Latest / HumanifAI HealthTech / What Responsible Healthcare AI Actually Looks Like w/ Dr. David Rhew, GCMO, Microsoft
Transcript
- 0:00Francesca: Well, hello.
- 0:01Francesca: AI in healthcare is full of promises and few people can tell you how to keep them.
- 0:07Francesca: Today a leader from a company at the very forefront of the AI boom is here to show us what responsible human centered AI actually looks like in practice.
- 0:17Francesca: And that company is none other than Microsoft.
- 0:28Francesca: Welcome back to HumanifAI Health Tech, where we explore how AI is rewriting the rules of healthcare and putting people right at the center where they belong.
- 0:37Francesca: I am your host, Francesca Vera, head of marketing at Net Health, and today we are thrilled to welcome Dr. David Rhew, Microsoft's global Chief Medical Officer and Vice President of Healthcare.
- 0:50Francesca: Now, Dr. Rhew is leading some of the most important work happening at the intersection of medicine and technology, specifically how we responsibly bring AI into patient care.
- 1:02Francesca: He's a physician, a technologist, and an inventor with six US patents, and he shared his insight everywhere from HIMS Global Health Conference to the Vatican.
- 1:12Francesca: We're especially excited to have him here because he'll be opening our keynote at net at Net Health.
- 1:19Francesca: Next, let me take that again.
- 1:20Francesca: That last little piece.
- 1:23Francesca: He's a physician, technologist and inventor with six US patents, and he shared his insights everywhere from the HIMSS Global Health Conference to the Vatican.
- 1:33Francesca: We are especially excited to have him here today because he'll be our opening keynote speaker at Net Health's next 2026 in April, where he'll be talking about AI for better health and how human centered AI is transforming specialty care.
- 1:49Francesca: Dr. Rhew.
- 1:50Francesca: Welcome to the podcast.
- 1:53Dr. David Rhew: Thank you, Francesca.
- 1:54Dr. David Rhew: I'm excited to be here.
- 1:56Francesca: That is so good.
- 1:57Francesca: I'm, I couldn't be more excited to have you.
- 1:59Francesca: Um, as I mentioned, you've worn many hats, infectious disease, physician researcher, and tech executive.
- 2:06Francesca: So when you're advising healthcare systems on AI integration today, what patient moments from your clinical days still shape how you think about implementing technology?
- 2:17Dr. David Rhew: Great question.
- 2:18Dr. David Rhew: You know, I, throughout my career I have oftentimes been very excited over technology.
- 2:23Dr. David Rhew: Uh, but at times I've also been frustrated with the technology.
- 2:27Dr. David Rhew: And I think the key lesson for me as a CLA clinician and as a researcher, has been to really appreciate the clinical workflows.
- 2:35Dr. David Rhew: If we are trying to enable improved care, you need to have individuals using the technology.
- 2:41Dr. David Rhew: There's no value in having technology sit on the shelf.
- 2:44Dr. David Rhew: And, uh, understanding those workflows is essential.
- 2:47Dr. David Rhew: And oftentimes we think that sometimes the technology will solve one problem, but it's really the workflows that enable us.
- 2:52Dr. David Rhew: So let me give you an example.
- 2:54Dr. David Rhew: Uh, you know, early in my career, I was fascinated by the fact that we had so much variation in care, and I was curious why there was so much variation.
- 3:02Dr. David Rhew: Uh, a lot of it was actually unwarranted and I started thinking about like, could, is there a way that we could actually improve the standardization of care or specifically around what we'll call evidence-based, uh, practices.
- 3:13Dr. David Rhew: And so I, I built this product, it was called an evidence-based order set.
- 3:17Dr. David Rhew: And, uh, ultimately this is a product that we found a way to integrate into electronic health record.
- 3:23Dr. David Rhew: And this was something that was.
- 3:25Dr. David Rhew: Being launched throughout the United States, specifically as electronic health records were starting to become more popular.
- 3:31Dr. David Rhew: And as a researcher, I was extremely excited because the evidence-based order sets were shown to, uh, not only reduce the variation, but also decrease mortality, uh, and improve overall efficiency.
- 3:42Dr. David Rhew: So there was a lot of benefits from using these well.
- 3:45Dr. David Rhew: What turns out hap is that it was widely adopted, not because of the clinical benefits, but it was adopted because it saved the clinicians 95% of the time, required to enter orders into the electronic health record.
- 3:58Dr. David Rhew: And, and so like appreciating that allowed us to be able to then get to the point where we had widespread adoption and ultimately then we saw improved quality of care.
- 4:07Dr. David Rhew: So a lot of times we think that it's actually the quality of care that's gonna drive the behavior change or the improved safety.
- 4:14Dr. David Rhew: Actually, it was the improved clinical workflows.
- 4:16Dr. David Rhew: And I think for me, that was the one thing that has continued to stick with me.
- 4:20Dr. David Rhew: How do we ensure that no matter what we put in front of clinicians, that it actually is workflow compatible?
- 4:27Francesca: You know, it's so interesting you mention that because one of the things that we keep touching on, um, is this notion of the myth of AI reality versus the, you know, the true reality of what it's delivering.
- 4:38Francesca: And the fact that you're talking about like workflows and saving time is such an interesting point because you've seen these medical AI ascribes, um, deployed across hundreds of healthcare organizations and settings through the partnerships with Microsoft.
- 4:52Francesca: And you know, here specifically at Net Health.
- 4:55Francesca: We are hearing everything from, as you've just mentioned, it saves me two hours in pajama time a day to it creates, you know, more work fixing those errors, right?
- 5:03Francesca: So there's a little bit of a dichotomy there as well.
- 5:06Francesca: So what separates successful ambient AI implementations from those disasters?
- 5:12Dr. David Rhew: Yeah, so I think the first thing that we all need to recognize is that the AI continues to get better.
- 5:16Dr. David Rhew: And so the earlier evaluations of scribes and a lot of the technologies that we saw first deployed have gotten to the point where a lot of those issues that we've seen in the past are no longer present.
- 5:28Dr. David Rhew: So staying on top of things is probably maybe the first thing that's important, but also the thing that is really important is to.
- 5:35Dr. David Rhew: Understand who is using it.
- 5:36Dr. David Rhew: Now, there are some clinicians that they're not necessarily, uh, super busy or burdened with the task of note writing or maybe their notes are pretty simple and it's just, for them, it's a very straightforward task.
- 5:50Dr. David Rhew: There's probably not a huge amount of benefit from ambient in those scenarios.
- 5:54Dr. David Rhew: But if you have somebody who is working in a very busy practice, uh, they're spending a lot of time writing notes, they're burned out, they're frustrated.
- 6:02Dr. David Rhew: Then in that scenario, ambient may be actually extremely valuable.
- 6:06Dr. David Rhew: And so I think understanding who we're seeing, and this also plays out because we see a mixture in all practices and settings.
- 6:13Dr. David Rhew: And sometimes it turns out that individuals that have marginal benefit are being their, their impact is being measured with those that have significant.
- 6:20Dr. David Rhew: And so I think it's really important to start recognizing that the importance is not on average did it improve things.
- 6:27Dr. David Rhew: But if you talk about the individual physicians.
- 6:30Dr. David Rhew: How many of them actually feel like their lives have gotten better and that the burnout has decreased and their satisfaction has improved, and also the patients, if their satisfaction has improved.
- 6:41Dr. David Rhew: And those metrics are extremely IPO impactful because we're starting to see with the newer AI ambient technologies that we're finding ways that we can actually reduce the burnout and honestly without, you know, by reducing the burnout, it's such an important issue.
- 6:57Dr. David Rhew: Because without clinicians, we really don't have healthcare, and so we have to think about ways to be able to address this, and ambient has become an extremely valuable tool.
- 7:06Francesca: I think you, it's, I mean, it's such a, a spot on point.
- 7:09Francesca: I mean, at the core of healthcare is really trust, right?
- 7:12Francesca: You're talking about how the clinicians themselves trust the technology enough to use it.
- 7:18Francesca: You're talking about certain workflows and some of their trust linkage there as well.
- 7:22Francesca: Um, and you actually helped train, uh, sorry, launch Train, which really stands for trustworthy and responsible AI network, and you did so with, uh, was it about 16 or more healthcare systems specifically to operationalize responsible AI principles.
- 7:38Francesca: Um, if I talk about from a net health perspective, we've recently introduced some comprehensive data and AI governance plans that really emphasize things like GDPR, HIPAA and ONC DS DSI certification.
- 7:52Francesca: But when you talk about that and the linkage between people using technology, right, and that compliance and that sort of adoption of it, what are the non-negotiable elements of an AI governance framework when you think about trust underpinning that?
- 8:06Dr. David Rhew: Yeah, so the trustworthy, responsible AI network was launched in response to the fact that many organizations were saying, Hey, you know, we, we think responsible AI is important, but, but how can we do this in an efficient manner?
- 8:18Dr. David Rhew: How do we do this with limited resources?
- 8:20Dr. David Rhew: 'cause there's just too much ai.
- 8:22Dr. David Rhew: So that was launched over a year and a half ago, and right now we have over 70, uh, health systems, maybe close to 80 across two continents.
- 8:30Dr. David Rhew: So, so it's, it's really expanded and largely because of the fact that we're all recognized the importance of this and to the point of like, what are the non-negotiable elements?
- 8:39Dr. David Rhew: I mean, let's start with just the basics in terms of privacy, security, regulatory compliance.
- 8:44Dr. David Rhew: Those are absolutely non-negotiable.
- 8:47Dr. David Rhew: We have to make sure that those are in place.
- 8:49Dr. David Rhew: But on top of that, we're starting to recognize that other non-negotiable pieces are we really do have to manage the ai.
- 8:56Dr. David Rhew: We cannot just simply let it, you know.
- 8:58Dr. David Rhew: It's run in our systems and just hope that it, it provides good, better care.
- 9:02Dr. David Rhew: Uh, so in order to do that, we have to ask five basic questions.
- 9:06Dr. David Rhew: Do we know what type of AI is running in our system today?
- 9:09Dr. David Rhew: So that's like the first question.
- 9:11Dr. David Rhew: The second is, have we tested the AI on local dataset or datasets that are diverse and ultimately representative of your population?
- 9:20Dr. David Rhew: And a corollary to that is have you monitored them post-deployment?
- 9:25Dr. David Rhew: And then we have to ask the question, have we looked for bias and bias in the models and biases, how it's being deployed?
- 9:31Dr. David Rhew: Bias in terms of impact in subpopulations.
- 9:33Dr. David Rhew: So we need to think about that as well.
- 9:35Dr. David Rhew: We need to start thinking about how have we developed a process for governing all of this.
- 9:40Dr. David Rhew: And, and ultimately the governance process is extremely pro important, but it also is time consuming and we have to think about how to do this all in a time and resource, cost efficient manner.
- 9:50Dr. David Rhew: So these are the types of things that we have to ask ourselves.
- 9:53Dr. David Rhew: If we don't answer those questions in a affirmative, then what we're doing is we are creating gaps of potential opportunities where the AI may drift, uh, may do things that we were, are kind of unexpected, and we ultimately will find that it doesn't deliver on the value.
- 10:10Francesca: You know, you brought up two great points.
- 10:11Francesca: One, I wanna come back and talk about bias in a second, but you also touched on the fact that you've expanded this over global, um, sort of healthcare environments.
- 10:21Francesca: You also spoke a little bit about different sort of community care, and I think you touched on the even sort of the rural concept.
- 10:27Francesca: Right?
- 10:27Francesca: So train specifically collaborates with OCHIN and Trubridge to ensure that organizations, regardless of their resources, can access AI benefits and avoid perpetuating that bias.
- 10:38Francesca: Um, but many of net health's, uh, environments as well.
- 10:41Francesca: The, the, we serve many clients in the rural and sort of community-based populations, but as we think about AI adoption, accelerating, as you touched on a second ago, what are those biggest threats to healthcare equity?
- 10:54Francesca: And what guardrails do we need to make sure we put in place to ensure innovation reduces, but doesn't reinforce some of those disparities?
- 11:02Francesca: And then we'll expand on bias in a, in a second.
- 11:05Dr. David Rhew: Yeah, that, that's a great question.
- 11:07Dr. David Rhew: And you know, obviously technology can do a lot, a lot of positive things can be derived from technology, but it, it can't if you don't have the basic infrastructure to run it.
- 11:16Dr. David Rhew: Uh, I mean, we're talking about things like access to broadband, uh, cloud infrastructure, cyber productions, uh, the resources to be able to manage this.
- 11:25Dr. David Rhew: And so to help with that, Microsoft, even well before AI was.
- 11:30Dr. David Rhew: Sort of being talked about in a regular basis.
- 11:32Dr. David Rhew: We had recognized that just digital technology infrastructure is essential, and we wanted to ensure that organizations and individuals across the globe had access to affordable broadband.
- 11:43Dr. David Rhew: We launched an initiative called Airband, which allowed us to be able to.
- 11:46Dr. David Rhew: Expand the capabilities and largely by working in collaboration through public-private partnerships to be able to offer discounted rates of these technologies, uh, for individuals across different areas.
- 11:57Dr. David Rhew: And we've seen millions of individuals now have access to act to affordable broadband.
- 12:03Dr. David Rhew: And more recently, with specifically emphasis on rural health, we launched something called the Rural Health Resiliency Program.
- 12:10Dr. David Rhew: Uh, and so as part of that, uh, we are specifically looking at ways that we can enable rural hospitals to be able to get access to.
- 12:17Dr. David Rhew: Uh, the cloud infrastructure that is necessary for running ai and then the cyber protections that are necessary to secure the data on the cloud, and then also to enable the discounts that we see.
- 12:28Dr. David Rhew: So you can't just simply provide, uh, these technologies at regular prices.
- 12:33Dr. David Rhew: You have to actually enable this to be done at a. An affordable process.
- 12:37Dr. David Rhew: And so that was something that we felt was extremely valuable as building on it.
- 12:41Dr. David Rhew: And then of course, we just talked about this recently, the trustworthy, responsible AI network, helping organizations work together to share in their ability to learn from each other and ultimately be able to do things in a time and cost and resource efficient manner.
- 12:54Dr. David Rhew: That's allowing us to be able to think of new models of care where we potentially can enable.
- 12:59Dr. David Rhew: Uh, what we call a hub and spoke model where higher resourced organizations can provide some of the infrastructure and support that supports lower resource ones.
- 13:07Dr. David Rhew: And these are the types of strategies and technologies and approaches that are gonna allow us to be able to democratize healthcare technologies across all regions, not just in the US but also globally.
- 13:19Francesca: I, I absolutely love that whole idea.
- 13:21Francesca: And I think that's so important, especially here on our podcast where our theme is really about, you know, how technology is so human-centric.
- 13:29Francesca: I wanna stay on that thread of trust, just for one, you know, second more and really start to sort of dig a little bit into some of the bias.
- 13:36Francesca: So your keynote that you're doing for us.
- 13:38Francesca: In April really emphasizes the notion of rebuilding trust in healthcare systems.
- 13:43Francesca: Um, and particularly at a time where healthcare misinformation is starting to be weaponized a little bit.
- 13:49Francesca: We're hearing a lot of that in the news.
- 13:51Francesca: Um, and this trust in healthcare organization seems to be declining.
- 13:54Francesca: In fact, there was a recent stat with the Civic Health and Institute's project that found that general confidence.
- 14:00Francesca: In doctors and hospitals drop from 70% in 2020 to 40% in 2025, which is alarming.
- 14:08Francesca: Now, Ned Health's governance, uh, really our framework really tries to address this better, but our clients often tell us.
- 14:15Francesca: That their clinicians are skeptical of another tech solution, quote unquote.
- 14:20Francesca: So how do you counsel health systems as they rebuild this?
- 14:24Francesca: Clinician's trust when they're thinking about AI tools that are being layered on top of EHRs, that they're already a little bit frustrated about.
- 14:32Dr. David Rhew: Yeah, you know what?
- 14:33Dr. David Rhew: Trust is essential at multiple different levels.
- 14:36Dr. David Rhew: There's, uh, trust with the clinician, which is what we're talking about here.
- 14:40Dr. David Rhew: Uh, and how do you build trust?
- 14:41Dr. David Rhew: Well, they have to be involved.
- 14:43Dr. David Rhew: They have to understand what, you know, uh, the, the data are that behind it.
- 14:47Dr. David Rhew: The transparency of the models.
- 14:48Dr. David Rhew: They have to understand how it's going to impact their lives.
- 14:51Dr. David Rhew: Uh, there's, you have to understand whether or not the AI will actually continue to stay, uh, effective in terms of his, uh, overall.
- 14:59Dr. David Rhew: Deployments, and I feel like the, these are the types of things that healthcare physicians and organizations need to think about.
- 15:05Dr. David Rhew: But then there's another level of trust, which is around patients and patients and consumers trusting the technology.
- 15:11Dr. David Rhew: And I think that is something that we also have started to realize the importance of, you can't just simply run it.
- 15:17Dr. David Rhew: Provider side and expect patients to be able to believe that this is gonna be leading to things.
- 15:21Dr. David Rhew: Uh, there there are issues that relate to one.
- 15:24Dr. David Rhew: Uh, do do I really understand when this is being used?
- 15:27Dr. David Rhew: Like, the first question is, have you told me about this?
- 15:29Dr. David Rhew: Like just the general awareness.
- 15:31Dr. David Rhew: Uh, we've also recognized that consumers are concerned about, like with the data that's being captured, what's gonna be done with it?
- 15:37Dr. David Rhew: Is it gonna be.
- 15:38Dr. David Rhew: Somehow exposed to folks that I, I'm not particularly interested in having this exposed to.
- 15:43Dr. David Rhew: So this whole concept of data management and governance is essential.
- 15:47Dr. David Rhew: And so that then left ties back into the transparency and then giving individuals the opportunity to opt in and opt out.
- 15:53Dr. David Rhew: I think that with an informed decision, they might be able to do that.
- 15:57Dr. David Rhew: Now many physicians, uh, or many patients will likely say, well, I trust you and I trust your decision making to, to do the right thing for me.
- 16:05Dr. David Rhew: But just like with anything, we do need informed consent and we do need to make sure people know what's going on so that they can actually feel confident and trust in it.
- 16:14Francesca: I agree with you on that and, and, you know, you specifically hold six US patents on clinical decision support, NER data interoperability.
- 16:22Francesca: Um, you actually recently joined Ignite data as board, specifically because AI and interoperability makes that process so much more efficient, inclusive, and accessible.
- 16:32Francesca: So how does poor data interoperability really limit what's possible with clinical ai?
- 16:37Francesca: And what should specialty eh HR specifically be doing differently?
- 16:42Francesca: In this, uh, scenario.
- 16:44Dr. David Rhew: Yeah.
- 16:44Dr. David Rhew: I'm glad you mentioned that question because I oftentimes find that people talk about ai.
- 16:49Dr. David Rhew: Uh, data and interoperability is three separate entities.
- 16:53Dr. David Rhew: Well, in reality, they're interrelated and what we find is that the AI is highly dependent on the data sets that they were built off of, that they were modeled after.
- 17:04Dr. David Rhew: And those data sets really do need to be.
- 17:07Dr. David Rhew: Large, diverse, and oftentimes we don't see that because it's hard to be able to get access to these large, diverse data sets.
- 17:14Dr. David Rhew: So what ends up happening is it's tested on smaller data sets that may not be representative of larger populations, but at the same time, we wanna generalize the AI on larger populations.
- 17:23Dr. David Rhew: So this whole interrelationship between AI and the data sets is important.
- 17:28Dr. David Rhew: Now, how do you create large data sets and, and make sure that we're actually able to pull in different data sets from different systems?
- 17:35Dr. David Rhew: Of course we've got regulatory privacy rules and things of that prevent it.
- 17:39Dr. David Rhew: So we do need to think about how we can enable different mechanisms to be able to pull data sets, whether it's centrally or through a decentralized process like federated learning.
- 17:48Dr. David Rhew: But interoperability is essential.
- 17:50Dr. David Rhew: 'cause if we, we have these data sets in different areas and we wanna pull them.
- 17:54Dr. David Rhew: It can only happen if the data sets are all speaking the same language, and that is interoperability.
- 18:00Dr. David Rhew: That's why it actually is both the ai, the data and the interoperability that need to be done, uh, in collaboration, in coordination.
- 18:08Dr. David Rhew: And we're starting to see improvements in all three areas, which is why we're starting to see so much great opportunities for ai.
- 18:15Dr. David Rhew: Now, what should EHRs be doing?
- 18:17Dr. David Rhew: Well, obviously in order for us to be able to think about how to improve that interoperability piece.
- 18:22Dr. David Rhew: It would be very beneficial if we could actually enable not only access to these pieces of information, but complete access.
- 18:30Dr. David Rhew: There's like, uh, when a, when an individual, if a consumer says, I want my medical record, we don't want just sort of like the minimal requirements, the minimal dollars, we want actually the full, comprehensive data sets, including notes and things like that.
- 18:42Dr. David Rhew: Uh, and largely because it's up to the consumer to decide where those notes should be sent to, if it's another clinician or if it's gonna be done through an AI model.
- 18:51Dr. David Rhew: That that is something that the consumer should be able to gain complete access to and real-time access to.
- 18:57Dr. David Rhew: And this is a challenge that goes beyond the EHR because how can you ensure that we're getting real-time access as opposed to delays which occur or piecemeal information coming here and there.
- 19:08Dr. David Rhew: And I think this is something that we're gonna continue to see more of an opportunity for us to think about how.
- 19:13Dr. David Rhew: Different systems work together to be able to enable consumers to be able to get complete and real-time access to their data sets so they can better manage it or PR transfer them to the individuals that can.
- 19:25Francesca: So I heard you say large sample sizes, the types of access, the broad ability to access different types of data.
- 19:33Francesca: Timing is a key thing, but isn't it also true that.
- 19:37Francesca: There is sometimes some inherent bias in some of those data sets.
- 19:40Francesca: And, um, Dr. Gita Naer, who's actually a friend of the show, an occasional co-host here, she really goes into, in her book, she talks a lot about misinformation and she talks about how we have to combat that and make sure that it's equitable, racially sensitive.
- 19:56Francesca: Right.
- 19:56Francesca: As we're thinking about how we're, we're.
- 19:58Francesca: Potentially finding AI perpetuating some of the disparities and maybe even victimizing some of the more underserved populations.
- 20:06Francesca: Right?
- 20:06Francesca: And then researchers at Mount Sinai found that LLMs often just repeat and then even expand on some of this false medical information and do it so convincingly, like they, with a lot of confidence, make it sound so real.
- 20:19Francesca: So what does clinical grade AI really need to include in order to ensure it's acting that it's actually reducing misinformation rather than.
- 20:28Francesca: Creating it with this veneer of authority.
- 20:31Dr. David Rhew: Yeah, I mean, you've, you've touched on so many important topics within that question.
- 20:35Dr. David Rhew: Uh, you know, the first is the, the difference between, you know, with, with bias, uh, looking at bias in terms of what we see in ai.
- 20:45Dr. David Rhew: Some of it is not intentional bias.
- 20:47Dr. David Rhew: Actually, a large part of it is not.
- 20:49Dr. David Rhew: It's just simply the fact that we had a very narrow D data set that we have con D determined the AI model, and now what we're doing is we're applying it to other.
- 21:00Dr. David Rhew: You know, individuals and other populations.
- 21:02Dr. David Rhew: And so in that application of that very narrow model to more generalized settings, we have found that it doesn't work as well.
- 21:09Dr. David Rhew: And then in that process, it has created disparities in terms of outcomes, specifically in groups that are not well represented in those data sets and who's not what well represented.
- 21:19Dr. David Rhew: Turns out that it's the vulnerable, it's the low resource, it's the underserved because they haven't been able to gain access to the health system.
- 21:27Dr. David Rhew: The data hasn't been, you know, they don't come in regularly.
- 21:30Dr. David Rhew: They don't get all the tests, and so because of that, that.
- 21:33Dr. David Rhew: Sort of influences the data sets to become much more reflective of a different type of population.
- 21:40Dr. David Rhew: So to me, that's like one element that we have to start thinking about.
- 21:43Dr. David Rhew: Now, the other is about misinformation and propagation of misinformation.
- 21:48Dr. David Rhew: Now that that can be done largely because of the fact that people don't know what the truths are and they don't believe the truths.
- 21:54Dr. David Rhew: And so there's a whole element here of how do we change the whole concept of like, what is truth?
- 22:01Dr. David Rhew: And I think it largely stems from the fact that we're not all on the same page in terms of what is ground truth.
- 22:07Dr. David Rhew: Some people believe that ground truth is the last thing that I saw.
- 22:10Dr. David Rhew: Somebody said that I had an experience and this is what, uh, uh, that that, you know, maybe there was a side effect with a medication.
- 22:17Dr. David Rhew: Now we think all medications are bad, you know, or all treatments are bad.
- 22:21Dr. David Rhew: Well, is that evidence or that's one piece of evidence.
- 22:24Dr. David Rhew: But now if we're applying it for a general population.
- 22:27Dr. David Rhew: Is it true that all the medicines are bad for every single person?
- 22:30Dr. David Rhew: Now, some people have allergies, some people have unique reasons why they have adverse reactions to different things.
- 22:36Dr. David Rhew: What we have found is that randomized control trials.
- 22:40Dr. David Rhew: Allow us to be able to study things in a way that allows us to be able to understand whether or not certain treatments are better or worse on a larger population, not necessarily an individual population, but a larger population.
- 22:53Dr. David Rhew: And in those scenarios, if we're gonna make a generalization about whether treatment is good for a population, those type of studies actually should carry more weight.
- 23:01Dr. David Rhew: Now, where are we heading?
- 23:03Dr. David Rhew: We're actually now heading towards more personalized medicines, and so we may be able to start learning from individual experiences just like you on what your best treatment is, and that may end up becoming the best option, but we're not quite there yet.
- 23:16Dr. David Rhew: And so that's why it's important to know where we're at right now.
- 23:20Dr. David Rhew: We are at the point where we're gathering large services of data.
- 23:23Dr. David Rhew: We're starting to better analyze it, and I think the education is not just on the clinician side, but it's gonna have to be on the consumer side.
- 23:30Dr. David Rhew: You know, I mean, imagine this, uh, you know, our, our, because of cuts to education, because of the la lack of focus on a lot of this, people are not only learn, not learning all the sciences necessary, but they're not learning what I could refer to as the scientific method.
- 23:45Dr. David Rhew: If more and more people were introduced to this concept of two pieces of information, which is the one that is more likely to be true for under general population, I think we would probably start under getting people to better understand that.
- 23:58Dr. David Rhew: And then the last thing is, where are you getting your information source from?
- 24:01Dr. David Rhew: Are you getting it from a, a, we'll call it a trusted individual that has an understanding of it, or are you just getting it from a trusted individual?
- 24:09Dr. David Rhew: And I think with nowadays.
- 24:11Dr. David Rhew: Trusted individuals can come in all shapes and sizes.
- 24:13Dr. David Rhew: Some may have some scientific understanding, some may have none.
- 24:16Dr. David Rhew: And I think what we're finding is that people are relying on just the fact that I've worked with this person, I listened to them, I trust them, as opposed to that plus this person understands the scientific method.
- 24:28Dr. David Rhew: So I think there's a lot here that to unpack, uh, this is obviously a very important topic, but one that I think that we collectively as a society need to address.
- 24:38Francesca: I feel like you've seen into my soul because with my recent doctor's office, I took my notes that I got and put them into, you know, Claude and got an assessment.
- 24:47Francesca: Um, and I think that a lot of healthcare workers, you spoke about consumers here, you spoke about, you know, who is their trusted individual, is your trusted individual, your ai, you know, model that's helping you.
- 24:57Francesca: Um, and I think these healthcare workers are grappling with this notion of shadow ai, right?
- 25:02Francesca: Like staff, I think, are using unapproved AI tools themselves because they're using, you know, transcription notes or different ways to sort of combat their own, uh, burnout and staffing shortages.
- 25:13Francesca: And that's forcing organizations to have to think about how to implement formal governance frameworks around those.
- 25:19Francesca: And then you've got patients who are also running doctor's notes through your chat GPTs, your CLOs, your Geminis, et cetera, without any clinical context as you just described.
- 25:28Francesca: So when you've got patients self diagnosing using AI or clinicians using unauthorized AI tools, who's accountable when something goes wrong?
- 25:37Francesca: In other words, how do we do what you just said, which is balance that innovation.
- 25:41Francesca: With patient safety when the technology is moving a lot faster than our governance structures, right?
- 25:46Francesca: How do we do
- 25:47Dr. David Rhew: this?
- 25:47Dr. David Rhew: Yes.
- 25:48Dr. David Rhew: And, and we are at that point right now where the technologies are getting so good that people are relying on them for medical advice, uh, independent of clinicians.
- 26:00Dr. David Rhew: And I think that's not necessarily the path that we should be looking at right now or encouraging because of the fact that there are some things that the AI can do very well.
- 26:09Dr. David Rhew: And then some things that the AI with the doctors and clinicians.
- 26:14Dr. David Rhew: Uh, will allow us to be able to do better.
- 26:16Dr. David Rhew: So if I look across the board, let me, uh, lemme just kinda reflect back on the days that I was a clinician, uh, when a patient came to see me and I gave them a recommendation in terms of what the treatment was, prescribe a medication, uh, certain amount of exercise, diet, and ultimately making sure that to, to know what, what are the warning signs?
- 26:36Dr. David Rhew: These are the types of things that I would give at a very general level.
- 26:39Dr. David Rhew: Now as a patient.
- 26:41Dr. David Rhew: I would gather that information or an individual would, and they may have a lot of questions.
- 26:45Dr. David Rhew: Yet, I didn't have enough time to talk to the doctor about it.
- 26:48Dr. David Rhew: I didn't understand, you know, you said take it three times a day.
- 26:51Dr. David Rhew: Am I supposed to take it every eight hours or am I supposed to take it three times a day during the daytime?
- 26:55Dr. David Rhew: You know, uh, breakfast, lunch, and dinner.
- 26:58Dr. David Rhew: Uh, you told me that I need to eat this with food, any type of food, you know, like what about spicy foods?
- 27:03Dr. David Rhew: I mean, you name it, there's a lot of questions that come up.
- 27:06Dr. David Rhew: People also, I have questions.
- 27:08Dr. David Rhew: You told me to exercise.
- 27:09Dr. David Rhew: How much should I exercise?
- 27:10Dr. David Rhew: You know, he told me to, uh, eat right and low sodium.
- 27:13Dr. David Rhew: What's low sodium?
- 27:14Dr. David Rhew: You know, uh, is this food, low sodium?
- 27:16Dr. David Rhew: All those questions about how you would translate what the doctor was saying into real life.
- 27:21Dr. David Rhew: AI can do very well.
- 27:23Dr. David Rhew: Similarly, if I were to see the doctor preparing for my surgery and I wanna ask questions about like, what, what questions should I be asking when I see the doctor?
- 27:31Dr. David Rhew: AI can help us prepare for that.
- 27:33Dr. David Rhew: AI helps villain fill in the void for so many of the things that we currently do not have resources for, but it is also done best in collaboration and coordination.
- 27:44Dr. David Rhew: With a healthcare provider so that whatever we're doing is supportive as opposed to replacing.
- 27:50Dr. David Rhew: And we oftentimes know that AI has not yet been tested in some of those independent matters.
- 27:55Dr. David Rhew: So for us to be able to start relying on things for everything would be an overshoot.
- 28:01Dr. David Rhew: But if we rely on AI to help us with things that have never been addressed before and and really give us guidance, but ultimately deferring back to the clinicians for final.
- 28:12Dr. David Rhew: Input.
- 28:12Dr. David Rhew: I think that allows us to be able to balance the issue of answering questions, but also doing it in a safe and effective manner.
- 28:18Francesca: That sounds fantastic, um, and such great, uh, counsel and guidance there.
- 28:23Francesca: Let's switch gears a little bit.
- 28:24Francesca: Let's think about what we're hearing and seeing a lot in the news of late.
- 28:29Francesca: There's a lot of discussion about all these great AI pilots, um, and there's a lot of discussion about how do we actually get these into production.
- 28:36Francesca: So, you know, the proverbial sort of implementation gap.
- 28:39Francesca: Um, and many of these healthcare organizations are getting stuck in this perpetual sort of pilot mode, right, with ai, and you've actually distinguished between the Coalition for Health AI C, which is about defining what responsible AI is, and then train, which we spoke about earlier, which is really focused on figuring out how to implement it.
- 28:59Francesca: So what advice would you give for moving from proof of concept to scaled AI deployment?
- 29:04Francesca: And what should organizations make sure they change in order to be able to support that happening?
- 29:10Dr. David Rhew: So I'm gonna reflect on one of my experiences.
- 29:13Dr. David Rhew: Before I joined Microsoft, I was Chief Medical Officer at Samsung and we worked with Kaiser Permanente on an initiative called Home-Based Cardiac Rehab.
- 29:21Dr. David Rhew: So let me start with the problem.
- 29:22Dr. David Rhew: The problem was that a lot of individuals after they'd have a heart attack or have some type of cardiac surgery, needed this cardiac rehab program.
- 29:32Dr. David Rhew: Uh, why?
- 29:32Dr. David Rhew: Because it turns out that that program saves lives.
- 29:35Dr. David Rhew: And it prevents hospitalizations and it's something that's very valuable.
- 29:39Dr. David Rhew: Okay.
- 29:40Dr. David Rhew: Why are they not doing it?
- 29:42Dr. David Rhew: It turns out because it's very difficult.
- 29:43Dr. David Rhew: It's a six week program.
- 29:45Dr. David Rhew: It every, every week you have to come in three times a week, you have to exercise for 30 minutes, and you basically have to commit to doing this.
- 29:53Dr. David Rhew: Well, people have jobs, you know, they, they are, uh, busy taking care of their kids.
- 29:56Dr. David Rhew: They live far away.
- 29:58Dr. David Rhew: Uh, and to be able to do this on such a regular basis.
- 30:01Dr. David Rhew: Is a huge commitment that one, that most individuals don't, can't do.
- 30:05Dr. David Rhew: So at kp uh, their baseline rate prior to the implementation of this home-based program was around 40%, which is really not good.
- 30:14Dr. David Rhew: Less than half of the people are receiving this life-saving treatment.
- 30:17Dr. David Rhew: So what we decided was to ask the cardiologist, does the person actually need to be seen in the hospital or in the clinic, in this facility?
- 30:26Dr. David Rhew: And they said, well, a lot of this could be done at home.
- 30:28Dr. David Rhew: So what we did is we built this home-based cardiac rehabilitation technologies kit, which involved digital technologies.
- 30:36Dr. David Rhew: It was actually a wearable, it reminded people on, on that to exercise, do so with something called a six minute walk test.
- 30:43Dr. David Rhew: Reminded them to take their medicines, but also.
- 30:45Dr. David Rhew: If you had any questions, you could push a button and somebody would get back to you.
- 30:49Dr. David Rhew: And so that bi-directional feedback, sending data to the clinicians so they were aware of what's going on, clinicians actually coming back and letting them answer the questions.
- 30:57Dr. David Rhew: That was something that this program was based off.
- 30:59Dr. David Rhew: And what happened?
- 31:00Dr. David Rhew: The completion rate went for, this program went from 40% to 80%.
- 31:04Dr. David Rhew: It doubled, which is extraordinary.
- 31:06Dr. David Rhew: And so what that meant is for every seven to 17 patients that completed this program, we were saving one life.
- 31:13Dr. David Rhew: It's extraordinary.
- 31:15Dr. David Rhew: We had about a hundred or so patients start in this pilot, so I'll call it a proof of concept.
- 31:20Dr. David Rhew: It expanded to a couple hundred, went to 1200, and then ultimately it was reviewed by the entire administration, and it is now considered the first line option.
- 31:31Dr. David Rhew: In addition, as an alternative to in to in-person.
- 31:35Dr. David Rhew: And right now, over 35,000 people.
- 31:38Dr. David Rhew: Have benefited the impact this has had, not only in terms of completing the rates, the readmission rates have gone from 11% to 1%.
- 31:45Dr. David Rhew: These results have been published in the New England Journal Catalyst.
- 31:48Dr. David Rhew: It represents one of the largest scale implementations that is successful in which they took this great idea from I, you know, concept pilot to now large scale production.
- 32:00Dr. David Rhew: And how do they do that?
- 32:02Dr. David Rhew: One, they proved that it worked.
- 32:03Dr. David Rhew: In terms of like demonstrating its value, they found ways to make it workflow compatible, and then within the KP system, people are incentivized to be able to then deliver on this.
- 32:13Dr. David Rhew: And I think it's really that last piece.
- 32:15Dr. David Rhew: That's the only missing piece for this to be widespread adopted across the entire globe.
- 32:20Dr. David Rhew: If we could find ways to encourage people to be able more supportive of preventive care as opposed to reactive care, I think that that would allow us to be able to start thinking about this.
- 32:30Dr. David Rhew: But the essential elements from a.
- 32:30Dr. David Rhew: Healthcare organization standpoint where, take it a great idea, incubate it, test it, get the workflows right, scale it to the next point, and then continue to scale it to the point where it is considered just as good at, if not better, than what you're currently doing.
- 32:47Dr. David Rhew: And then make it first line option.
- 32:48Dr. David Rhew: And.
- 32:50Francesca: So really what you would say to the skeptics out there that are predicting a sort of a mass extinction event, so to speak in 2026 for a lot of these AI startups, as they transition from this promise to proof, is essentially that real outcomes, right?
- 33:04Francesca: Real clinical outcomes is, is really where people should focus, right?
- 33:08Francesca: Is that what you believe separates.
- 33:10Francesca: Financial
- 33:11Dr. David Rhew: and financial because, uh, at this day and age, as I mentioned before, if it's just about, uh, improving the overall, uh, quality, uh, unfortunately that doesn't move the adoption needle as much as making things easier for people.
- 33:25Dr. David Rhew: And in this case, it made it easier for patients.
- 33:28Dr. David Rhew: It also made it easier for, for, uh, clinicians so that they could actually focus on other things that were allowing them to be able to become much more effective.
- 33:37Dr. David Rhew: Uh, but then at the same time, the readmissions dropped.
- 33:40Dr. David Rhew: Readmissions has a huge implication in terms of.
- 33:43Dr. David Rhew: Reimbursement and penalties and finding ways that specifically we can improve upon that is, uh, both a quality issue as well as a, as a, as a cost and, uh, expense issue.
- 33:54Dr. David Rhew: So I feel like you have to have those type of pieces in there.
- 33:57Dr. David Rhew: The, of course, the quality improvement, but, and it's needed to be clinical, you know, that the workflow has to be done right.
- 34:04Dr. David Rhew: But the key is also there has to be some economic element here, and in this case, the economic element is what drove a, it continues to drive a lot of interest in this.
- 34:12Dr. David Rhew: So it's the combination of all those.
- 34:15Francesca: I think it's tough though because you've got this environment right now with hospitals operating in margins around 2.3%.
- 34:22Francesca: Um, 26 budgets are just being ruthlessly cut.
- 34:24Francesca: Right?
- 34:25Francesca: And, and as you mentioned, people are focusing on soft ROI like burnout and those types of things.
- 34:30Francesca: But then really what seems to matter is more that ROI, which is the cashflow improvements.
- 34:35Francesca: Right.
- 34:36Francesca: And then, um, you've also emphasized, as you just said, you know, that.
- 34:39Francesca: AI needs to make the process much more efficient, but it has to also focus on being inclusive, accessible, improving the outcomes.
- 34:47Francesca: Um, but are we setting ourselves up for a false choice there or is there really still some genuine tension between AI for humanity versus AI for profitability that we are gonna have to reckon with?
- 34:57Francesca: What's your counsel to healthcare organizations out there on this hard trade off?
- 35:02Dr. David Rhew: Yeah, I think that a lot of organizations are looking at AI in too narrow a viewpoint.
- 35:08Dr. David Rhew: They look at AI as solving a specific problem, when in fact AI may potentially open up opportunities for us to be able to address the bigger issues.
- 35:17Dr. David Rhew: So I'll give you two examples.
- 35:18Dr. David Rhew: One with ambient.
- 35:19Dr. David Rhew: With ambient, we look at ambient and say, oh my gosh, you know, ambient has not been shown definitively to increase revenues.
- 35:27Dr. David Rhew: That wasn't the purpose of ambient.
- 35:28Dr. David Rhew: Ambient was to decrease clinician burnout so that the long term, we actually have clinicians so that we can take care of patients and be sustainable, but that, that, that's not the short term metric that people are looking at.
- 35:39Dr. David Rhew: But what it also does is it converts that entire conversation into a digitized format.
- 35:44Dr. David Rhew: It's just that we can run things like revenue cycle management, AI on top of it, that we can run other areas to optimize reimbursement, to make sure the notes are written in such a way that actually revenue is gonna be captured, and also then allows us potentially to create opportunities for real time prior authorization.
- 36:01Dr. David Rhew: These are the types of opportunities that could only have been opened.
- 36:05Dr. David Rhew: Had we now digitized it through the ambient.
- 36:07Dr. David Rhew: So think of ambient as, uh, an opening for a whole series of challenge opportunities that we have not even explored Revenue cycle being one of'em, but there's numerous others.
- 36:18Dr. David Rhew: Another one, we oftentimes think of AI as replacing, you know, an individual task like they're, we're making their life better or replacing their task.
- 36:26Dr. David Rhew: When in fact, maybe what we should be thinking about is AI being brought on the front end to change the whole dynamic of.
- 36:33Dr. David Rhew: How number, the number of patients that we're seeing and who manages them.
- 36:37Dr. David Rhew: So let's, let's use, uh, some of the things that we're seeing right now.
- 36:41Dr. David Rhew: AI can be applied on large populations to identify individuals that have certain conditions that we, and they may never have known.
- 36:49Dr. David Rhew: Uh, we can find patients that have.
- 36:50Dr. David Rhew: Diabetic retinopathy by looking at macular, you know, the, the, the retinal images.
- 36:55Dr. David Rhew: Uh, we can identify patients that have, uh, depression, stress, and anxiety by looking at voice as a biometric.
- 37:02Dr. David Rhew: We can identify patients with heart failure by looking at their EKG and different things that normally we would have required an echo, but we're now using different signals.
- 37:11Dr. David Rhew: So these are things that could never have been envisioned before.
- 37:13Dr. David Rhew: But AI can do all of those things.
- 37:16Dr. David Rhew: The problem is.
- 37:17Dr. David Rhew: We're finding so many patients that it overloads the system.
- 37:20Dr. David Rhew: And so because of that, what we're finding is that people are saying, no, no, no.
- 37:24Dr. David Rhew: Hold back.
- 37:25Dr. David Rhew: We don't actually run a roll of AI'cause we can't handle the existing load.
- 37:28Dr. David Rhew: Now what we're looking at is AI to risk stratify and bring the highest risk individuals to care and then manage the medium to low risk through other channels.
- 37:38Dr. David Rhew: And so what does that do?
- 37:39Dr. David Rhew: It expands the pie, it expands the funnel.
- 37:42Dr. David Rhew: And so by doing so, the proceduralists, the cardiologists that do procedures, the ophthalmologists that do procedures actually by shuttling the high risk procedure oriented patients to them will likely get.
- 37:55Dr. David Rhew: Not, I mean, if we can remove some of the medium, the lowers and, and not send them to that, those patients, those, those individuals, we can actually increase the level of complexity of their visits and the number of procedures, and ultimately they get paid more in a fee for service world.
- 38:09Dr. David Rhew: So it's a win-win, even in a fee for service world.
- 38:12Dr. David Rhew: By expanding the pie.
- 38:13Dr. David Rhew: Now what you've also done is you've added more individuals on the primary care side, which could potentially be managed by other individuals throughout other, through, through other processes, whether it be digital, virtual, or even through, uh, other advanced care practitioners leveraging the, the fact that they're now using ai.
- 38:29Dr. David Rhew: So what we're finding is that AI can actually, in any ways.
- 38:33Dr. David Rhew: Increase revenue by increasing the funnel and ultimately allowing us to be able to treat the sicker or the sick by finding them and bringing 'em to care earlier.
- 38:41Dr. David Rhew: And at the same time, it delivers better care.
- 38:44Dr. David Rhew: So there's so many opportunities for us to start thinking about how AI can be applied beyond simply looking at it from a task perspective.
- 38:52Francesca: I think your point there about being open to the opportunity, it's one thing to start a pilot with certain implied outcomes, and then as you said, be flexible as the data comes in to be able to expand and see future opportunities.
- 39:04Francesca: But also I think the reality is that with ai, it's still very much an emerging field, and we're gonna discover along the way as we're building the plane.
- 39:13Francesca: Um, and flying it.
- 39:14Francesca: At the same time, we're gonna discover all these additional opportunities we can't even possibly begin to think about right now.
- 39:20Francesca: But on that, um, particular note, and candidly this, I think we're running outta time.
- 39:24Francesca: I'm getting note from our producers here, but I could talk to you all day.
- 39:28Francesca: I do have one more question, though.
- 39:30Francesca: I'm gonna.
- 39:30Francesca: Sneak it in before we have to go.
- 39:32Francesca: Um, and it's on this notion of the future, right?
- 39:36Francesca: So thinking about your vision for 2023, you are coming to our conference in Arizona.
- 39:41Francesca: You're gonna be speaking at the Net Health Next Conference, and your theme is all about where health tech becomes human tech.
- 39:48Francesca: This whole idea that obviously clinicians will give, get, um, more face-to-face time with their patients.
- 39:53Francesca: You've touched on the empathy piece.
- 39:55Francesca: You've spoken about human connection.
- 39:57Francesca: But when you look out to 2020 and you look at what.
- 39:59Francesca: Does success look like?
- 40:01Francesca: What would make you say, yes, we got AI in healthcare, right?
- 40:05Francesca: And it made us much more human, not less.
- 40:10Dr. David Rhew: Uh, again, I love that question because it incorporates all the things that we are talking about right now.
- 40:17Dr. David Rhew: I think what we need to think about are how is AI data and interoperability gonna allow us to be able to, with safeguards, be able to increase access to care?
- 40:28Dr. David Rhew: Reduce costs, reduce, improve quality and safety, and ultimately create new opportunities for jobs so that individuals in our society can continue to grow.
- 40:36Dr. David Rhew: That's the bar that we need to raise.
- 40:38Dr. David Rhew: We need to think about beyond going from, it solves a task and removes people from those, those processes to, it elevates the opportunities for all of us, and as long as it's done in a way that's safe, inclusive, and ultimately one that's gonna be cost efficient, it allows us to be able to deliver better care for all.
- 40:57Francesca: That's terrific.
- 40:58Francesca: Thank you so much, Dr. David Rhew.
- 41:00Francesca: It has been such a pleasure and such an honor speaking with you today.
- 41:03Francesca: I cannot wait to see you in person in April, um, at our conference.
- 41:08Francesca: Thank you again for your time today.
- 41:10Dr. David Rhew: It's been my pleasure.
- 41:11Dr. David Rhew: Thank you, Francesca.
- 41:17Francesca: What a great episode today.
- 41:18Francesca: So today we covered three real key points.
- 41:21Francesca: First, AI works best when it's an invisible partner, something that's built right into the workflows that clinicians can already use, and actually gives them some time back to feel much more human.
- 41:32Francesca: Secondly.
- 41:33Francesca: We discuss that trust and good governance are not optional.
- 41:37Francesca: Programs like train show that responsible deployment matters even if clinicians and patients are going to rely on ai.
- 41:46Francesca: And finally, the real impact of AI goes well beyond efficiency and cost savings.
- 41:51Francesca: It's giving clinicians more face-to-face time, helping them make better decisions and improving the overall experience for both patients.
- 41:58Francesca: And providers in driving to better outcomes.
- 42:02Francesca: Thanks again for joining us, Dr. Rhew, and thank you to all of our listeners.
- 42:07Francesca: If you're tuning in from Spotify or YouTube, we'd love to hear your thoughts on this episode in the comments below.
- 42:14Francesca: And once again, I'm Francesca Vera, head of marketing at Net Health, and I'll see you next time.