← Tomorrow's Cure
Loading summary
Transcript55 lines
- [00:01]
A
Hello, I'm Kristen Meinzer, host of Health Matters, a Mayo Clinic podcast investigating topics big the influence of gravity and small brain, eating amoeba. We talk about our health bodies and the world around us. Get off the toilet, you're gonna have hemorrhoids. So join me in conversation with some Mayo Clinic experts as we strive for happier, healthier lives. Mayo Clinic's Health Matters new episodes every other week.
- [00:28]
B
Wherever you get your podcasts, now is the time where we have an opportunity to make a real difference. We can bring care to patients in the comfort of their homes. We are at a point where we are able to blur that separation on physical care and digital care. But for this to happen, it's going to need more than technological capability. It's going to need that partnership, the regulators, the policymakers, and everybody to come together.
- [00:57]
A
Everywhere you look, automation and AI are reshaping healthcare from smarter diagnostics to streamlined workflows and even big decisions about how machine learning fits into human centered care. But how do we make sure it serves patients and caregivers, not just algorithms? Are we chasing efficiency and hope or overlooking the risks and responsibilities that come with it? That's ahead on this episode of Tomorrow's Cure, a podcast from Mayo Clinic that brings the future of medicine to the present. Thanks for being here. I'm Kathy Werzer. We have two great guests who've spent a lot of time thinking about the intersection of healthcare and AI. Dr. Anjali Bagra is a professor of medicine at Mayo Clinic and directs enterprise automation, keeping patients and people at the center of every technology shift at Mayo Clinic. Dr. Ravi Bapna is the Curtis Carlson Chair in Business, analytics and Information Systems at the Carlson School of Management at the University of Minnesota. He's also the author of Maximizing well Being in the Age of AI. Truly a pleasure to have both of you with us. Thank you so much.
- [02:02]
B
Thank you, Kathy. It is a pleasure to join you.
- [02:05]
C
Absolutely. Pleasure.
- [02:06]
A
Zolas, I know that you too have been keeping track of what's happening in our society. And as you know, the people building AI are saying that the technology is advancing more rapidly than the vast majority of people realize. And I'm wondering if the technology develops at the pace that lab leaders are predicting, how prepared is society for the changes?
- [02:26]
C
Let's take a step back. I mean, I think AI being this very powerful general purpose technology, unlike anything that we've seen before, has created a lot of sort of angst at some level and also has this crazy potential. Right? And a lot of this is because Unlike prior general purpose technologies like electricity or computing, it's fundamentally intangible. You know, these algorithms are lurking in the background, let's say, you know, doing diagnosis or they're predicting. People don't understand them necessarily, right. They can't see them, they don't have a relationship with them. And therefore it's pretty much, I would say, misunderstood. And I think the challenge when you have something like this is that the ability of organizations to integrate and adopt this into, you know, whatever they're doing to improve their, let's say, delivery of healthcare, right. That takes time. That's a challenge, right. It's non trivial. And I think that's the stage we are in right now. Right. The technology is moving at a really fast pace. Organizations are trying to keep up, but I think there's always a little bit of sort of this lag that I think we are seeing playing out.
- [03:35]
A
So Dr. Bagra, how prepared is our healthcare system for these changes that AI and automation are putting in place?
- [03:41]
B
Great question, right? Because that's a fundamental question. A, why do we need this? And B, are we ready? So just to give a little bit of context to our listeners, this is not new intel, but we know that healthcare is at a breaking point. When we look at the cost of running healthcare today, it's over $4.5 trillion. So nearly 1/5 of our GDP is going into sustaining health care. And we've tried all the traditional fixes in the past, we've tried hiring more people, we've tried new payment models and we've tried digitizing the analog and we know that none of this has worked. So I loved how Ravi kind of kicked off with, you know, AI. Is this potential not fully harnessed yet? Because industries need to do their preparation to be able to harness all the benefits. So I would say today's question is not how do we do more of the same kind of a model that we've used in the past within healthcare, but it really is how do we do things differently and how do we harness the capabilities that we have today that we've never had in terms of how ready we are? Before I jump into that, I do want to be clear that AI and automation are not replacing people because I think that's a fundamental concern or skepticism that we hear. So no conversation is complete without addressing that upfront. So these are not replacing people, but they are adding to our toolbox. They are adding tools that we've never had before to allow creating scalable solutions for problems that have been age old and how ready are we? Well, I'll say that automation is already quietly changing healthcare in the background. There is so much happening in the background that may not be visible to our patients. For example, things like automation and AI enhancing images and enhancing our ability to read images for picking up things that human eyes can't see. And this is not just in radiological images, but also in pathological images and pathology specimens. We generate vast amount of data in healthcare and we use less than 9, 10% of data that we generate because of the impossibility of the human mind and brain to be able to harness the data. So, so I would say we are very much ready to use these in terms of opportunities where the work continues to be needed to done is processes for seamless integration, for enabling the human in the loop where necessary for enabling that these technologies actually help us be more human and more nimble with taking care of our patients. That these technologies help us do what we've never done in the past, diagnose disease way before we've traditionally done them and really provide the personalized care that every patient deserves.
- [07:01]
A
Can you help me out here, both of you? Let's take automation, perhaps when it comes to automating some repetitive tasks, I can see that helping the situation. But I'm wondering how it might enhance the value of physicians and say researchers expertise, right? How, how is the use of AI automation helping to amplify human skills and strengths?
- [07:24]
B
So one example, quick example that really jumps out is what I use in my office every single day. I see complex patients from across the national practice at Mayo Clinic. On an average, my patients have around 20 to 25 complex problems that we need to address because they are referrals from other medical centers. So there is a lot of information I gather during my visit with my patients. Traditionally back in the day I would need to go back and dictate all of that information in a clinical note that could then be used as a reference by other specialists and other physicians taking care of these patients. But now I don't do that anymore. I get my note generated automatically. It's automated through use of ambient AI that works in the background ubiquitously. My patients love it. I show, show it to them. You know how this is in the background and we are going to have the note pop up. I don't need to memorize every detail. I don't need to scribble details about their medical conditions. I give my full attention to my patients when I'm in the room with them without being distracted. And the need to Write things down so I can recall important information to be able to include it in their clinical note. I think that to me has allowed me to be more human, more present, more in tune with my patients emotions without having to be distracted by a computer screen or scribbling the notes. And quite honestly the same note can be used with the use of AI to generate a patient discharge sort of after visit instructions in a very non medical jargon terminology. So there is a lot of day to day application in my clinical work. Similarly, there are many applications in research and I'm sure Ravi can share a few of those.
- [09:22]
C
Just to build on, I think couple of points Dr. Bhagra made. I mean I think one, you know, this, this example that she gave. If you think about the value of this data that's being generated, right. One of the superpowers of generative AI at least, right. So we can think of sort of two buckets of AI, traditional AI, which is still very useful, you know, doing predictions, you know, like she was talking about, for disease and variety of other settings and then generative AI. So we can think of this as a, you know, supercharged cognitive engine. Right. So if you have all these transcripts from all these conversations between patients and doctors and of course once they are duly anonymized and sort of, you know, cleaned up, that itself becomes a really, really rich repository of data for us to in the future understand, you know, again, how disease spreads and how in a personalized way you can have these interventions out there. Right. So, so I just wanted to sort of again just build on this idea that we are still harnessing only a very, very small fraction of the data we generate as a society. And I think one of the things that generative AI has done is as it actually has reduced the cost of doing the analytics on this data. I think that is a really exciting potential. Right. And just another thing that I think will be maybe we can get into a little bit more. And also related to your question Kathy, was this idea that, you know, automation, yes, we have to do that. Every organization is going to look to do that to free up, you know, scarce resources in this case time of physicians. In addition to that, the lens that we really propose, and Dr. Bhagra and I have been working on a framework for this around human centered AI is the lens of augmentation, right. So how do we augment human capacity in a way. And I think that lens becomes a really good way to think about sort of this future with AI.
- [11:14]
A
Say I don't want to get Too much in the weeds. But I was thinking about this when it comes to generative AI. I was reading somewhere that about 95% of generative AI pilots failure in the healthcare industry. And I'm wondering what is it about healthcare that makes automation uniquely challenging?
- [11:32]
B
I think that number will shrink over time, hopefully. And that's our goal collectively. But one of the things that gets missed sometimes in this cycle of innovation is technology is just one part. What we need to align is people and processes. We really need this triple aim to come together. It's technology, people and process. Technology brings capabilities. But if the processes are not ready, if the processes are not updated, and if we don't do the very important step of process mining and then layer technology that's brought in by people, that's often the reason why we see high failure rates in pilots. It's either the process that was not mined or the people that were not appropriately scanned skilled already. So in our case at Mayo Clinic for example, this process is very decentralized. This is not a top down initiative where design studios are coming up with gen algorithms and then we are plugging and playing. It's actually quite the opposite. We basically start with defining a problem. We start with a question. What is the problem are trying to solve and how can we get to the desired state? And that really involves deconstructing the current workflow. I feel organizations that perhaps are doing it in a more expedited way without deconstruction, without really understanding the process, without upskilling the people and without the buy in of the people, that's where failure rates are higher. Aside from the people and process, we also know that technology is rapidly advancing. So what was the best technology six months ago is obsolete in today's world. So I think there are many complex factors, but largely speaking within healthcare, if people, process and technology aren't moving together and if we aren't keeping up with the pace of the change, rapid change in technology and managing change around in the organization, those are big areas of cracks. I feel at a systemic level and of course there are many others, you know, ability to scale pilots and cost prohibitiveness and people and process, we got to keep that always ahead.
- [14:03]
A
It sounds to me that in terms of the implementation of some of this has to come, as you say Dr. Bagra, from the ground up, right? From bubbling up. But if you've got managers, I would think that there has to be a level of trust and transparency. Right when you're implementing say automation in a hospital setting, where does that come from who actually drives that forward to help the workforce?
- [14:29]
B
Yeah, that's such a great question, Kathy. I would say within healthcare, trust and transparency are the basic tenets on which we build our relationship with our patients within our team and everything we do. Trust and transparency is core. Now in terms of who is responsible for trust and transparency, I think this is everybody's responsibility. We take it very seriously as care team within our organization. But I think it's very bi directional fundamentally with a patient, we know that I will not be able to achieve any outcome with my patient if the patient doesn't put their trust in us. To be able to get to that trust. We are always very transparent with the knowledge we share with them, with the prognosis of their disease, for example, the kind of treatments that are available with what is not possible. That's also a very important part of the discussion and why I'm sharing this with you is we use exactly the same framework when it comes to technology. It's not any different. It's a mindset. It's a mindset of building trust. So for us, within all of our innovation teams at Mayo Clinic, this is core to what we do. Having everything on top of the table, what works, what doesn't work. We never take technology specifically as proof of success in any way. We also discuss openly the challenges that come with it. Challenges for our staff in adoption, challenges for our patients when we bring these in the room because it is another extra layer or a paraphernalia in the patient's room. So transparency, being open, discussing failures as much as successes, discussing pitfalls, discussing where these can be used to our advantage, but also then discussing what some of the unintended consequences some of these tools are within our practice, discussing how we are at a risk of making things transactional versus experiential and always being proactive in managing around that. I think all of those are behaviors that need to be adopted by organizations and everybody in the organization from top to down, from bottoms up and all around. This is, I would once again say a combined responsibility. And this doesn't happen just during casual conversation. I think what's also wrapped around this trust and transparency is strong governance. What are the things we don't allow with these tools? What are absolute non negotiables in healthcare? And I think there is good deliberate attention to this. For example, Ravi was talking about his framework. We don't do this work alone. We do this transparently with all of our partners that we are learning with.
- [17:23]
C
And from this question of where does innovation Come from who? Bottom up or top down is a eternal question in many ways. And I think the answer is actually both Right. And in the specific case of AI, what we find is that in many situations, the awareness of the art of the possible is sort of, you know, what's the gap when you start getting up into the director, senior vp, into the C suite, Right. Because again, many of these folks, when they were trained, this was not a thing. Right. And the hype is so much. Right. So what's the reality and what's the hype? So we do a lot of work at the Carlson School just trying to help people see through this and say that, look, you know, these are the, you know, sort of low hanging fruit. These are sort of the, you know, more ambitious projects. You have to have a process to get to these. And I think just demystifying that for the senior leadership, that's the top down angle, I think that sometimes is missing. Right. So when these things come together as they do at Mayo Clinic, and I as a patient have experienced all of this, right. Which is kind of a weird thing to say, but when Dr. Bhagra was talking about, you know, kind of this explainability of a treatment, last time I was there, I got a report from my, my visit. It actually had links to break down complex medical terms that I could understand, right. As a patient. Right. So I think that's where again, this human centricity of deploying AI becomes really first and foremost.
- [18:47]
B
I also want to share that Mayo Clinic was one of the founding members of CHAI Coalition of Health AI. And this is really around responsible use of AI, because yes, we can do this and a lot of times we have capability. But the question becomes, should we? What are the defining tenets of responsible as well as ethical use of AI? So trust and transparency is everybody's responsibility from top down to bottoms up, as Ravi mentioned. And it also needs creative partnerships because healthcare impacts all humans. You know, it's really an experience by our society. So trust and transparency also cannot be built alone. We need partnerships. Hi, I'm Dr. Bill Maurice from Mayo Clinic Laboratories. Curious to learn more about healthcare innovation. I'm Dr. Bobbi Pritt, host of Answers.
- [19:41]
A
From the Lab, a podcast that explores.
- [19:43]
B
Trends and innovations in laboratory testing and clinical clinical diagnostics. New episodes drop twice a month. You can subscribe on your favorite podcast app or visit mayocliniclabs.com to learn more.
- [20:00]
C
Busy healthcare professionals. This one's for you. Find Mayo Clinic talks on your favorite podcasting app or visit CE Mayo Edu podcasts to learn more. Every week we share succinct, relevant and practical medical insights tailored for healthcare clinicians that you can immediately apply to your practice. Each episode covers common health issues seen in a primary care practice shared by Mayo Clinic experts.
- [20:38]
A
You mentioned the human element here. And Dr. Bakwar, you also talked about transactional versus relational care, right? I wonder if there are worries. Do you worry that without proper guardrails some of the our care could become dehumanized because of the use of automation in AI?
- [20:58]
B
This is a very important question. Always in front of us is what we are doing adding to our patients experience, adding to our patients outcomes, adding to their longevity, early diagnosis and are we unintentionally in some ways trading off the experience in healthcare? I would say there are many aspects in healthcare delivery that are very transactional, right? So there we know 30% of work done within healthcare is administrative work, is pure administrative work. Now if you think about that administrative work, a lot of that is repetitive, very transactional and you know, doesn't really include a lot of relationship building. I would say it's a no brainer that that pocket of health care work is ready and ripe for automation with appropriate human oversight. However, when we come to taking care of human beings when they are most vulnerable and needing human connection, that's where we don't want care experience to become transactional. Now it's also true that one of our biggest challenges in healthcare is burnout, right? For the longest time we've admired the problem by generating massive amounts of data around what is burning out the healthcare workforce. But we haven't really had tools to address burnout at scale. So I would actually argue and introduce this concept of AI and automation paradox. I think these tools allow us to rehumanize healthcare because I shared with you how before I was using these ambient tools in my clinic, I would be distracted to some extent because I didn't want to miss any information. And I'm scribbling and I'm on a computer screen trying to make sense of all the data, massive amount of data that my patient is sharing. But now I can have my entire attention to my patient. So that's I think a very real example, relatable example where I actually am in fact more human with my patients. And I can tell you Kathy, I decided to become a physician in elementary school and I was not envisioning documentation and paperwork. When I decided to become a doctor that young, it was the hope and healing, always the hope and healing aspect of the profession that drew me into this calling. So I would say yes, when not done right, when not done responsibly, when not done ethically, that's a risk. But when done right with appropriate human in the loop, with appropriate mining of processes, and I love what Ravi said earlier, with clear definition of the outcomes we want to drive, I think these tools really allow us to be more human.
- [23:58]
C
And I think the power of frameworks is that they help to communicate complex sort of ideas to a lot of people in a simple way. Right? People remember them. So one of the things that Dr. Bagra and I have been working on, which is around this again, idea of human centered AI, is what she had referred to earlier as the four A's, right? And we already talked about augmentation. So this augmentation mindset rather than just pure automation, we have to do automation to save up that time. And that the 30% that she was talking about, right? But then can we use that excess capacity to augment the human sort of enterprise? I think that's one of the forays. Right? But if we start from the, from the very beginning, you know, the awareness of the art of the possible, I study, you know, companies that succeed in, companies that fail in the tech sector. The most powerful CEOs are going back to school, right? They're in the business of lifelong learning. So this awareness of what can we do and having that sharp nose to say, you know what? This is not actually a good application for our setting. And this one is, I think that's really important. So that's the second A that we talked about building this technology in an accountable way, right? We know AI systems, when not designed with the appropriate guardrails, can actually exacerbate biases that we have in society. Right? And so there is a lot of research happening at the eu, at other leading institutions around the world on making sure that these systems are accountable, that they are fair, that they are transparent, they're explainable. Right? So there's. So I think that accountability becomes really important and just projecting to the future. I think the fourth A we bring into this framework is this idea of thinking in an agentic way, right? So I think agentic AI is like the hot new buzzword. People ask me what is this? But the simple idea that your listeners can sort of think of is we have this very powerful cognitive engine gen AI capabilities, can understand and even get into the perception of human language in a very powerful way that we didn't have before. If you marry that with existing knowledge, research documents, you Know, the repository of collective wisdom that exists in a digital format at an organization like Mayo or at the University of Minnesota. And you also combine that with the, all the tools you have already built out using traditional it, right? So anything that has, you know, what we call an API, which is just a, basically a way to talk to a system. Right. So now I'll give you an example. Just to make it really, really simple. You know, I was teaching my students the other day that they could build an agent that can go out and do a comparative analysis of, you know, two different, let's say tech companies make some projections about, you know, forecasting their revenue and then connect that insight, summarize it and send it into the slack channel of a particular team in a particular organization. Right. So I can connect that to, in this case, the tool, external tool would be slack. Right. When you start thinking about combining these things together, that really opens up tremendous capabilities that we haven't even started to scratch the surface of right now. So I think those four A's, I think become really powerful to thinking about human centric AI.
- [27:14]
A
Can I talk a little bit more here about human centered artificial intelligence in that I'm wondering if there's pressure to, to know when to trust AI and when to override it. And I'm wondering if that might be an ongoing mental calculation, it might be mentally taxing, say for a physician as it pulls physicians out of possibly intuitive patient centered care and into a constant kind of risk analysis situation. Might that be a pain point? Might that lead to some burnout?
- [27:48]
B
So information overload definitely causes more cognitive demand. But as physicians, you know, we are always trained to compare contrast data. So before AI was evidence based medicine, before offering a treatment to any patient, you know, we always want to ensure that we are staying up to date with the latest clinical trials, the latest studies. So I think just by nature our job, we ingest massive amounts of data, not just objective data points on our patients, but really to pick the best treatment. There is a lot of analysis going on to, like I say, offer the best possible treatment to our patients. There is definitely a learning curve in how to use these technologies to our advantage. But I'll give you an example of how this actually has flipped our abilities actually in a way that we can be more efficient with less cognitive demand. Because with tools such as Open Evidence is a great example where instead of me manually going in and researching the latest and the greatest clinical trials to help guide treatment for my patients, which is what I would ordinarily do, now I can use tools that can give me the five most important studies to know. And based on my patient presentation, here is a suggestion of the best possible roadmap for my patients. So I'm kind of shifting gears now in how I'm spending my, I would say, cognitive bandwidth. I'm making different kinds of decisions compared to when I didn't have these tools available. So yes, to your point, there is a training curve to this learning curve to this how do I use these technologies. But essentially, at the end of the day, I'm the one making the best decision for my patient in combination with, with their desires. So I think these tools are actually allowing me to have less cognitive burden versus more cognitive burden because they are allowing summarization of data that I would ordinarily spend massive amounts of time on. But yes, it does involve me getting more adept at using these tools.
- [30:07]
A
So the buck still stops then with the provider?
- [30:10]
B
Absolutely.
- [30:11]
A
What happens when an automated system contributes to patient harm? You know, who's responsible? The nurses, the software developers, the healthcare institutions, the provider? Has anyone figured that out yet?
- [30:23]
B
I think this is a question that's asked across industry. When things go wrong, who is responsible? I don't think it's just healthcare, but like with everything else, you know, at the end of the day, the buck stops with the care team taking care of the patients. And that's why this responsible integration with the appropriate human in the loop and, you know, the appropriate governance and the oversight are absolutely critical. There have been record number of lawsuits who can sue an AI algorithm? Like, who is responsible when things don't work out the way they were intended to for us in healthcare, you know, we know that it stops with the team. It stops with the physician, with the people taking care of our patients.
- [31:08]
C
You know, as an educator, you know, we have to really think about changing the model of our educational system in the light of this new technology. Right. So a big skill that we have to teach people coming out of the Carlson School of Management is how to use the AI in the appropriate way. Right. So I have to, I have to bring these tools into the classroom. I have to push the students. My baseline has shifted. If I was evaluating you at a certain level right now, I've already moved that up because there is a certain set of capabilities I know the AI can do, and I want you to use that. And then I'm also going to coach you and then evaluate you on how you use the AI and make sure that you have good evaluation of the AI technology. So that is going back to this idea of, yeah, these systems are probabilistic, they can make errors, you're responsible for those errors. If you go out, make a decision by blindly trusting this, okay, the buck will stop at you. Right. I think that becomes important. The only other thing I would add out here is that, you know, I was on a, on a similar panel with Dr. Bhagra a couple of years ago and I heard this example that, hey, you know, maybe the doctors that, you know are not using or not augmenting their capabilities with this latest model that can detect, let's say, that can look at mammograms and detect breast cancer in a very accurate way, maybe the lawsuits will come for those doctors for not using the technology actually. Right. Because you're probably again, underserving the patients, actually. So that's another angle that we should keep in mind.
- [32:43]
A
We've been talking about augmentation, automation, awareness and accountability, the four A's when it comes to AI. I'm wondering, as we look at what's happening in the healthcare industry, do you worry about rural or underserved facilities that may lack the infrastructure at this point to adopt some sophisticated automation technologies? I mean, if they can't keep up, might that lead to worsening healthcare disparities?
- [33:09]
C
I mean, in my view, you know, it's an opportunity to use this technology to better serve those communities that already we are not serving. Right. At the appropriate level. Right. So partly, like, you know, like Covid was a big shock to the system and suddenly telemedicine became very accessible and it allows now, actually people who are not really close to a particular, let's say, hospital or a clinic to actually use, like in this case, this virtual technology to even talk to these doctors out there. Right. And so I feel that this becomes a real opportunity to extend that high level quality of care. I feel that this is the time to do it. The resources and the tools are there for us to actually make this a much broader capability accessible to a lot of people who are underserved right now. Will that happen? I think that's going to be determined by other social, political forces that unfortunately sometimes don't work in the way that we want.
- [34:10]
A
And Dr. Bhagra, what do you think?
- [34:12]
B
I completely agree with Professor Bhapna on this. I would say we've always had some kind of tailwinds historically when we look at times of massive change. Covid was a classic burning platform, you know, that really united humanity across the globe. And that's where industries came together, regulators, providers, insurers, like everybody wanted to do their fair share in saving humanity. And I think that is a demonstration of what is truly possible. Because when we go back to our old ways of doing things, I think we go into those silo mentality. But specific to rural health care. Those are the kind of the tailwinds that we need. And we know that we are projecting workforce shortage not just within healthcare, but across industry. But we know that it affects healthcare disproportionately. For the longest time we've been talking about shortage of primary care physicians and think about the solutions that are available to us. Now is the time where we have an opportunity to make a real difference. It may not be human workforce, but we can bring care to patients in the comfort of their homes. We are at a point where we are able to blur that separation and physical care and digital care. But for this to happen, it's going to need more than technological capability. It's going to need that partnership, the regulators, the policymakers and everybody to come together. So personally, as a physician, I've practiced across the globe and I've seen very disparate areas of medical care. I think we have the capability, we just need that result. We need the continuation of tailwinds and understanding of other factors that need to come together and align.
- [36:17]
C
Busy healthcare professionals. This one's for you. Find Mayo Clinic Talks on your favorite podcasting app or visit ce mayo.edu podcasts to learn more. Every week we share succinct, relevant and practical medical insights tailored for healthcare clinicians that you can immediately apply to your practice. Each episode covers common health issues seen in a primary care practice shared by Mayo Clinic experts. Are you trying to become a better human? Well, the Human Optimization Project is an exciting new podcast offering from Mayo Clinic which aims to help you do exactly that. We're all struggling to balance the need to perform and do more while simultaneously maintaining our wellness. In the Human Optimization Project, we're focusing on 10 key domains of human performance to help people achieve more, improve well being and become the best humans they can be. If that sounds like what you're looking for, search for the Mayo Clinic Human Optimization Project on your favorite podcasting app.
- [37:35]
A
We've had a really interesting conversation about the possibilities of AI in healthcare here and I'm so grateful to you both. But I was doing some some research before our conversation and the literature reveals though some significant gaps in our understanding regarding patient safety, ethical considerations, data privacy, a lot a lot of unanswered questions yet, right? I think you both would probably agree with that and there's no way to put the genie back in the bottle. But I'm wondering what kinds of guardrails might be needed and can they be implemented at this point to make a difference?
- [38:08]
C
Yeah, sure. I think, you know, it goes project by project. You know, I think the lens that we started off, okay, what is the current friction in the system? What is the need that we need to solve for? And once we take that perspective, which I strongly endorse. Right, because the opposite is, you know, there is something shiny new object out there. Let me try to, you know, get some insights from it or get some value. You know, that's not a very effective way to start. So if you start with that problem driven approach, then for that particular context, you know, there will be a set of relevant guardrails. For example, if you build an algorithm that is going to, let's say, you know, look at a variety of socioeconomic indicators, it's going to look at imaging, it going to look at pathology to detect some type of disease. Is the representation of different subgroups of people, you know, sufficiently balanced in that algorithm? Right. And if it is not, there are methods now, like including things like synthetic data generation and other things that, you know, my PhD students, others at R1 universities are working on that can A, help you, you know, detect whether you're likely to see these types of failures or sort of, let's say, differential, like false positive rates across different subgroups and B, then to go back and retweak those algorithms and to improve them. Right. And so I think that has to be baked in, into every literally problem we want to solve using this technology out there.
- [39:38]
A
Dr. Bagra, what do you think?
- [39:39]
B
Professor Bhapna nailed it. It really needs to be baked. So there is this idea of mindset in healthcare. You know, it's always safety first and it's physical safety, but it's also psychological safety. So we are all about safety, quality. That's at the core of everything that we do. And I think with these technology integrations that we are seeing in healthcare, we are actually introducing a whole new set of capability. I'll give you a tangible example. When I trained many decades ago, not gonna give it away. You know, we were always in the ICU setting. We were always preempting things like sepsis so that we could mitigate a, like prevent sepsis, but then have metrics around early diagnosis and appropriate management, minimizing mortality from bad outcomes from sepsis. And there is massive amounts of data generated on every single patient from Every ICU bed, including ventilator data, hemodynamic data, their biochemical data, and lot of other very complex data points. So it can be very challenging to stay on top of all of that. But fast forward, in today's day and age, with appropriate use of AI, we can detect early signs, very early signs of sepsis and act on them. So when we talk of patient safety, you know, there is an element of a human error, rapid deterioration, and processes that need to come together. And I think, like, these tools are really now allowing us to mitigate all of those patient safety episodes by early diagnosis, early detection, but then also with agentic that we were talking about earlier, also mitigating some of these processes in an algorithmic way. Definitely. This is, again, there are two sides to every coin. We just need to always stay on top and have that responsible integration of technology. But realistically, we definitely have way more capability at this point to mitigate patient safety events with the appropriate monitoring. That was traditionally not possible with just human oversight, but with human and machine oversight together, we have way more capability. The other thing I would say is we talked about workforce shortages. Our nurses are overworked. You know, they have a lot of burden of taking care of the patient at the bedside, but also documentation. So using these technology, we can actually minimize some of those safety concern by again, augmenting and automating, where appropriate, some of those processes within healthcare delivery. So we actually are in a very good position now with the use of these tools to prevent many safety events that were not possible with human alone oversight. So human and machine brings great possibility.
- [42:45]
A
Can I center patients for just a moment? Because I want to bring them into the conversation. If they're listening right now, I'm betting they're thinking, okay, so what can I do? What could my family do to prepare for a more automated future in medicine? Do you have any advice for patients listening who might think, okay, well, this is going to be the future. What do I need to do as a patient professor? Any thoughts on that?
- [43:09]
C
We know to some level what they're already doing. Right. I think it wouldn't be a surprise to anybody saying that people go to ChatGPT and ask for, you know, questions. So patient journeys, we, we like to think of consumer journeys, but patient journeys have, for a long time now started in the digital realm. Right. Google used to be the place where people would go and look for answers. I think that shift is now moving towards platforms like ChatGPT and others. Right. And so I think that's a great resource. I Think for patients to inform themselves and to, to come in, you know, at a higher like baseline level of knowledge and information when they are meeting their care providers out there. Right. So that's my initial thought. I mean Dr. Bagra, I don't know if you agree.
- [43:53]
B
Yeah, I think if we go back to our four as that we've been talking about, the one that I encourage my patients and honestly learn a lot from as well as the awareness piece, are consumers of technology outside of healthcare for a variety of different reasons. Right. If I watch Netflix, I will get 10 recommendations of you know, what I need to watch or if I want to make a reservation in a restaurant in any town. It's not like I'm going to physically drive and then think where I want to go. So I think humans, all of us are patients and providers, we are becoming very agile, almost like dependent right on technology for our day to day comforts and conveniences. And if we just kind of bring that whole thing within healthcare, that experience in healthcare, I think our patients can engage way more effectively with their healthcare experience if they kind of adopt these tools to their benefit and advantage. I can tell you when I trained, I trained in the day and age of paper charts. My patients did not have access to their records at the time and they would have a health related conversation pretty much only when they walked into the clinic on their scheduled healthcare appointments. And today my patients, it's very different. They're on their portal all the time. My patients get access to their tests before I can see their tests. They send me questions all the time. We have conversations all the time. And of course it was Dr. Google for the longest time, but now it's Dr. Chatgpt. So I think we are blurring the boundary and we are also reducing that hierarchy in the relationship that existed in the past. To me this is a very exciting time to be a patient. Right. Like I have more access than I ever did to my own medical records. I can see behind the scenes, I can see under the hood.
- [45:56]
C
And that gives you agency, right? That gives them agency.
- [45:58]
B
Yes, exactly. So I think that to me and having practiced at a time where my patients did not have access to any of this, you know, it was pretty much a one way communication of information from me to them every time I saw them in clinic or rounded on them in the hospital floor, today feels very different. It's combined decision making. We are in this together and I think it's more powerful. It's the right thing for us to do. But there is still candidly a pocket of patients who may not have the right awareness of what the capability of these tools is. And I'm always focused on, you know, how can we do a better job of empowering and educating our patients about our partnership and how we could be better partners in their care experience.
- [46:49]
A
Dr. Bagar, you mentioned this, that there are some people who are just, that might not understand what's happening or don't understand the technology. They might be afraid of it. What would you say to a patient who might have some misgivings about where we're going right now?
- [47:05]
B
Yeah, I think it's very understandable. You know, human fear is what keeps us alive. Like that's the reason why we stay alive. I think sometimes the fear becomes irrational and the reason for that could be just lack of information. So when I come across patients who fear this technology, my fundamental approach is to first try to understand, meet people where they are because they could be very realistic reasons for the fear. A lot of times I find it was just lack of information. And once I share more information, and we talked a little bit ago about trust and transparency and when we discuss the possibilities transparently as well as the trade offs and how we are managing proactively responsible integration of these technologies, I would say a lot of times my patients come along, in fact, they, they get excited about what's possible and what they weren't really aware of. Sometimes though, it does raise a question, you know, whether technology is the best tool in a certain situation. And like I said, you know, technology is always and will always be just a tool, another tool. It does not replace what our patients come to healthcare for. It does not replace the hope which we humans are here to give and the healing that our patients should experience.
- [48:32]
A
Say what? I end our conversation on a personal note and I would be remiss if I didn't riff off of Dr. Bapna's book Thrive. How do you both plan to personally thrive in what is a very rapidly changing environment? Dr. Bopna, I'm going to start with you.
- [48:49]
C
My life's mission, I think for the last 25 years has been to help leaders in companies, executives, managers, try to sort of, you know, convert their data into an asset. Right. To really embrace data driven decision making, to embrace the scientific approach. There's lots of evidence that shows that as being a business school professor, that this is strongly linked to higher, better, you know, performance. Right. And not just profits, but also, you know, just sort of overall greatness, I think, right. In companies. And so for me, you know, this this new technology has been like, you know, again, a serious dose of tailwind. People have started appreciating, you know, what I've been trying to say for 25 years. So. So I think there's a big audience now to try to demystify what all of this is and to give people direction. And again, I am a lifelong learner, I think, as everybody is in this panel. And that's what I try to convey to others as well. Right. Like that the only way to sort of, you know, thrive, truly, is to embrace that. And then, you know, I think if we collectively do this as a society, we can have a level of sort of prosperity and care that we have not had before, I think. So I'm really excited about that potential, Dr. Bagra.
- [50:01]
B
Now is the time where these tools are enabling us to thrive in what seemed like a broken sort of system to some extent. So I would say now is the time to focus on smarter, more human care. What's most exciting to me and what helps me thrive is that we blurred the boundaries. As I mentioned a little bit earlier, we are in this together. There are so many of us who are excited about the possibilities. We have more partnership in this work, and that is very invigorating and personally inspiring to me. And Ravi and I have been, like, doing this for some time now and. And just being able to learn from each other, doing this together to challenge the status quo and knowing that we have real capability to make experience just exponentially better for human beings. I mean, who would argue, like, this is what thriving looks like? And we are certainly doing this together and having fun with it.
- [51:07]
A
I have had a wonderful time talking with you both. Really, it's been such a fascinating conversation. You're both doing great work, and I appreciate your time. Thank you so much.
- [51:19]
C
Thank you.
- [51:19]
B
Thank you.
- [51:21]
A
Tomorrow's Cure is a production of Mayo Clinic with production help from the podglomerate. Be sure to follow Tomorrow's Cure wherever you get your podcasts. I'm Kathy Werzer. Thank you so much for listening.
- [51:34]
C
Sat.