Transforming Healthcare with AI: Insights and Innovations
Microsoft's Dr. David Rhew recently joined CEO Rajeev Ronanki for an engaging fireside chat. Sharing their valuable insights on trends, opportunities, and challenges of AI in healthcare.
View the completed transcript below or download a copy.
Transforming Healthcare with AI: Insights and Innovations
0:00:03:01 - 00:00:38:10
Steve Ambrose
Hi everyone. Welcome to this highly anticipated fireside chat or the transformative power of artificial intelligence in health care. Now, I'm thrilled to have with us two visionaries who are leading this revolution. Doctor David Rhew and Rajeev Ronanki. Doctor Rhew serves as the global chief medical officer and vice president of health care for Microsoft. Prior to this, he served as chief Medical Officer and Vice president for Samsung and before that, Senior vice President and chief medical officer at Zinke's Health.
00:00:38:12 - 00:01:11:03
Steve Ambrose
Now, in addition to his experience in senior leadership, Doctor Rhew is adjunct professor at Stanford University, and he holds six U.S. technology patents that enable authoring, mapping, and integration of clinical decision support into electronic health records. He's also been recognized as one of the 15 most influential Clinician executives by Modern Healthcare. Joining Doctor Rhew is Rajeev Ronanki, a visionary leader and influential thinker who's at the intersection of healthcare and technology.
00:01:11:05 - 00:01:49:18
Steve Ambrose
For over 25 years. Raj is dedicated his career to reimagining healthcare through artificial intelligence, simplifying the business of care and reducing its complexity. Rajiv's insights have been recognized globally with features at the World Economic Forum in Davos and as a speaker at the inaugural Ted AI conference. Formerly, he was the President of Caroline Digital Platforms. Raj is also the author of the Amazon bestseller you and I a Citizens Guide to AI, blockchain and Puzzling Together the Future of Healthcare.
00:01:49:20 - 00:02:03:06
Steve Ambrose
Welcome to you both. This should be an engaging conversation, tapping into a number of challenges and opportunities in how AI is reshaping healthcare. Gentlemen, it's good to have you both here.
00:02:03:08 - 00:02:05:14
David Rhew
Thanks, Steve.
00:02:05:16 - 00:02:09:15
Raj Ronanki
And thanks for having us. And David, great to see you again.
00:02:09:17 - 00:02:11:19
David Rhew
Yeah, same here as okay.
00:02:11:21 - 00:02:48:09
Steve Ambrose
Let's dive on in. David, we're going to start with you with the launch of the trustworthy and responsible AI network or Train, Microsoft has put itself into this environment of the safe and fair use around artificial intelligence. What I'd like to have you do, David, is tell us a bit more about train. Just chat a bit about its members, its purpose, and then if you could share a little bit deeper about the initiatives and how it looks to improve AI safety and trust through some of the biggest challenges today in health care.
00:02:48:11 - 00:03:10:12
David Rhew
I think we're all excited over the potential that AI has to transform health care to address some of the biggest challenges that we have. But we also know that it can propagate bias. It can change over time. So I'll drift, but it needs to be there. these are just some examples of some of the things that we're recognizing a couple of when you implement artificial intelligence.
00:03:10:12 - 00:03:32:05
David Rhew
And the question that most organizations are asking now is, well, how do we do that? How do we operationalize or how do we implement responsible AI in our own settings? And that is really the foundation and the reason why the trustworthy and responsible AI network has been established. It was established by providers or health care providers that were all thinking about the same thing.
00:03:32:05 - 00:04:00:19
David Rhew
And many of these organizations had already been implementing responsibly AI. They had processes in place. But when you ask them about the challenges that they foresaw, as more and more I was coming through the process, they realized that this was a very manual process. It was very difficult to scale. And so they had recognized that even when organizations had put together teams to do this, they themselves needed to figure out ways that perhaps technology could help them scale this.
00:04:01:01 - 00:04:20:06
David Rhew
And so it really is based on the concept that we need to implement it. We need to do this at scale and hopefully technology and through collaboration, we might be able to do this in a way that's efficient. And ultimately at the end of this. We'd love to be able to ensure that every organization is able to implement responsible AI.
00:04:20:07 - 00:04:45:07
David Rhew
So that means that we should ask ourselves, every organization should ask at least these four basic questions. And the first is, you know, all the artificial intelligence that's running in your system today. You know, is there a way that you can do a inventory, a registration with model cards to really understand what AI is available? And then the second ties a lot into this concept called value assessment and also drift analysis.
00:04:45:07 - 00:05:05:23
David Rhew
Because we want to do is we want to test the algorithms to see if they're working on your own local data sets before and after the implementation. So this is something that's really important because we oftentimes assume that just because I has been tested or has been FDA cleared, that it's going to work in your own setting. Well, that's not necessarily the case.
00:05:06:01 - 00:05:24:16
David Rhew
And so what we recognizing is that the need to test these algorithms before and after and and the reason why I say after is because I can change over time. And it's important to see whether or not it's actually having the results over time. But also we sometimes see unintended consequences. And so we need to start looking beyond that.
00:05:24:16 - 00:05:44:10
David Rhew
So so that's why that's important. And the third question gets into the whole issue that I talked about before, which is bias, because we know that inherently these algorithms, these AI were tested on a certain data set and it was specific for that data set. But when you start looking at whether or not it works in other data sets or other populations, we can't necessarily say that's the case.
00:05:44:16 - 00:06:04:09
David Rhew
And what we oftentimes find is that unless you measure it, unless you assess for it, unless you find ways to mitigate it, we won't be able to have a process that truly addresses the issue around bias. And the last one is really about governance, accountability, understanding that we actually have to do this at scale. And so there has to be a process for doing this and allowing us to be able to address this.
00:06:04:09 - 00:06:19:01
David Rhew
So this is really the foundation for train. Let's take those key principles, those responsible AI principles find a way that we can operationalize it. And let's do it in a way that we can collaborate together and learn from that process.
00:06:19:03 - 00:06:46:11
Raj Ronanki
You know, I think, all the points that David made are just kind of spot on. And it's I'm really glad that Microsoft and Fran are taking a lead on this, because at the end of the day, the whole ecosystem of health care needs to come together to make this work. The providers, the payers, life sciences, pharma and all the technologies that are support players that are in the mix, such as Microsoft and ourselves, that all need to have a part in this.
00:06:46:13 - 00:07:11:03
Raj Ronanki
And so by laying this out and convening the industry as an ecosystem to make this happen around a common framework, along the four principles that that David just laid out. I think that's a key, you know, step towards sort of addressing this, this more holistically. And my hope is that through these efforts that we're able to, but what quote, a quote unquote nutrition label on every algorithm so that we understand its performance current?
00:07:11:08 - 00:07:23:07
Raj Ronanki
Well, how it's been tested, the data sources are the news and as well as its performance in production, to manage drift and other things that David mentioned. So really encouraged by it. And, David, thanks for taking the lead on this.
00:07:23:09 - 00:07:52:09
Steve Ambrose
Let's go ahead and shift gears to health equity. On the topic of health equity, as I becomes more integrated in health care, there's also a risk of exacerbating existing disparities. So what I'd like you to do for the audience, if you could, is to maybe codify and explain this risk, in particular in greater detail. And then how is Lyriq addressing the potential for AI to identify mitigate health inequities?
00:07:52:11 - 00:08:24:19
Raj Ronanki
So I think, it's it's a deep and multifaceted, you know, topic. You know, Steve, you know, some of the, solutions to this lie in technology and as a subset of that, probably with AI. But the broader framing of it is, is one of around equity of care, which is regardless of the, the patient's ethnic and demographic and social status, zip code, etc., that the care is rendered based on sort of the personal health history and, the related sectors that doctors invest.
00:08:24:21 - 00:08:45:19
Raj Ronanki
And it's long been known that, there are differences in that care based on where we live. And, you know, other factors. So one of the risks is in that in automating and, and sort of scaling that have internet speed and internet scale and that we risk perpetuating some of those inequities that are inherent, you know, within the system.
00:08:45:19 - 00:09:26:04
Raj Ronanki
So we have to take a deliberate pause and redirect that and make sure that we're not scaling something that, you know, we know it to be, you know, as, as certain issues. So that I think is a is kind of a broader topic, to which then, Lyric's role isn't necessarily to solve that broader problem. but we do have, I think, a role to play in making sure that the care that is delivered, conforms to all of the medical necessity guidelines, and the various policies and regulations across the system, which both providers and payers agree to us as being the standards by which, you know, we should be arbitrating that.
00:09:26:06 - 00:10:09:15
Raj Ronanki
So as an example, through the work that that lyric does, we're on track to save the system collectively, you know, something in the range of $20 billion this year. so with that cost avoided and cost saved, now, providers and health plans together can determine, where that should be redirected to those savings, better access to care, for example, investing in access to care in rural areas where in some cases people have to drive 8000 miles to get to a medical facility and building more facilities where where there's lack of access and also investing more resources and making sure that, the care that is delivered, is, the best
00:10:09:15 - 00:10:16:01
Raj Ronanki
standard of care across the system, not just based on how you live and whether you go to Stanford. You know, for example.
00:10:16:02 - 00:10:36:05
Steve Ambrose
David, turning to you, you've had a lot of involvement with all types of stakeholders and organizations when it when it comes to improving health equity. So should I be considered crucial to effectively address health disparities, particularly when we talk about the underserved or the rural communities?
00:10:36:07 - 00:11:04:07
David Rhew
Yeah, absolutely. you know, one of the things that we've recognized is that if we want to democratize artificial intelligence, we have to start with infrastructure. And what I mean by infrastructure are some of the digital infrastructure that's necessary for individuals in underserved rural communities across the globe. And that is maybe a starting point. And one of the things that we recognize is that, of access to affordable broadband is, is a is an important element.
00:11:04:09 - 00:11:29:18
David Rhew
In fact, Microsoft, initiated a global initiative called urban in, in already, 51 million people globally have been able to gain access to affordable broadband. Now, something that has happened here in the United States is sort of that next step, access to cloud, and in fact, leveraging a cloud to be able to solve some of the more critical vulnerabilities that we see it today, which is particularly problematic in, smaller resource settings.
00:11:29:18 - 00:11:58:15
David Rhew
So, for instance, cyber security. So, in in June 20th, during 2024, Microsoft announced, with administration of the white House administration, the Microsoft cybersecurity program for rural hospitals, in which case, these organizations can access free and low cost security services and solutions through the cloud. specifically, these rural hospitals. So, so this is, a great first step to providing some of that infrastructure.
00:11:58:17 - 00:12:22:01
David Rhew
But what we also realized is that with that in place, we actually have to have I apply and do things that allow us to be able to increase access to affordable healthcare to noninvasive, low cost, highly accessible tools. I'll give you an example of that. So, one of the things that we've been recognizing is that there are many different ways that we can take a look within our own bodies.
00:12:22:01 - 00:12:47:06
David Rhew
And one of those that's extreme efficient at doing this is through our AI. So I'm sure many of us, if not all of us, have had an AI exam, you know, so you remember that machine where you put your chin on there and you say, is it better or worse? well, that machine captures the data. And traditionally that data has been, just sort of like, disregard it apart from just being able to to tell you whether or not what your vision was.
00:12:47:08 - 00:13:10:08
David Rhew
Well, now we actually have the ability to take that image with the patient's permission through the provider, run AI on top of it and screen for conditions like diabetic retinopathy. We can also screen for potentially cardiovascular disease, chronic kidney disease, even Alzheimer's. And the reason is because the images are taken at the micron level for both the arterials as well as the nerves.
00:13:10:14 - 00:13:30:08
David Rhew
And we can also see the surrounding structures. And with AI, we can determine whether or not we have certain diseases. And so this is actually being done. So I'll give you an example. Like at Stanford for instance, they have a Maine Medical center. And then what they have are multiple different satellite hospitals, I should say clinics, throughout the entire North, Bay area.
00:13:30:10 - 00:13:49:08
David Rhew
And, and many of these are in kind of rural areas or areas that are maybe less, populated. And what we found is that when a person comes in for their routine visit, they will in addition to getting their their vital signs shed, they'll get their eyes examined, and then they'll get a screening with AI for diabetic retinopathy.
00:13:49:08 - 00:14:12:22
David Rhew
And they're picking up a ton of retinopathy. And it's largely because these are places where individuals that normally don't go see the physician for conditions like diabetes or they don't, maybe they have diabetes and they haven't gone for the routine ophthalmology visit. In fact, the highest score is, traditionally the lowest he'd score traditionally is screening for diabetic retinopathy.
00:14:13:00 - 00:14:31:05
David Rhew
Not because the clinicians don't order it. It's because it requires the individual to actually go see the ophthalmologist. Well, this allows you to be able to change that paradigm and dramatically improve it. And on top of that, it's something that has a CPT code so the clinicians can get paid. you've got an opportunity for recruit quality of care.
00:14:31:06 - 00:14:51:07
David Rhew
And most importantly, we're increasing access to affordable health care through a large screening of populations that we normally could never have done before. And so it's he has given us the opportunity to do this. And if we put the infrastructure in place, if we apply the technology the right way, we can have a dramatic impact, in particular on those that are at underserved rural communities.
00:14:51:09 - 00:15:03:15
Steve Ambrose
David, with the federal government plan to regulate AI in health care. What role do you believe that public private partnerships should play in shaping these regulations?
00:15:03:17 - 00:15:25:01
David Rhew
Yeah, I think one of the first things that we have to recognize is that from a regulatory and policy standpoint, we often kind of start looking at risk. And risk is, the lens that everything's looked at. But what I'd love to be able to have is a conversation about not only the risk of doing AI well, the risk of not doing AI.
00:15:25:03 - 00:15:56:11
David Rhew
And that is something that we have to always balance. Because when you look through the risk lens, everything looks risky. You know, there's always these reasons why that we shouldn't be doing it or we need to overregulate this. But if you look at it in terms of the opportunities, it's very important for us to know what is the baseline rate of, whatever type of process we're looking at, because it turns out that the quality of care and the level of patient safety in many cases is really not not at a level that is acceptable.
00:15:56:13 - 00:16:13:19
David Rhew
And so so we actually have to always look at it a risk benefit ratio. And I think that's really important for us to start thinking about. So that would be the first step. Now the second is that we oftentimes lump all AI together. And because of that we come up with regulations and policies that are very confusing and sometimes inappropriate.
00:16:13:21 - 00:16:34:07
David Rhew
And I think it's important for us to recognize that we have been doing AI, traditional AI ML for quite some time, you know, since the 1950s. And
that has evolved from being an algorithm to doing things that are extraordinary, for instance, like being able to convert natural language processing through natural language processing voice, text, images into things that are understandable.
00:16:34:11 - 00:16:56:06
David Rhew
I mean that that is extraordinary, and that is something that is also highly reliable. But we also see that with generative AI, we can do things that are highly creative and that are able to take administrative loads and convert them into things that are automated. And that is a very important as well. But we also have to recognize that there's two different use cases that are evolving from this.
00:16:56:06 - 00:17:16:06
David Rhew
And it's important to know that with these different types of AI, we have different capabilities but also different limitations. And the third is that we have to establish a standardized approach for evaluating the impact that AI has. It's not just about the technology, but it's about what we're measuring, how we're measuring, how we're implementing it. So I'll give you an example.
00:17:16:10 - 00:17:39:07
David Rhew
Like back in the day, we used to always look at electronic health records as being the savior. That would save all of us, because it would create such a digital environment that we could apply technology and we could streamline processes. And there was early studies that showed that the implementation of computerized provider order entry, or CPO, decreased mortality.
00:17:39:09 - 00:18:01:21
David Rhew
And then we saw studies that came out that it showed that it increased mortality. These things stacked same system. Well, how is it that one system could decrease mortality? It could another time it could actually increase mortality? It's because the we didn't look at what the implications of how they implemented it. And that's so important. We actually have to compare apples with apples.
00:18:01:22 - 00:18:20:22
David Rhew
If we had organizations that took the same type of technology, took that same type of methodology for measurement and took the same type of approach for implementation, and they got results. We could actually start comparing and determining whether or not these would work. And that is an essential part of how we need to think about how we move forward together.
00:18:21:01 - 00:18:28:14
David Rhew
It's not just about the technology, it's about all the other things around it that allow us to be able to determine is it working or is it not?
00:18:28:16 - 00:18:48:07
Steve Ambrose
David, you hold several patents related to clinical decision support systems, so you'd be best to ask this to how do you envision the future of AI driven clinical decision support evolving, especially around, aspects of significant transformation and change?
00:18:48:09 - 00:19:14:21
David Rhew
Well, one of the things that we're recognizing with regards to clinical decision support is that historically, it has been for getting the right information at the right time for a provider. And that has been the paradigm for what we think of CCDs. But with AI, we can extend that paradigm to start thinking about a new world in which clinical decision support or providing the right information at the right time is for everybody, is for the extended team.
00:19:14:21 - 00:19:35:23
David Rhew
It's for individuals that we don't oftentimes think of as being part of the team. Yeah, we're talking about receptionists, call center agents, case managers, as well as to patients and families. So let me give you an example. Let's imagine we we are a patient making a phone call. And we're calling that and there's a call center agent that's then picked up, the call.
00:19:35:23 - 00:19:56:09
David Rhew
And then they send it over to the receptionist and they said, oh, hey, you'd like to schedule an appointment. Let me look. Let me look at the available schedule. Oh, here's the first available. Done. Let's move on to the next call. Well, let's imagine if when you made that phone call, it ran through the database and it found that you were someone who had actually missed your last three appointments.
00:19:56:11 - 00:20:22:17
David Rhew
You live 60 miles away and you don't have a car. Well, and you're eligible for the rideshare program. Well, at that moment, a decision support mechanism comes up and presents to you the option that the individual here should be encouraged to take advantage of this program and then not only sign them up at that moment, but make sure that we can get the, cell phone number so we can they can get the proper texting prior to that and that they can ensure that they actually get the ride.
00:20:22:19 - 00:20:46:15
David Rhew
I mean, these are the types of things they're missed opportunities and largely is because we assume that people just know that the programs exist or is done through the provider. It actually is best done through
every interaction, every opportunity that we have with the individual. And it's not always the health care provider that the health care provider interaction is so limited, so infrequent that if we relied on that, we would miss many opportunities.
00:20:46:17 - 00:20:51:14
David Rhew
They're more likely to actually interface with other members of the extended care team.
00:20:51:16 - 00:21:09:01
Steve Ambrose
Quick follow up to that at David. When you talk about extended care team, would you also put in, pharmacists as well as maybe health plans, particularly health plans, as they're focusing more on becoming true health organizations?
00:21:09:03 - 00:21:29:15
David Rhew
I mean, you could some people could consider them part of the core team. Some people could consider part of the extended team. Absolutely. I mean, we're talking about everyone that has a a role, a stake, a stakeholder role in the care of the individual. And we oftentimes find that there are interactions that are missed opportunities for individuals to be able to take advantage of it.
00:21:29:17 - 00:21:51:18
David Rhew
The role of the pharmacist is greatly undervalued, viewed because we often tend to think of them as simply being responsible for the drugs. But in fact, the interaction, the opportunity for the pharmacist to have a conversation with the patient to talk about so many other things and identify issues and bring those to light and manage programs is extremely valuable.
00:21:51:18 - 00:22:13:15
David Rhew
And that is why we do. We really need to think about how we redistribute tasks and load across a broader team and not think that it's only done by one individual. But sometimes what you can do is with AI, you can practice at up at the top of your license, and sometimes even beyond the top of your license in a way that is liable, secure and something that is trustworthy.
00:22:13:15 - 00:22:40:06
David Rhew
So I feel like we have an opportunity with the AI to rethink what the current process of how we deliver care. Because the current care system is not designed for scale. It has a lot of bottlenecks. And largely those bottlenecks are people. And if we can figure out ways that we can democratize this knowledge across larger groups so that people can access it using these technology in a way that's understandable and actionable, we could actually change care.
00:22:40:08 - 00:23:05:17
Raj Ronanki
But I think, that that was a nice, you know, an awesome summary and some fantastic insights there. It brings to mind this indelible image that we have of doctors with these data scopes, ultimately sort of measuring lung function and heart function. But imagine that that sort of scope is replaced by AI. We have so much more knowledge and information about not just the patient, but everything else that's happening across the industry.
00:23:05:19 - 00:23:14:14
Raj Ronanki
New developments, new science, latest therapies, all that goes fingertips. That ultimately is the potential of AI and and the practice of medicine.
00:23:14:16 - 00:23:41:12
Steve Ambrose
It's almost like with so many of the things, David, that you and Rajat both mentioned here, what comes to my mind is it's really this is really a tremendous gap filler, but it's also getting doctors and patients to question, basic understanding of what health is and what health isn't. as you mentioned, you could feel great and have no symptoms, but you could be developing chronic disease and not even know it.
00:23:41:14 - 00:24:14:17
David Rhew
That's right. And it's impossible for a physician in a very limited time period to be able to make those diagnoses. In fact, it is not something that we should be expecting of a physician to do that. It should be done continuously through technologies, where people have that are highly accessible. These are things that have to be, you know, relatively inexpensive, you know, affordable, but also something that is, is something that would provide value that that actually shows that it works in particular context.
00:24:14:22 - 00:24:32:06
David Rhew
So screening for diabetic retinopathy with, you know, that through the eye. But we also see that we have tremendous number of individuals that just routinely get films. Like for instance, if you think about an urgent care center, all the individuals that come in, they have a cough, they get a chest x ray. While working with one of the health glands.
00:24:32:06 - 00:24:57:00
David Rhew
We found that 62% of the time with those films that are routinely taken, that they're missing other three key elements. So for instance, osteoporosis and fractures. And so they found that in the Medicare population, about 1,315% of individuals have fractures. That's a that's a pretty extraordinary amount. So you've got individuals that have fractures. And 62% of the time you're missing them.
00:24:57:00 - 00:25:19:22
David Rhew
Are these felt that that's an opportunity. So in fact this the system this health plan they went and they did an intervention with education exercise programs calcium bisphosphonates. And they actually showed clinical and financial benefits for doing this. We have an opportunity to start thinking about all the things that we do today, the images that we take, the voice conversations that we have with patients.
00:25:19:23 - 00:25:40:11
David Rhew
You can actually take a person's voice and run biometrics and screen for depression and anxiety. You know, we have an opportunity to start thinking about how I can be applied broadly to be able to help identify individuals before they even know it or before they know that is progressed. And that's an extraordinary opportunity. I call it opportunistic screening or pre screening.
00:25:40:17 - 00:25:48:22
David Rhew
These are all ways that we can think of how I can dramatically change the paradigm of how we improve care for populations.
00:25:49:00 - 00:26:08:14
Steve Ambrose
Raj, we mentioned before that Lyric's focus and vision is on simplifying the business of care. that said, how do you envision I transforming the relationship between providers and health plans, as well as going into the experience for the plan? Member.
00:26:08:16 - 00:26:33:11
Raj Ronanki
You so the, you know, I think a predominant view that both payers and our clients have is that, both are engaged in an arms race to, to get the transactions ultimately codified accurately. You know, so we've got applications of AI and revenue integrity and revenue cycle management on the provider side. And then you've got investments in payment integrity and payment accuracy and program integrity.
00:26:33:11 - 00:27:05:02
Raj Ronanki
And on the other side, and ultimately, you know, I think there's a commonly used, statistic that 80% of the providers have good intentions and are quality doctors that are taking great, you know, care of their patients. And so the investments in these technologies on revenue cycle, on revenue integrity, is, is really perhaps misplaced and comes because of a lack of trust and lack of collaboration between providers and payers.
00:27:05:04 - 00:27:28:15
Raj Ronanki
So I think if we can improve that, improve the trust levels, improve transparency and make it easier for the to the business across those two critical stakeholders in our system, that we could eliminate the the focus on the vast majority of the providers in the country that are doing
the right things, you know, for their patients and are simply, battling bad technology.
00:27:28:17 - 00:28:07:00
Raj Ronanki
the lack of information, sort of the business part of the care, which is a financial settlement piece of it. So if lyric can play a role in creating more transparency and more trust between the two parties and enable the vast amount of of providers in our country that are good actors to be really focused on, patient care and increase, you know, the time that they spend with their patient, which is increasingly a challenge with all the technology constraints that we're replacing for them to free them up and say, rather than spending the average of 15 minutes or so of the patient, you know, spend 30 and, and, and so
00:28:07:00 - 00:28:46:02
Raj Ronanki
doing improve the quality of care and the types of, conditions you might be able to diagnose and get after all the undiagnosed and misdiagnosed conditions, like not detecting the factors that, that David was talking about as an example. So those are all the things that we can enable payers and providers to do more effectively by eliminating all of this noise around what's accurate, what's necessary, does it conform to all the rules and regulations, does it comply, etc., etc. and automate all that with a great amount of pedigree advancements using technologies like AI as well as other technologies that are deterministic, that have been around in our software for a long, long time.
00:28:46:04 - 00:29:15:21
Raj Ronanki
So that's where, you know, we're in the middle of doing. And then what that as a byproduct of all of that, then we can, you know, as a system, focus on a small subset, that are the bad actors and, you know, systematically minimize their ability to, you know, to, to, to practice in that way, and make sure that our attention and our, our fraud, waste and abuse analytics are all sort of oriented towards a small percentage of the population that's practicing that.
00:29:16:00 - 00:29:41:01
Raj Ronanki
And for the vast majority of providers and, and payers, make sure that the transactions are settled, in real time, as accurately as possible. And then if we did that, then the patients that unfortunately sometimes have to deal with, prior us that get delayed or they have to wait on scheduling their care out because providers and payers have to haggle over what's appropriate and what's necessary.
00:29:41:03 - 00:30:05:20
Raj Ronanki
All that could be streamline and, and and right after, you know, Cara's, patients can walk away with, with all the information about what they're going to be on the hook to, you know, look for, financially and take care of it rather than getting, meaningless explanation about a certain explanation and payment statements for 6 to 9 months following, you know,
care being delivered and then not having any idea, you know, what they're on the hook for.
00:30:05:20 - 00:30:12:22
Raj Ronanki
So all that could be greatly improved and simplified, you know, to the solutions that that lyric is focused on creating.
00:30:13:00 - 00:30:35:13
David Rhew
Yeah, I want to build off of that because what lyric is doing is so important. This is, a major opportunity for us to apply technology to improve transparency, but also to serve as an arbiter to that, a neutral arbiter that could potentially allow organizations to say that this is an appropriate or not appropriate mechanism for us to be able to address this particular issue.
00:30:35:18 - 00:30:59:07
David Rhew
And if we could do apply these approaches to be able to remove a lot of that administrative burden and the waste and the time delays, that will improve care for everybody. And that is a fantastic objective. And and I'm very hopeful that rise and lyric can be successful in this endeavor because, getting gaining trust is a very, very tough challenge.
00:30:59:09 - 00:31:03:21
David Rhew
And it's going to require all the tools in your toolbelt to be able to help do that.
00:31:03:23 - 00:31:15:22
Steve Ambrose
David, what measures is Microsoft implementing to ensure data privacy and security, especially when it comes to generative AI in sensitive patient information?
00:31:16:00 - 00:31:37:13
David Rhew
Yeah. So I guess one of the first things that organizations need to know is that if you go to the internet and you start, entering information in through a ChatGPT engine, that is something that, in some cases will just go out to the public domain. Not now. There are certain mechanisms that you can, go that, allow you to ensure that that doesn't happen.
00:31:37:15 - 00:32:12:15
David Rhew
But if you're thinking about a business, and how you want to leverage this GPT type engine, typically, what the better approach to do, the more reliable approach would be to create your own secure enclave, you know, your own cloud environment that allows you to be able to then take the models and write it within that off. I mean, what the nice thing about doing that is that it actually ensures that neither the data that you enter in through the grounding, neither the, prompts that you put in,
none of the information that you actually enter in this or the user's entry will ever get exposed.
00:32:12:16 - 00:32:40:09
David Rhew
And in fact, it doesn't even come to Microsoft or it doesn't go to OpenAI or any organization. In fact, all of that is within your own local environment, your own data sets. So from a data privacy security that is the way that enterprise organizations need to start thinking about how they implement it. The public domain is a completely separate issue, and that is one that we really do need to recognize that even if you don't, think that you'd like to, readily adopt this, many of your employees in your patients probably will.
00:32:40:11 - 00:33:08:19
David Rhew
And so it's good to create a mechanism so that you can actually enable organizations and individuals to feel comfortable doing this in a way that is privacy preserving and, allowing you to have data security. Now, there's one other thing that I want to highlight as well, and that is that if we think about AI in general, that we always recognize that risk is something that is kind of now throughout we we have to find ways that we need to better manage risk.
00:33:09:00 - 00:33:29:02
David Rhew
Cybersecurity threats are a high concern for a lot of folks, and the best way to do that is you need to be able to have access. You need to control the data and and not and then on the portals of entry as well, and not necessarily leave these open to to others to enter. Well, it sounds logical, but but we do this all the time.
00:33:29:07 - 00:33:59:06
David Rhew
For instance, when we collaborate, we typically share data. So a good example would be that the NIH that the data management and sharing policy that was put into effect January of 2023, it actually promotes the sharing of scientific data, which is fantastic. But at the same time, information like your omics data, your imaging data, your data is now going in a de-identified way, is going out to, another organization or another place that you don't necessarily have control over.
00:33:59:08 - 00:34:31:09
David Rhew
And that's something that we need to start rethinking the approach for how we do data collaboration without data sharing. And Raj touched on this. But federated learning, federated processes where data stays in your own premises is essential. And it's this a lot of organizations are now starting to move toward federated process, but that's only good for one part of the, the, the the issue, the, the data stewards will be happy with federated because the data never leaves your premises.
00:34:31:11 - 00:34:52:21
David Rhew
But the algorithm developers are unlikely to want to send their algorithms to all the different sites because they'll it'll expose the algorithms. And so we actually need to have privacy preserving strategies to ensure privacy for not only the data, but also for the algorithm. And to do that, then we can truly collaborate in a way that's privacy preserving.
00:34:53:01 - 00:35:19:23
David Rhew
And what we found is that you can do encryption. And, there's obviously there's encryption, you know, at rest and encryption in transit, but there's also encryption in use. And what that is, is, is allowing you to be able to put the data and the algorithms directly into the microchip, run the analysis and put out the different outputs, but neither the data nor the algorithm ever get exposed because it's all done within the microchip microchips, the technology enforced nepotism.
00:35:20:01 - 00:35:47:19
David Rhew
And so these type of strategies federated learning, privacy preserving, secure enclaves allow us to be able to think of a new world in which we can actually do data collaboration without data sharing. And that will be essentially the Gulf War basis. They think about how to advance AI, because ultimately AI has to be tested in multiple different environments and can't expect everyone to move all their data in one environment and expect all algorithm developers to share, or expose their IP on their AI.
00:35:47:19 - 00:35:51:23
David Rhew
So we have to think about this in the context of both.
00:35:52:01 - 00:36:29:05
Raj Ronanki
And I think in some ways, you know, Clay Christensen is in his book on the innovator's dilemma, like, decide as to why it's so difficult for, for big organizations like Microsoft to innovate. And there's lots of reasons for it. but I think, and the team at Microsoft that really shown the path that big companies can in fact, innovate and do it effectively, but it's going to take, an open and a humble approach where you have to make investments in the ecosystem and sort of take this approach that you have to move the collective, you know, ecosystem in a way that's meaningful, impactful, while laying in all the
00:36:29:05 - 00:36:43:06
Raj Ronanki
safeguards and, protective mechanisms to, to do it safely. so really, kudos to you and the team on doing that. And, you know, for us, as users of it, we benefit a great deal from the work that you do, gentlemen.
00:36:43:06 - 00:36:54:00
Steve Ambrose
And engaging conversation. It didn't disappoint. And, I really appreciate you both taking your time to, to, be here today and to share your thoughts.
00:36:54:02 - 00:36:54:14
David Rhew
Thank you very.
00:36:54:14 - 00:36:57:01
Raj Ronanki
Much. Thanks, David. I really enjoyed it.
00:36:57:03 - 00:37:27:23
Steve Ambrose
Our thanks to Doctor David Rhew and Doctor Rajeev Ronanki for sharing their enlightening perspectives and valuable insights. In this fireside chat, you can learn more about Lyriq and its solutions by visiting Lyriq dot AI, and more about Microsoft and all they're doing in health care by visiting microsoft.com. And thanks to you, our viewers will continue to explore these and other crucial topics, opportunities, and trends as we collectively shape the future of health care.
00:37:28:01 - 00:37:29:08
Steve Ambrose
Have a great rest of your day.
