to interact with a device that can—we can take an image of the ear or she can put the symptoms in and then suddenly just have that be the interaction and a prescription is now at the Walmart or a prescription is on its way home. I think we’re within months of that kind of capability. And I don’t think that mother is going to have a problem. You know, on the flip side, if I’ve got prostate cancer, I’m probably not going to interact with my iPhone. So it’s where we are in the learning curve and we’re going to get more and more comfortable with it as we leverage and use these technologies and people, our patients are as well, they’re gonna find that to be a very pleasant experience and much, much less costly.

Sanjog Aul [00:28:55]:

Let’s take a quick break, listeners. We will be right back. And, David, I’ll come to you and talk about the approach physicians and surgeons and the hospitals who are dealing with them are taking with respect to the liability. Because we are inviting a tool, maybe a data resource, which we are even doing today, but then we are allowing it a little more power that we have allowed so far. So then if something goes wrong, who is left holding the bag? Who is going to take responsibility and accountability? Please stay tuned, listeners. We’ll be right back and explore.

Sanjog Aul [00:32:27]:

Welcome back listeners. So, David, the question here is about liability, accountability, responsibility. Who’s gonna take it when you’re going to put AI in the forefront, allow it to make decisions, and use anything which is dependent on AI to try to heal patients and things go south?

David Chou [00:32:46]:

Well, I’ll give you one example. When the hospital is performing a DaVinci using the DaVinci robot on a surgery case, obviously, they’re controlling the robot and they’re controlling all of those features. The hospital and the surgeons are liable. I would expect almost the same scenario where the AI is learning, is it being programmed based upon some sort of algorithms or some intelligence from a human, and is being performed inside of the hospital facility. Therefore, no matter what we say, the liability still falls on the hospital and the physician, the surgeons that are utilizing this tool. We can’t get around it. It’s almost, I mean, we break it down simply to let’s look at the EMR. Someone is the user of the EMR. So when something does happen, who is liable? It still goes back to the facilities. I see AI falling in that same pattern. There’s no way to get around it and hold something other than an entity liable. So, I don’t see much change with the use of AI. I think we may have to get a little bit more clever. We may have to get a little more tactical in terms of how we interpret that and design AI, but liability still falls in the same realm.

Sanjog Aul [00:34:08]:

So, Mark, to that end, do you think the hospital would really want to take liability for a piece of software or churning algorithm or anything which is coming from a third party, and it says, oops. I’m sorry. But here, they lose a million dollars. Do you think we are that forgiving in this world yet? 

Mark Probst [00:34:28]:

I don’t know that we’re that forgiving. But there’s gonna be some interesting challenges with this. The FDA takes a close look at anything that’s a closed loop system where a computer’s making a decision for you. And I think we’re going to see new policy, new regulation, new ways of having these systems tested that will increase our level of confidence in their ability to do more of this closed loop kind of information giving back to clinicians, or even in the case of some of these robots that we’re teaching where they actually do something physically to a patient. And that day is going to come for sure. But no, we’re not going to be that forgiving. And because as David said, we’re ultimately held liable. We’re held liable if we bring a tool and then we use that tool, whether we use it properly and it malfunctions or whether we use it improperly, it’s still within our facility and the patient’s making a contract with us for the care, not with that tool, whether it’s an AI tool or any other tool. 

Sanjog Aul [00:35:34]:

So I see that hospital is taking the responsibility, but let’s go to an individual physician because they also carry a malpractice insurance. And something could be coming to them if they did not do a good job interpreting what was suggested by AI. So where is—and this is a question for you, Daniel—where do we draw the line or where are we crossing the line from AI being an aid or a tool to a physician who makes a decision to AI becoming solely the reason why you did whatever you did?

Daniel J. Barchi [00:36:08]:

So I’ll channel part of a question that you asked earlier about artificial intelligence replacing jobs, and then I’ll pivot to responsibility. So we think of artificial intelligence like the example I gave as a helper, not a replacer for physicians and their own decision. So if I’m driving into New York City and I’m heading for the Holland Tunnel and I’m using Waze on my phone as a navigation system and Waze says, hey, why don’t you take the Lincoln Tunnel instead of the Holland Tunnel? It’ll be 15 minutes faster. I don’t look at my Waze and say, hey, you’re trying to take my job as a driver. What I say is, hey, that’s really helpful. You’ve just saved me 15 minutes. Thanks for that insight. In the same way, nothing that artificial intelligence does for physicians replaces them as physicians, and in fact, it enhances what they’re doing. Now, just like I won’t follow Waze down the wrong way off an overpass that looks like the wrong thing to do and I certainly wouldn’t allow Waze to drive me into a brick wall, physicians can use artificial intelligence and other tools to guide them. But at the end of the day, they are intelligent, well-trained people who know what’s best for patients and are unlikely to blindly accept what they’re being told by a computer and quite frankly from another physician or nurse who might have a different opinion unless that’s really gonna change the outcome for a patient. So I think the paradigm is less about ultimate risk, and it’s not binary. Is the machine right or is the machine wrong? It’s how does the machine help the physician do what he or she already does and make them better at it? But at the end of the day, it’s the physician’s responsibility to acknowledge whether that’s the right thing or not.

Sanjog Aul [00:37:54]:

Mark, what would be the reasons one would tread carefully adopting AI versus going all the way and going gung-ho? What would you be cautious in us—like a healthy skepticism? Should we be healthy skeptics or feel this is nirvana and let’s go for it?

Mark Probst [00:38:14]:

Well, I can only speak for me, but I’m a healthy skeptic. I’m very optimistic about the technology and what it can do, but I think we have to understand its capabilities. Plus, it needs to learn. It doesn’t have all the knowledge it requires. I had a friend that was in Australia recently, and they got into a cab. They ordered up an Uber. And what they ended up with was an Uber Tesla, and I guess that was a special brand in Australia. And they said they got in the car, it was a nice Tesla, and they started off to the airport. And the driver turned around and just started talking to him. And they got a little nervous because he wasn’t driving the car. The car was driving the car. And it was leveraging, obviously, some of the things with AI, but it was all brand new technology. And the driver is saying, oh, yeah, 90% of this trip, this car is just going to drive all by itself. I know for a fact they felt much more comfortable, one, that the driver was able to explain what was happening, but that the driver was there. Even though the car was driving itself and they were a bit nervous about that and they were paying attention, they were able to accommodate it because there was a physical driver in the car taking them forward. And they themselves were getting comfortable with AI with the car doing its own thing. That’s what’s going to happen in healthcare. We’re not just going to turn it all over and say, okay, now all of our prostate surgeries are going to be done by a computer. We’re going to learn, our physicians are going to learn, and our patients are going to learn. And when the outcome starts to suggest that this completely computerized process can do a better job if that date ever comes, then the physicians are going to be comfortable with it and the patients are going to be comfortable with it, and we’ll continue to move forward. But I think it’s going to be a very iterative process and not just on how smart the tool is, but on how comfortable and how trusting we become in its ability to do whatever we’re asking it to do. If it’s reading an image, great. If it’s performing a surgery, great. If it’s just, you know, moving supplies around the hospital, great. All of these are going to require trust and experience.

Sanjog Aul [00:40:26]:

David, what are we doing in terms of the dollars and cents? Do you think we are struggling with the business case given that, I mean, of course, the topic’s about improving patient experience? Is the corner office ready to accept the costs and say, okay, even if it improves patient experience, I’m good, or are there certain other things you have to demonstrate before this is gonna be tackled, or you’ll get this back?

David Chou [00:40:54]:

We definitely have to demonstrate the ROI. But there’s a lot of soft ROI that cannot be calculated. So I think that’s where the challenge lies in being more efficient, getting patients more engaged. Sometimes there’s not a hard ROI. It’s a very hard thing to measure. So I think definitely we have to put that use case in. Let me give you a personal example of AI that’s not related to patient care, but I have a virtual assistant AI robot that’s managing my calendar, and I also have the executive assistant. Now I’m at a point to where my executive assistant is really organized and we’re starting to promote her to be more of a project management type in terms of her role and responsibilities. Now this is an example of where AI has really augmented and helped an individual sort of take away the mundane passing of scheduling my appointment and allow someone to really elevate their skill sets. I really see that as a method to where AI can be used for patients here. You know, there’s gotta be plenty of ways where they could help a physician, a medical practitioner, nurse practitioner, anyone providing care become more efficient and really, you know, I’ll use an HR buzzword, practice to the highest level of their skill sets. So this is where I see AI come in, and sometimes it’s hard to measure the ROI, but I think we have to figure out how to come up with some sort of calculation to show the efficiency of having this type of technology in the workplace.

Sanjog Aul [00:42:30]:

So, Daniel, in your specific case, how many different unique business cases you have to develop for you to invite AI into different areas where it can add value, or you just get a blanket carte blanche to go ahead with it?

Daniel J. Barchi [00:42:46]:

It’s a great question, and I think one of the important notions about all of this is that it really shouldn’t be a technology-driven process. At New York Presbyterian, we’ve got physicians who are on my team in innovation who are focused on leveraging artificial intelligence in everything that we do, focused early on in ED throughput, OR throughput, and on length of stay. And so it’s not necessarily sitting down in a conference room and charting out where we’re gonna go with the process. It’s spending time with clinicians who know what processes can be improved, exposing them to how artificial intelligence works, and then having them guide us to how to leverage neural networks to improve what we do. So at the end of the day, like most things IT, it’s best guided by clinicians, and it’s not discrete business cases with proven ROI.

Sanjog Aul [00:43:43]:

And, so coming to you, Mark, if I were to look at different ways to figure out if AI is not working—so I’m sure there is some experimentation that’s going on—do you think we have a way out? Are we going in a unidirectional approach with AI? Because if you’re going to embed it in your health care DNA, if you will, and if it is not gonna show the promise because we are still in that experimentation stage, are we keeping that safeguard that we will be okay and the patients will be okay to go back to the traditional model?

Mark Probst [00:44:18]:

Yeah. I think we’re always keeping those safeguards in place. And because we’re so early, much of what we’re doing is either rudimentary AI or it’s, as you mentioned earlier, bolt-ons, or it’s piloting. It’s us learning about the capabilities so that we can do what Daniel was talking about and instruct our clinicians or our operational people on the use of the technology. So I don’t think we’re anywhere close to a point where we couldn’t back out of it. I do think the day comes, assuming AI continues to show the benefit that it’s looking to show where the systems are actually coded as AI. It is much more integrated into all of our EMRs, all of our ERPs, our phone systems, a lot of the things that we work on. And at that point, it’s gonna be more of a challenge to back out. But I don’t think we get there until we’ve proven that this technology is really providing the benefit that we believe it can.

Sanjog Aul [00:45:25]:

And, David, would you say there are any specific areas in which patients, after they have been given a dose of this new cool technology in a way AI can help them or maybe even without an individual physician present and they got their resolution or some sort of diagnosis and treatment, have you received any feedback for us to say, okay, whether we go one direction or the other?

David Chou [00:45:54]:

So, you know, health care in general, hospitals in general, we do a really good job of monitoring, taking care of the patient when they are in our facilities. Once they leave the facility, it becomes a black box. This is where AI can really come and play whether it’s creating a better experience or just helping the patient change behavior, getting them more education about their care plan outside or creating some sort of action and predict some sort of outcome. So that’s where we’re moving towards, because we do a really good job right now when patients are in care. But when they leave, this is when a lot of things could happen. Therefore, the net outcome is driving up costs. So I definitely see AI play in that realm, whether it’s giving patients some sort of technology to take home to monitor their behavior or making sure that they’re taking meds on time, things of that nature. That is definitely where we’re headed because the future of care is we want to provide care to the patient where they’re at. Currently, the patient has a contact facility to seek care, but in the future, it’s gonna be at the point of location where the patient is at and at a convenient location wherever they wanna be.

Sanjog Aul [00:47:09]:

And to that end, Daniel, when you are looking at an organization which is attempting to go in this direction and they are doing something, including yours, have you received direct feedback from all stakeholders for you to know if you are doing the right thing? Because if it’s so new, it is so unprecedented, totally new territory. How do you know if it’s gonna take you in the direction where you will not lose your shirt or at least you’ll not cause yourself more harm than good. And I’m being a healthy skeptic again and not saying AI is not good. But that’s something we have to worry about, isn’t it? 

Daniel J. Barchi [00:47:49]:

Right. You’re asking the right question, but that’s actually where academic medicine excels in that we are willing to try new things that are going to advance care and it’s not always bottom line based. So you asked how we get buy-in from all parties and the answer is, simply we can’t and don’t. And we’re doing not broad applications of artificial intelligence but narrow and targeted for two reasons. One, it is a new enough technology that we wouldn’t roll it out broadly. But by doing it in very limited and targeted ways, we can work very closely with clinicians who understand the impact when there’s not any risk of somebody blindly following guidance from the system and not knowing its downsides and risks. But by doing targeted rollouts, we can also instantly see results. So instead of doing a broad rollout where we might have to wait months or even years to see how it affects a broad population, by seeing how it affects individual physician encounters and patient outcomes, we can get a better sense for the efficacy of using these tools. So, again, we’re doing this not because it’s a money saver or because it necessarily makes people more efficient. It’s simply the right thing to do. Here’s a great tool that has the opportunity to really change the way we’re able to deliver medicine, and it’s incumbent upon us to embrace it, try it, test it thoughtfully. And then if it works well for our physicians and our patients, roll it out wherever we can.

Sanjog Aul [00:49:25]:

And I love the sentiment that you shared that is, you know, we are not out there to just think the bottom line. Would you think that overall health care, whether you talk about population health, patient centered care, and even bringing AI to help in, you know, improve patient experience, do you think if we were truly altruistic in our vision, would we have really been struggling as much as we are today in different areas? And the only reason I bring this up is this is a great disruptive technology, and we are trying to make things best for the patient, but somewhere things are falling through the cracks. What would those be and how can we plug those holes so that we do not continue to bleed even though you have such a disruptive technology here? Mark, why don’t you answer that one?

Mark Probst [00:50:19]:

Yeah. That is a pretty broad question you asked. So there are a lot of reasons why health care is expensive. There’s a lot of reasons we have challenges as an industry. And some of those are demographics. Some of those are simply where we’re at and understanding how to provide care. Some of it is just expectation on the health care system that may or may not be realistic. But there are problems that I think these technologies can help us solve. David brought it up earlier. Our patients, we care for them when they’re in our clinic or when they’re in our hospital or an urgent care center, but when they’re not there we have a very difficult time understanding what they’re doing and assuring that the care continues to happen, that they’re self caring, and that they’re doing the things that they were prescribed to do. I believe this technology very quickly is going to be able to monitor our patients while we’re not here. And through AI they’re gonna be able to understand behaviors that our patients are involved in or not involved in, and nudge them as Daniel said earlier, into doing the right things. So part of it is if people can take care of themselves better they’ll become less of a burden on our national healthcare system. They’re not a burden to me or Daniel or David because that’s what we do for a business. But they are a burden on society because they take costs that could be spent on other things and they’re driving up the cost of healthcare. So these technologies are going to play a role in helping us solve some of these very macro issues that you were bringing up earlier, and I’m excited about the potential of doing that.

Sanjog Aul [00:52:03]:

On behalf of the show and our listeners, I’d really like to thank you, Mark, David, and Daniel for sharing your respective views on how artificial intelligence can be utilized to improve patient outcome and even help health care to help solve the broader issues, reduce costs, and come get some more benefits. So thank you so much again.

Daniel J. Barchi [00:52:25]:

Thank you.

Mark Probst [00:52:26]:

Yep. Thanks.

Sanjog Aul [00:52:27]:

Thank you. And, listeners, please like us on Facebook. Search for CTN, CIO Talk Network, and be sure to follow us on Twitter and join our LinkedIn community. Thank you again for listening to this segment on CTN. This is Sanjog Aul, your talk show host. Till next week. Take care, and God bless.