Optimistic Canadians

Canadian healthcare innovators are proving that optimism, data, and thoughtful technology can reshape pediatric care. In this episode, three leading physician-innovators share how digital health, artificial intelligence, and smarter care models are transforming outcomes for children and families.

Dr. Shazhan Amed discusses how her startup Haibu Health is using digital health platforms and data integration to improve the lives of children living with type 1 diabetes, reduce hospitalizations, and support care across the entire lifespan.

Dr. Joshua Liu, CEO at SeamlessMD, explores the evolving landscape of AI in healthcare, from the rapid rise of AI scribes to the next generation of tools focused on care delivery, workflow automation, and patient engagement.

Dr. Devin Singh, Founder & CEO Hero AI, shares groundbreaking work using real-time AI models in the pediatric emergency department to accelerate diagnoses, reduce wait times, and improve care for vulnerable populations.

Together, they offer an optimistic perspective on how Canada’s healthcare ecosystem is driving meaningful innovation in pediatric care.

Episode Resources:

Scribe - Smarter documentation software, powered by AI

Revolutionize how you write text - AI Sidekick

Connect with Dr. Shazhan Amed:

Dr. Shazhan Amed LinkedIn

Haibu Health Website

Haibu Health LinkedIn

Dr. Shazhan Amed BC Children’s Hospital Research Institute

Live 5210 - BC Children's Hospital Research Institute

Connect with Dr. Joshua Liu:

Dr. Joshua Liu LinkedIn

SeamlessMD Website

SeamlessMD LinkedIn

SeamlessMD Instagram

Connect with Dr. Devin Singh:

Hero AI Website

The Hospital for Sick Children Website

The Hospital for Sick Children LinkedIn

The Hospital for Sick Children Instagram

Dr. Devin Singh LinkedIn

Connect with us:

KidsX Website

KidsX LinkedIn

Children's Hospital L.A. Website

Children's Hospital L.A. Instagram

Children's Hospital L.A. LinkedIn

Transcript

SPEAKER_02 0:00

Oh Canada. Our neighbors to the north are friendly. They're great at hockey. They love their maple syrup. They're Wayne Gretzky and their booblet. But they also have a deep pride in their national healthcare system. Admittedly, I knew very little about the Canadian healthcare system before preparing for this episode. So, first, Canadian healthcare isn't free. It's paid for through taxes, and most Canadians understand that they're making that trade-off. Higher taxes for services like healthcare. And most are happy with what they get. I even saw a poll recently that showed 85% of Canadians are generally happy with their health care. Another thing, there's no way to skip the line. Not really. You can't pay your way through, you can't private pay in Canada. I believe it's illegal. So even the wealthiest Canadians have to wait in line. Another thing, most physicians in Canada run their own practices. They hire their own staff, they choose their own schedules. The government is just the payer. They're not the employer in Canada, which is different than a lot of other countries that have universal health care. And what's interesting, and you may have heard this, Canada spends a fraction of what we do here in the United States on administrative functions related to healthcare, billing, coding, insurance administration, probably because there's only one main payer in each province. So a baby born in Canada, the delivery, the anesthesia, the pediatric checks, the hospital stay, typically there's no copay, no deductible. It's all part of what they get in Canadian health insurance. And so as a result, medical debt in Canada is largely low. It's mostly non-existent. There's a few cases here and there, but compared to what it is in the United States, it's not there. Now, having said that, there are some drawbacks. So even though they have universal hospital and doctor coverage in Canada, there still is not a complete plan around drug coverage. So many Canadians rely on employer insurance or provincial plans that they pay for to get medications and drugs. So that's one issue. The other is around waits. So many Canadians for most specialty care services will have to wait longer to get those services than many people in the United States do. And that's especially true for Americans with commercial insurance. Americans on Medicaid often have comparable wait times for certain specialty procedures or specialty care access. But wait times are definitely such where there is a difference. Now, two things to point out that I noticed there is a heavy emphasis in the Canadian healthcare improvement space around reducing wait times in the emergency department. There is a big national focus around wait times. There's transparency in that data, incentives around ETs that provide improved access and reduced wait times. And so that'll play a role in one of the interviews we have today. The other thing that is really interesting as it relates to innovation in Canadian healthcare is there is a real focus on longitudinal health. So there is a deep interest in making sure that children born with certain diseases, say diabetes, which we're going to talk about, in making sure those children are cared for, not only from a clinical quality standpoint, but to make sure that the overall longitudinal elements of their care are thought through. Because if a child's disease is not managed in childhood properly, it is going to cost their health system financially through more complications, more stays, more utilization as that child becomes an adult. So compared to the United States, where care is episodic, where we have fragmented elements of care, there is this real desire from a longitudinal population health standpoint, the Canadian health system, to make sure that a lot of childhood diseases and childhood illnesses are well managed because they could turn into very costly, expensive diseases to treat as these children grow into adulthood. And that's going to show up in all three of the interviews you hear from today, and something I'm really excited about. So today's episode is focused, as you can imagine, on Canadian healthcare and some of the innovations and innovators that are building really interesting solutions north of the border. I hope you'll enjoy. Welcome to the Dose of Optimism, where I talk to the optimists in healthcare. My name is Omkar Kulkarni, and I work at one of the world's best children's hospitals where I lead innovation. I started KidsX, which is a premier international startup accelerator for pediatric innovation. And over the years I've met thousands of startups, investors, and innovators. Every one of them has a story, and every one of them is optimistic about the problems they're solving. On this podcast, you'll meet amazing people who will share their stories and what makes them optimistic about the future of healthcare. A little note before you get into this episode. Please talk to your own physician about your health or the health of your children. All right, let's get started. So we start first in Vancouver in the British Columbia. Dr. Ahmed is a clinician leader, is great talking to her. She's an entrepreneurial leader over in BC Children's Hospital. She's a pediatrician and really focuses on kids, particularly with diabetes. And she's doing some really innovative things as we think about longitudinal care. Kid who gets diagnosed with diabetes today, how we can treat them as they become teenagers and specifically that transition into adulthood. So I loved this conversation and I hope you'll enjoy hearing from Dr. Ahmed from BC Children's. That's amazing. So current state, there's patients diagnosed with type 1 diabetes. We provide them with sometimes a glucometer, an insulin pump. We want as much data as possible about them. And on the same token, we don't have the staff or the people or the families don't have the ability or the time to really be able to make sense and manage all that data. And so it's, it seems like this interesting problem where we know there's opportunities to be able to use data in digital health to help these kids, but we may not have the tools to really aggregate and create insights out of all that data.

SPEAKER_00 6:44

Exactly. Kids with diabetes today have access to these great technologies. You mentioned glucose sensors, insulin pumps. Glucose sensors generate a glucose level every one to five minutes. That's a lot of data. And it's been a game changer for diabetes management. We've seen great improvements, both in quality of life as well as outcomes. Yet we see these kids using these greatest and latest technologies, but they're still suffering. And what does that mean? A certain proportion of them are still getting admitted to hospital for a life-threatening condition called diabetic ketoacidosis. We're still seeing kids affected by diabetes distress and sometimes leading to depression, anxiety. We still see eating disorders at a higher rate compared to those who don't live with diabetes. And so there's still work to be done to improve outcomes in kids. And we believe at our company, HIBU Health, that we can unlock the potential of all of this data that they generate by unifying it and using it intelligently.

SPEAKER_02 7:57

Repeatedly, in what you've told me, preventing hospitalizations for these kids seems like the number one driver, as you think about the value created by the technology. Am I hearing that right?

SPEAKER_00 8:09

Hospitalizations are a measurable impact where we have the data. We know how many kids are hospitalized every year. We know how much that costs the system. So it's a good one to look at to say, hey, we can save the system this much, in addition to the suffering for patients and families when they're admitted to the ICU and go through that experience, which can be quite traumatic. But there's potential beyond that, where in diabetes we have this great measure called the hemoglobin A1C, the average glucose level over the preceding three months that research from the 80s and 90s showed us that value is directly linked to complications. And a reduction in the A1C translates into a significant reduction in risk for those serious diabetes complications like kidney disease, blindness, amputations, cardiovascular disease. Using digital health and innovation and technology to optimize that hemoglobin A1C level during childhood also translates into significant savings 15, 20 years later, where we have the ability to reduce the risk and therefore the occurrence of these serious complications. And that's where I really see the potential in leveraging data during childhood and adolescence to benefit the patient across their lifespan and as a result benefit the healthcare system in terms of cost savings. And that's really where I think we can really move the needle because not only when you prevent these serious complications, do you prevent costs in the system, you also have more economic productivity out of these patients because they're working, they're not in the hospital getting treated. You have young adults going to university like they should be, rather than having to deal with the complications of their diabetes. So there's so many benefits that can come out of this. Some of them are measurable and some of them are not, but that doesn't make them less important.

SPEAKER_02 10:15

I want to really emphasize what you're talking about here. This episode is focused on innovations coming out of Canada. And because you have a national health system, and I imagine that you think about the value of innovations longitudinally over the cost and value of a person versus annual health plan related savings that we here in the United States sometimes think about because patients here are always going into and out of different health plans. It's an interesting difference. The value propositions you talked about in terms of lifelong, even clinical impact, it's a very different approach to thinking about the value of an innovation. And I just want to double-click on that a little bit with you.

SPEAKER_00 11:01

I love that you pointed that out because I'm so embedded in the Canadian system. I don't think that was top of mind for me, but you've really pointed out an important distinction of how countries like Canada, where we have universal health care, that we do look at how the patient uses the system over a lifespan in terms of cost versus these time-limited or insurance plan limited, maybe is how I would say it, interactions or durations of time in the system. And for us, looking across that lifespan is so important. If we keep kids healthy when they're kids, that translates into healthier adults, which then of course translates into many more benefits for our system. So this lifespan approach is actually very important to us in the Canadian system.

SPEAKER_02 11:52

Do you talk about things like that when you think about anything, not just that your product here, but just health in general? And are you thinking about the lifelong impact of a child with any kind of disease? Are you guys thinking about what impacts that would have on them as adults?

SPEAKER_00 12:09

Absolutely, because it's a continuous system of care. And take transition, for example. Transition is an important moment in time. I shouldn't even call it a moment in time because it's happening, it should be happening over the course of a few years preceding that young person transitioning to adult care. And then a few years after they end up in the adult system. And that's one of the first kind of areas of focus for us is how do we successfully transition them? So, what does that mean? We defined it in our research that we just published last year was seeing an adult provider within a year of being transitioned out of the pediatric system. And because we are a national healthcare system, we can look at these big, massive administrative data sets. So that's what we did here in BC. And we were actually shocked to see that in our province, 70% of young people transitioned out of pediatric care did not see an adult provider within a year of that transition. So did not have a successful transition based on a definition that they have to see an adult provider within a year. And so that was like a really great signal. Like we are missing the mark when it comes to this lifespan approach to care. Or losing these young people in our system at a critical point in of time in their life, right? They're going to university, they're living independently, they might be working, they're leaving the nest, and our system isn't working for them. And so again, that's another example of how we see care in Canada across that lifespan.

SPEAKER_02 13:53

And what an opportunity for a high boo health or a digital solution, right? Because now you've got a digital, almost like a passport that follows you from pediatric care to adult care. All your data from your teens and your childhood is available to that next doctor who can provide continuous history of all your diabetes care data looked like from the moment you were diagnosed until that transition point into adult care. Seems like it could really enable and help with that transition from pediatrics to adulthood as these young diabetics are turning into adult diabetics.

SPEAKER_00 14:27

Absolutely. You nailed it. That's exactly how we envision Kaibu House technology in terms of supporting. And we see our technology touching the patient across the entire user journey, right from diagnosis where patients and families are like so overwhelmed. They just got this life-changing diagnosis. Now they have to learn how to live with diabetes at home. And then they go home. And in our system, we definitely need to adjust their insulin frequently, but they're communicating with us by email or over the phone. And it's clunky and it's not timely and it's frustrating. And then they come to their first visit and there's information everywhere, but it's not all in one place. And so it feels disorganized. And that just ends up creating more distress for this already stressed out family. They see us every three to six months, and it's this episodic kind of care paradigm, right? I see my diabetes team every, let's say, four months, but what happens in between? And how do we enable a more continuous care paradigm? And then they transition into adulthood and they get to take all this information with them. We've leveraged the technology to optimize their transition, either through education or through connection or through transition readiness surveys or whatever it might be that we can leverage the technology for to prepare them. And so we see our technology touching the patient and the family across all of these time points in their journey, so that we can provide the best possible care across the lifespan.

SPEAKER_02 15:59

So I'm curious over time, as you introduce high-boo health, if you look at the total cost of care for a child who is diagnosed with type 1 diabetes and then grows into an adult and lives with type 1 diabetes their whole life, if there is that really rocky transition in young adulthood where they, you know, that 70% number that you threw out there, if they don't go see regularly their provider and they're not transitioned efficiently, if through digital means you're able to reduce that rockiness in that transition from pediatrics to adulthood, I wonder if you could bend down the cost curve and improve the clinical outcomes that you see amongst these patients who are diagnosed with type 1 diabetes in childhood.

SPEAKER_00 16:42

Yeah, absolutely. And when we did research in our province of British Columbia looking at DKA, we found that the most number of DKA events occurred in the 15 to 19 year age group. And so if we could move the needle on that number, just that would bend the needle in terms of driving costs down at a very vulnerable time for these young people.

SPEAKER_02 17:08

Super cool. Dr. Ahmed, thank you so much for joining us. It's always fun to hear about all the things you're working on and really excited to think about longitudinal care and how digital health can help translate all the learnings from pediatrics into adult care for the same patient in this case of diabetes. You've probably seen him on LinkedIn weighing in on really interesting topics relates to digital health, artificial intelligence, the electronic health record. He is a thought leader as we think about healthcare innovation. And he is a leader in Canada as we think about how health IT intersects with care delivery. He is such a fun person to talk to. He's thinking five steps ahead of the rest of us, and he's got a really deep understanding, particularly on the provider side, on how startup companies can be effective in how they deploy solutions into healthcare systems. Dr. Josh Liu is my next guest, and I hope you'll enjoy this conversation with him. I'm Caram, excited to be here. Thanks for having me. Were you at Vive? I was at Vive. What was your take on all the things that you saw there?

SPEAKER_01 18:22

There was no shortage of AI. If you weren't an AI company at Vive last year, you were definitely an AI company this year. I think what is interesting is that at least on the provider side, which is more in my wheelhouse, we have started to move beyond just the scribing piece. So I was excited to see that AI scribes weren't the only thing being talked about. We moved on to some other things. A lot of folks are working on AI for care delivery and so forth. So I'm glad we're moving on. I'm glad we're trying new things.

SPEAKER_02 18:50

Why do you think the AI scribe wave was the way that it was over the last 12 to 18 months?

SPEAKER_01 18:56

Part of it was it was just an incredible low-hanging fruit use case. I think one piece is that AI is a very umbrella term, but when people think about AI in the last few years, they think about the chat GPT moment in November of 2022. That's particularly a generative AI. And I think what we found is that AI has been really good for reading and writing summaries of information. And that was a perfect use case for AI scribes, right? Lots of burden from having to document and take notes after a medical encounter. And low risk, it's more of an administrative thing. It's not an actual medical treatment thing. And so it was a very low-hang fruit, high-value use case for healthcare teams and healthcare systems. But as I'm sure you've seen so far, all the other use cases of AI that we can imagine, predicting risk, autonomous care for patients, anything outside the administrative realm, we haven't quite seen the same level of adoption just yet for good reason.

SPEAKER_02 19:54

As we look at the next use case after Scribe, are we still going down the administrative route, you think, with some additional use cases, or do you think we get into clinical and care delivery as the next big moment?

SPEAKER_01 20:05

There's a lot of work being done on AI for care delivery and direct patient care. But in terms of what we'll see take off in terms of broader adoption, I still think more of the low-hanging fruit on administrative use cases for AI to read and write more stuff. So whether that's chart summarization with AI, discharge summaries, post-visit instructions for patients summarized with AI, anything that we can just summarize information, I think you'll see a lot more proliferation of that first before we see proliferation of the more patient care use cases.

SPEAKER_02 20:36

I think there's a pattern in any technology adoption, but particularly with AI, you're seeing this as a good example, where adoption in a specific vertical industry can follow what's happening broadly across a consumer base in their personal lives or in their personal consumption of technology. So to your point, individuals started using Chat GPT as a chat bot that could synthesize information and could reliably retrieve information. And so they felt comfortable when they were doing that professionally with their notes and their summarization there. And so it's logical that you could extract that to say, okay, now let me summarize my clinical information, my chart, or the information I need to look up for perhaps let's say open evidence or something where you're trying to figure out what's the best practice for clinical care, what's the latest clinical trial that's out there for this unique type of condition that my patient has. What do you think about the voice layering? There's a lot of voice AI companies out there and a lot of use cases that are being thrown around there. Do you buy in the hype?

SPEAKER_01 21:41

I do for the right use case. Anything right now where there's a ton of phone call back and forth, it's relatively low risk. It's certainly taking off in healthcare. Inbound calls to call centers at health systems is another low hanging fruit use case, especially for voice AI. The risk around scheduling, rescheduling, I think that's a very good one. Not just for like large organizations like hospitals like yours, but even lots of outpatient practices are doing this now too. It just makes a lot of sense. And I think to your point, we learned on that from other industries, right? Before healthcare, folks were doing both voice and chat-based conversational AI for retail, telecom, airlines. So I do think it's a good idea to understand like what's the consumer behavior with AI and technology more broadly. And then that will be a signal for, okay, which of those use cases have a direct line into healthcare? And then that way we're not having to totally test the concept. We already know our patients or consumers are using these things. Let's just carry on with that wave and just bring that into the healthcare context.

SPEAKER_02 22:41

So let's say somebody out there is listening right now and they've got a technology that they've built that's not a scribe. It's not one of these already big use cases that's out there. It's something new and novel, but it's an AI. And they're trying to get it into the hands of a doctor or a health system. Let's play the what to do, what not to do game around how best this listener can get their technology into the hands of the health system who ultimately is a customer. So what are some things to kick it off that are things they should be doing?

SPEAKER_01 23:08

I think a couple of things are centered around mostly around a trust. And so I'll say two things. One is trust, and I think the second is value and ROI to an organization. So if it's like a big healthcare organization, like a hospital or a health system, how are you gonna build trust in the AI? And trust can mean different things, I think, depending on the use case. So if it's a very clinical use case, it's AI that helps an organization predict risk of medical issues, or it's being used to directly impact patient care, folks are gonna look for evidence around feasibility, safety, and outcomes. How do you obtain that data and produce it? So you can go, you can get trust from the frontline clinical folks and the patients that are going to use it. Trust from a security point of view, do you meet the long list of requirements, SOC2, high trust, all those things, HIPAA, et cetera, et cetera? Whereas if it was a pure administrative thing, like call center AI, it's less about the clinical risk and it's more about does it just solve the value problem? When it comes to like the ROI equation, I think you have to deeply understand not just the frontline clinical staff metrics, but I think a lot of folks fail on how does this translate to ROI for the person writing the check, which is often not the clinician, it is for the COO or someone else. The reality we've all realized over the last couple of decades or so in healthcare technology is a clinical ROI doesn't necessarily always equate a financial ROI. Sometimes reducing re-admissions too much does not improve the bottom line. And there's tension there with the finance folks. So I think deeply understanding, okay, beyond the clinical ROI, how does this impact the financial bottom line for a health system? Because innovators have to realize that margins are really tight. And if you can't paint a really good story on how you lift the financial piece, it's going to be tough. It's obviously easier to do that if you're something that increases patient acquisition or revenue, and harder to do if you're something that just improves outcomes. I don't have a good answer for that part either, though.

SPEAKER_02 25:04

There's two things to follow up on there. One, on the last piece around cost, what I'm curious about as we get into the next couple of years, so many of the use cases that I see are not necessarily replacing existing expenditure as it relates to technology. So you're inflating your already somewhat large technology costs. There are some use cases where you could argue there's other cost savings that are there, but they're not as obvious, perhaps. It feels like there is more to be spent, at least before material revenues are generated, which then again squeezes at these margins that these health system providers have, and the margins are already pretty narrow with reimbursement pressures that are going to continue to be existing. So that's one question that I think we need to figure out is without replacing something that this just a better widget on, how do you justify the spend? The other thing I want to get your thoughts on, you mentioned around trust. So trust around, let's say it's a clinical use case of AI, something that's going to impact decision making, clinical decision making, or perhaps provide patients with information about their clinical care. How would a startup create evidence that their product is trustworthy? The old model is you do a trial, a research study that meets rigorous research criteria, but that takes time. And a lot of these AI products are coming so fast. Is there another way that we're going to imagine how technologists can reliably demonstrate the evidence of their product in a way that's somewhat objective without having to go through the rigor of a peer-reviewed trial or peer-reviewed research study?

SPEAKER_01 26:47

I'm not sure if too many of the dynamics or rules change in the age of AR in terms of the importance of evidence to build trust. I do think what I've found being in digital health for north of a decade is that the amount of the quality of evidence that folks need is very variable. So you could have the very serious academic physician who says, hey, if there isn't a very large RCT on this technology, I'm not going to believe it. And yet you can have health system executives who, as long as there's a hundred patient sample showing benefit pre- and post-intervention, they're very happy with that and they'll believe it. And then you've everyone in between. And so I think a lot of it actually more depends on the culture of the organization. There are some that are highly academic and need a lot of published RCTs, and some who say, hey, if this the data looks reasonable in a setting similar to ours, that's enough for us to at least try it in our own environment and see if it works for us. And I think that's a function more of the culture of an organization and less about the times that we're in. I do think though that we will see rapid advancements in AI for patient care more quickly on the independent practice side than we will on the large enterprise health system side. So you can imagine in a large organization, lots more checks and balances, formal AI governance committees, for good reason, more steps to get through. You can imagine there's going to be some direct-to-primary care clinic that's primarily virtual and asynchronous, where the doctor who runs it is going to vibe code his own AI clinical agents for his patients. And he is the AI governance committee. So if he or she decides to do it, it's going to happen. And hopefully they're doing it thoughtfully. But you're going to see that happen way more quickly, I think, in these smaller practices that are just very AI forward. And probably lessons from that are going to impact what ends up being adopted in the larger enterprise space.

SPEAKER_02 28:38

Yeah. And to a certain degree, ultimately that doctor is responsible for the care they're providing, which includes the tools they use to deliver that care, whether that's a set the scope or an AI tool. And they're basically responsible for making sure that the tools are efficient. So it'll be an interesting landscape, particularly for those vibe coding docs out there that are trying to do what's right for them, look at taking on one more patient every day or maximizing their time or whatever the case may be. What are some pitfalls on the other end? So, what are some things that a startup or an early technologist should avoid as they try to get their AI product into the hands of doctors and hospitals?

SPEAKER_01 29:13

One of the things I think has definitely changed is when I first started in digital health about a decade ago, we didn't have chief digital officers. We didn't have CMIOs only and CIOs only cared about optimizing the EHR. They didn't care at all about consumer-facing technologies or things like that. And the old playbook, which is the pitfall would be trying to do this today, which it doesn't work. The old playbook was you just have to find a clinical champion. They would bring you in. No one else cared about digital health back then except clinicians. And so they were the early champions for digital health in general for patient care. And then since 2020, there's been uh an evolution in how decisions are made. Now, because things are so much more complex, AI can impact folks across the organization. You have these centralized committees, you have chief divisional officers, you have CIOs and CMIOs who care about anything beyond just the EHR, you have to actually build trust and alignment with those folks. The positive there, though, is that it's no longer about siloed niche innovations. Now innovation is looked at through a much more holistic lens. I do think one of the pitfalls, which I don't have a good answer for, is what that means is the more niche your innovation is, the harder it is to get adoption. So for example, so you're at children's hospital. So this is actually not so much unique to you, but if you're in a, let's call it a non-specialized pediatric institution, a generic health system, which has like all kinds of clinical areas, and you had a very niche, niq you AI tool, it may be very hard to get that approved in a large institution because you won't be seen as something that can help a large majority of clinicians and patients. Whereas if you were an AI scribe that can help every department, all of a sudden, oh, like this is a broad tool. So I'm not saying don't do it if you have a niche problem you're passionate about, but I think a pitfall is not recognizing how difficult it is to get buy-in for something that does not have enterprise applicability. And I'm curious to get your take on that because I'm sure you get lots of like niche, like pediatric cardiac innovations and all kinds of niche things. And it's hard for you probably in your role to think about how do I actually support something for every department.

SPEAKER_02 31:17

It's a really good point. So within a children's hospital, you've got to figure out where the biggest impact's going to be because ultimately it takes resources, it takes time, it's an opportunity cost because you're choosing between what to spend time on this project versus other projects on. So especially these days, you really want to pick your projects well because you want to ideally maximize the impact of your portfolio. You want to be able to say at the end of the year, look, these are the four things we did, and look at all the impact we generated. That impact could be financial, it could be around revenue or cost savings or whatever the case may be. It could be around clinical quality and patient safety and patient experience. Look at all the patients that we impacted. But I personally love solutions that impact large groups of people. There's something powerful about products that can impact just large volumes of people. The other end is even if it's small, something that can have a dramatically different impact on whatever the outcome is, whether that's a sizable, it's a life saved or the experience is so much dramatically better. So it's not incremental change, it's sizable change, perhaps a sizable financial benefit. So those are the things that I think I look for. So if it's a small population and a small, small incremental value, it's likely not going to be there. And that's where the evidence piece comes in too, because if there's promise for something to have a big impact, but if it's never been tried before and it's going to take a sizable amount of resource from my organization, the political capital that it'll take to get something off the ground, I'm not sure I want to take that swing because I would rather go after something where I feel like there's some evidence behind it already. And that's different than it was five years ago or seven or eight years ago. As margins shrink as we think about making choices in this space, you've got to be balanced in your portfolio as an innovation leader in a hospital because you want to make sure that at the end of the year there's data to show the impact.

SPEAKER_01 33:22

A decade ago, a KPI folks may have had in the annual report was how many new projects did we launch? How many innovations? Now there's a higher bar, which what was the impact? Sounds obvious now, right? But back then, you're right, when margins were better, people were excited to try and fail lots of things. It's also time to impact too.

SPEAKER_02 33:39

So it used to be you could have a year, maybe even a year and a half before you really saw that ROI realize. Now you want to see ROI 90 days from go live, six months from go live, which is hard. It means you really want to bet on things that have strong evidence that they're gonna work, they worked elsewhere. So it makes it hard for that very first pilot, that very first initiative to take hold. But then the other is you may not want to take the moonshot risk projects either, right? Because if you put all this effort into it, the other reality is a lot of these AI tools require some deep integration with your EHR infrastructure, andor they require data, which is also an important asset that a provider will govern. And so those are not easy lifts. As you think about the projects that hospital wants to prioritize, those are things that they're gonna think about. How much data do I have to provide? What is the requirement around integration workflow-wise? Is this gonna disrupt other clinical workflows, other things that are happening? Those are all things that we think about.

SPEAKER_01 34:45

It's interesting. As you're describing that, I'm also imagining almost like a reverse of what a VC does. So VC investor says, hey, I have a portfolio of 10 companies, they're all moonshots. If only one of them hits out of the ballpark and returns like way more than the value of the fund, and and then I'm happy. A couple might just return the investment, eight will completely fail, and that's okay. But in healthcare, I don't think we're at the point where we can say, hey, we're gonna do a 10 moonshots. If one works out and we're okay with the fact that the others failed or really had no impact, we don't have that sort of uh buffer in healthcare to say, hey, one in 10 works out great. We have a higher bar than an investor would.

SPEAKER_02 35:23

Yep, totally. That's exactly right. The other elephant in the room is the EHR itself. So Epiconcerner and other EHRs that are out there, they're developing their own AI tools as well. How do you think about that as it relates to other solutions being developed at play and showcased at places like Vive and Health and other conferences where they're out there, but it's possible that an EHR may replicate a version of this at some point in the future. How does that play into the whole landscape around technology development and implementation?

SPEAKER_01 35:56

I have a few thoughts on this in general. I think one is it took me a while to realize that it's not enough to have a product or innovation that's two or three times better than what an EHR can produce. You you really have to be 10 times better if you're a third party, and it has to be 10 times better on some dimension that the health system really cares about. So if you have 10 times more features, but the health system doesn't care about all those features, it doesn't really matter. And innovators don't realize that the bar to exceed is that high. Now, in terms of like where the EHRs will shine, where they won't shine in the age of AI, anything that's predicated on the data being stored in the EHR and using that data, I think particularly on summarization with AI, is where the EHR would just have so much of an advantage. There's a reason why they're the ones who are winning with chart summarization, because it even if a third party does summarization two, three, four times better, it's too much of a lift to have that. What's compelling about what we'll see with AI scribes is that the EHR is owning the AI scribe space would have been a very natural thing to expect. But some of like Epic didn't really invest in that directly early on. Now they are, obviously, they came up with their own AI charting. But I think what's interesting about scribes is it feels like it's not as simple as just having a scribe, right? We're seeing that you need personalization for different specialties, you have to somehow support multiple languages. We're really good at outpatient scribing, but if you look at inpatient scribing, it's a lot more complicated because inpatient staff are not verbalizing everything they're doing to support a patient in the hospital. And the question I have is given all the complex customizations and change management needed, even just for scribing, which sounds like a simple use case, are the AHRs going to invest the change management, the customization, et cetera, for all these different use cases? And if they are, how long will it take to get to the point where it's as good as currently provided by the third parties? And I think that would be very telling for the landscape because for something that everyone assumed that the HRs would just own, I'm not sure it's that simple as it turns out.

SPEAKER_02 38:01

I agree with you. I think the scribe is a great use case or great example as we think about your point. The benefit the EHRs have is they have a ton of resources and a ton of access to the data and the infrastructure of care delivery in these hospitals, right? So they've got that going for them and decades worth of experience. The flip side is they've got to be able to provide their solution for a wide range of services. They're not gonna come out with a tool that's just for a very specific niche in a specific market. These EHRs have to be robust in terms of how well they can address all the specialties that they serve, all the hospital types that are out there, rural community, academic, government, all the things, right? So to a certain degree, it's hard for them because they're gonna really need to perfect tools that are really robust and have wide breadth. And that's sometimes harder to do. Whereas if you're a startup coming in, you can innovate and take whatever path you want to get to scale. Whereas I think the EHR folks have to have a slightly more difficult challenge because they are serving the entire health system already. So it'll be interesting to see how with new use cases they they build product.

SPEAKER_01 39:11

I think one neat thing though is that once the note is made though, I feel as if the HRs have a massive advantage. Because I was looking at some of these ambient tools lately, and now as they're all incorporating clinical decision support where you describes open and you could also ask questions about what's the latest evidence for this treatment plan. And that's great if maybe you're asking those questions during the encounter. But the moment that you finish the encounter and you're back in the EHR, you're doing all the other things you have to do, I don't want to pull up the scribe just to ask a question. I'm gonna go default into the EHR chat bar and type my question about what's the latest guideline for treating this patient's heart failure. And I'm going back to the patient chart. And so one of the things that I wonder is as much as the ambient companies want to add all these new features in, unless I have a reason to stay in the ambient interface, why would I be using your additional tools to do all those things? I don't think they have a good answer just yet.

SPEAKER_02 40:03

They don't, and that's gonna be the ultimate question as we think about fragmented delivery of all these new AI tools. Josh, it's always fun talking to you. This was fun. Thanks, Amkar. I've always found that emergency department physicians can be incredibly innovative. And Dr. Devin Singh is no exception to that. He is a leader at Sick Kid Children's Hospital in Toronto, and he has really spent his time focused on how particularly artificial intelligence, combined with operational workflow management, can really improve wait times, care delivery gaps, and the experience for families in the emergency department. He is leading all sorts of amazing efforts. There's been a recent story about how he's leading some collaborations between the Canadian health system and the National Health Service in the United Kingdom. And so Devin Singh is my next guest, and he is an internationally recognized healthcare innovation leader who is driving some impressive optimizations in the way patients get emergency room care in Canada. I find that emergency room physicians tend to be very innovative and have lots of ideas around how to solve problems. Why is that the case?

SPEAKER_03 41:22

Well, I'm gonna take the compliment first of all. Um, so that's really generous of you to say, but I think this happens out of necessity. If you think of what it means to be an emergency physician, and particularly a pediatric emergency physician, if you don't get things right, the outcomes can be absolutely devastating. And then combine that with the time pressures of increasing wait times, lots of patients waiting in a waiting room, you're almost by necessity forced to innovate and try to figure out in this constrained health system that we all work in, how do we do better? And I think that's probably why you see innately in in most emergency physicians this desire to push the boundaries and to try and find solutions. Cause we, at the end of the day, we're all really trying to provide the best care we can and the most timely care we can to our patients who are waiting.

SPEAKER_02 42:12

Yeah, you've been really focused on using artificial intelligence and data and analytics to help improve so many different things, process and outcomes-wise, in your emergency room. Can you talk more about the work you're doing?

SPEAKER_03 42:24

There's sort of two hats that I actually wear, which are which is kind of interesting as it relates to this. With me as like a hospital innovator hat and an AI researcher, I'm really focused on trying to figure out are there ways where we can leverage clinical automation and real-time AI modeling to advance patient care forward. So, for example, we've got this incredible research project where we're using triage data from the electronic health record system in near real time to predict things like might a child have appendicitis? Might they need urinary testing? Might they have a broken bone that needs an X-ray, or maybe a condition that needs an ECG? Thinking of these low-risk testing modalities and these diagnostic tests that we often order in the ED, and yet a patient will wait four, six, eight plus hours if it's a really bad day to get those tests ordered because you us as the human providers are just busy and constrained. And so we've really thought through can we use machine learning algorithms like logistic regression, random force, XG boost, or deep neural network models to actually make that prediction. And then we challenge the ecosystem to say, if we can predict this really well with high precision, which means low false positives, should we not just order the test? Like, should we not just automate that test order for that patient and drive and accelerate care forward? And if you think about what I'm saying and you step back, I'm basically saying is it okay to give the medical legal authority to an AI algorithm to provision care for a child without a human in the loop besides the algorithm and that patient and family?

SPEAKER_02 43:58

And on the positive side, if you Do that, you're able to reduce the time that people are waiting for a test or even their overall wait time in the emergency room. On the positive, you're also creating consistency and perhaps making sure that best practice is always followed. So there's there's a lot of positives to be had there if you're able to give that type of authority to the AI.

SPEAKER_03 44:19

That's absolutely right. If this patient is going to be sitting and waiting, anyways, why not get testing going? And if you think about it, this is an absolutely radical. We have nursing-based medical directives where after triage, a nurse may order a urine test. But those things don't exist as commonly in the pediatric world as they do in the adult emergency department world. And we actually don't see medical directives for things like abdominal ultrasound for roulette appendicitis necessarily. And so what we're really asking for and challenge ourselves with this research project is if we do give that authority to an AI model to automate ordering that test to just get it done while a patient is waiting, might we get that patient to diagnostic imaging and test results faster? If we get them to a diagnosis faster for appendicitis, can they get to their OR faster, treatment faster? Can they leave the emergency department faster? That's really powerful because for that individual patient, they get their answers and they get to treatment faster. And ultimately that's gonna drive safety and quality of care. But it's also really powerful for that other kid, that other patient who's sitting in the waiting room who needs that free bed space to open up. We've done some analysis where even just five really low-risk common use cases with high precision, which means you're gonna get maybe 10 to 20%, maybe up to 40% of all the positive cases. Even that alone could actually automate just under 24% of testing for patients who come to our emergency department. That's gonna be a game changer for safety, for flow, and efficiency.

SPEAKER_02 45:50

Now you've translated some of this information into hero AI. Tell me more about that.

SPEAKER_03 45:56

Yeah, that's right. And so in the early days after my master's of computer science that I did at U of T, we started to build these algorithms. And very quickly, I realized this is gonna work. Like the models actually work incredibly well. We have validated them over hundreds of thousands of retrospective records, and they've been prospectively deployed, but in a silent trial for almost two years now. And we see that the models work incredibly well. But what we were missing in the early days was an ability to actually deploy the thing. We didn't have the platform to get real-time data to flow into our model at the time. And and more importantly, we didn't have the platform to actually translate the model output into the hands of a patient or into the hands of a clinician or a provider. My main passion is changing care. Like I really want to see this stuff help kids. And so I knew early on that the publication wasn't gonna be enough. We needed to find a way to translate this stuff to the bedside. And that's what Hero AI does. So one of the incredible use cases that Hero AI has become a bit famous for is around the impact we've had on mental health. And so, right now, live in the emergency department at Sick Kids Hospital in Toronto, if sit and wait two hours, four hours, maybe eight hours on a really busy day, sitting in a waiting room, waiting, having a mental health crisis, rather than having that long wait time, our AI sidekick can review the charts in an automated way, recognize that the criteria have been met, and just automate the consultation to psychiatry, even before they've seen the emergency physician. And so what that's done has been incredibly impactful. It's actually reduced the time from arrival to psychiatry consultation by 55%. On average, we're able to discharge patients home after they've received their psychiatric consultation and interventions by almost 90 minutes to 120 minutes faster per patient. So not only are the patients getting better care and faster care, but again, the system is being given back extra room capacity time. And I kind of joke to my colleagues sometimes because it actually is a reflection of just sometimes how busy we are as emergency physicians. We get pulled away to really acute high-intensity cases and it creates these wait times for other patients. But what an amazing common sense thing to do, a small block of automation, have an AI sidekick, drive that care forward on your behalf so that care is delivered in a more timely way. It is common sense.

SPEAKER_02 48:28

So you're seeing now using this in your emergency room that you're seeing less wait times?

SPEAKER_03 48:35

Yeah, less wait times for really targeted patients. And so one of the latest sets of interventions that uh we've deployed is focusing on other sort of vulnerable and high-risk groups. So, for example, we have a large patient population that will come to our ED looking for help who have sickle cell disease. And so we've set a target as a group that we don't want any patient with sickle cell disease sitting in pain for longer than 30 to 60 minutes. It's an ambitious target to me, but it's really important for us as a hospital and as clinicians that we hit that target. And so one of the things that we're doing is we have another AI sidekick that can help monitor the waiting room and help draw our attention to when a patient enters into that state where they have a history of sickle cell disease, they're having acute pain, they haven't necessarily received the ideal opioid management or analgesia, our system will trigger an action and start a pathway forward. Almost, I see it as like an AI tool and an AI sidekick automating the advocacy for certain patient populations. We've got other examples where if someone comes in with signs of an acute testicle, but they haven't been seen in an hour and an ultrasound hasn't been ordered, and we're starting to deviate from the standard, an AI sidekick will engage and we'll start to advocate for that patient to move care forward. And to me, it's really about thinking through how real-time applications of artificial intelligence in the emergency department adds to our capacity, advocates for our patients, and helps us deliver a higher quality of care with very similar resources and physical constraints that we all have that are difficult to change.

SPEAKER_02 50:12

That is fascinating. Dr. Singh, thank you so much for all the work you're doing and love seeing the application of data and analytics and AI to help improve throughput, turnaround times, wait times, and ultimately patient satisfaction for patients waiting in the emergency room. Oh, it's been an absolute pleasure and thanks for having me on the podcast. All right. Thank you for joining us for your dose of optimism. Make sure to check out our show notes to get more information about our guests and the work they're doing. Visit our podcast page on the Kids X website to join our podcast community and to learn more about pediatric innovation. Thank you to our sponsors and to our presenting partner, Kids X. Please subscribe wherever you get your podcasts. And remember, it takes a village to make sure our kids grow into healthy adults. So volunteer at your local library, help out at the community center, and if you're so inspired, donate to your local children's hospital. Alright, see you next time. The content, views, opinions, and information presented on this podcast do not reflect the views of Children's Hospital Los Angeles or of the sponsors of the podcast.