Disruptors

Can AI fix ER wait times?

SPEAKERS

Mara Lederman, Amol Verma, Fahad Razak, John Stackhouse

John Stackhouse 00:00:09

Hi, it’s John here. If I ask you about Canada’s healthcare system, you probably don’t have to think too long about the answer. You might be grateful and, sure, you’re probably appreciative, but you probably also have a fair number of frustrations. This is a pretty familiar Canadian experience. We value our public healthcare system, we cherish the promise behind it, but healthcare in Canada involves a lot of waiting.

Earlier this spring, I had one of those experiences myself. A family emergency led to a long wait in a hospital emergency department. Now, everything worked out in the end, but I kept thinking about how common that experience has become. In that particular hospital, the first thing we saw when we entered the emergency department was an electronic board telling patients the expected wait time was seven hours and 32 minutes. Where else can you go where the wait time is advertised that boldly with no apologies?

Nearby, another sign asked patients to speak up if they did not want AI used during their visit. This May, the federal government released Canada’s national AI strategy and healthcare was the central priority. If you saw Prime Minister Carney roll out the policy, you’d have noticed by no coincidence he made the announcement in a hospital. And why not? Canada has a public healthcare system, world-class research talent, and hospitals generating hundreds of thousands of data points for every admitted patient. All that can help us develop better AI systems that can lead to better healthcare. But none of it, of course, is easy.

So on this episode of Disruptors, we look at the challenges from two sides. First, we’ll hear from Mara Lederman, an economist by training, and co-founder and Chief Operating Officer of Signal 1. It’s a Canadian health AI company working with hospitals in Canada and the US. Mara helps frame the pressure inside the system, rising demand, constrained supply, and the practical ways AI and technology can reduce wasted capacity.

Then we’ll hear from Dr. Amol Verma and Dr. Fahad Razak, two physician researchers connected to Unity Health Toronto. That’s the healthcare system built around St. Mike’s Hospital in downtown Toronto. They’ve built an AI infrastructure platform called GEMINI and are now helping lead VITAL, a more robust data infrastructure platform that aims to connect healthcare innovators across the country. VITAL just received a landmark federal investment of $100 million as part of a combined $200 million commitment that is the largest data innovation investment in Canadian history. Together, these platforms are aiming to reduce those horrible wait times that so many of us experience.

The opportunities here are twofold. There’s the operational challenges of wait times and then there’s the research that can lead to medical discoveries that life itself depends on. On both counts, Canadians want to be the world leader and here’s our shot to do just that. Here’s Mara Lederman. Mara, welcome to Disruptors.

Mara Lederman 00:03:19

Thanks, John. Good to be here.

John Stackhouse 00:03:21

I want to say actually, welcome back because we’ve had you on before and I want to kick off with that, Mara, because I actually can’t remember how many years it’s been, but we’re kind of talking about the same challenge. What’s going on in our hospitals that continues to lead to the levels of frustration that I think we all recognize across the country, even though we’ve all spent a lot on digital technologies over the last many years to ease those pressures?

Mara Lederman 00:03:49

I think we take a lot of pride in our healthcare system and our universal healthcare system. And then when you ask people if they feel like they’re getting good care, they’re not. And that’s not because we don’t have great research or great hospitals. I think if you ask any Canadian, they’re going to report being dissatisfied. Our ability to kind of give the average Canadian high quality care seems to be diminishing. We have a complete mismatch of supply and demand. We have incredible demand for healthcare. We have an aging population. They’re living longer with more ailments and more chronic disease. And so the need for healthcare, the demand for healthcare is growing. It’s effectively unpriced and we have a very constrained supply. And the simplest way to think about that is just sort of the number of hospital beds, the number of surgical spots, the number of procedure spots, and the number of clinical staff.

You go to an emergency department, you wait five to 10 hours. You need to see a specialist, people get booked six to nine months later. You need a quote, unquote elective, which is not really an elective procedure, but a non-emergency. You often have to wait months. And the way we’re going to do it is not just going to be by building more hospitals, staffing more beds because we simply can’t afford to do that.

John Stackhouse 00:04:54

And yet we are building more hospitals and trying to hire more doctors and nurses and there’s a logic to that. Why hasn’t digital technology particularly not made a greater difference over the last decade?

Mara Lederman 00:05:08

One of the benefits of adopting technology is it is supposed to make us more productive. What does it mean to be more productive? It means we can kind of generate more outputs with the same inputs, right? So in healthcare, you’d like to ask, how do the same sort of resources, when armed with technology, allow us to create more units of care? Probably the biggest digital technology to hit healthcare is the electronic medical record. The EMR is basically a system of record. It’s not a system of action. So there aren’t a ton of examples where you are busy clinical teams with another piece of technology often on top of the EMR and find that they can just do so much more.

John Stackhouse 00:05:48

That’s the pressure point. Demands keep rising. Supply can’t expand quickly enough and a decade of digital investment just hasn’t made the system move fast enough. At the same time, healthcare is generating more data than ever before. The problem is that too much of that data stays trapped inside hospitals, inside provinces, inside systems that were not built to learn from one another. That’s where VITAL enters the story. Amol and Fahad, welcome to Disruptors.

Amol Verma 00:06:16

Thanks for having us.

Fahad Razak 00:06:17

Great to be with you, John.

John Stackhouse 00:06:18

You’ve had an incredible busy couple of weeks now that you’re center stage in AI in the country in a very positive way. I want to start with a better understanding of VITAL. Amol, let me start with you. Take us through the VITAL story.

Amol Verma 00:06:34

The heart and the core insight and focus of our work is that healthcare does not use data very well. We don’t connect data across our healthcare organizations, and we don’t use it to inform good decisions, to develop solutions that can improve the way healthcare is delivered, and to use it to enable large-scale cutting-edge research and innovation. And so we’ve been working for about 10 years in Ontario on a hospital research network that we named GEMINI before Google showed up on the AI scene with their Gemini model and that connects hospital data now from about 45 hospitals, that’s about 60 to 70% of the patients in Ontario and makes that available for all of those different applications, quality measurement, research, innovation in healthcare. And what VITAL is doing is taking the work that we’ve built with GEMINI and really accelerating it and scaling it larger. So essentially saying we can connect hospital data at large scale across Canada and we can make it much more timely, and make that available so that we can support research and innovation to bring new solutions, new technologies, new investments into healthcare.

John Stackhouse 00:07:41

Walk us through GEMINI and how that connects with  VITAL.

Fahad Razak 00:07:44

So GEMINI was started roughly 10 years ago and the digitization of data, very different state than what it is today. So most of the chart of a patient admitted to hospital was handwritten, printed. It was in binders in the nursing station. 10 years later, almost all of information that’s relevant is now digital. So GEMINI was capturing that progressive digitization over that 10-year period. When we started to look at the world though the last couple of years, we saw that even though GEMINI was at best in Canada capacity, it really wasn’t competitive. What we were seeing in the Nordics, a place like Denmark as I think is the exemplar, the United Kingdom in terms of their ability to harness that kind of digital data at scale and use it for trials for artificial intelligence.

And just to put this in context, we’re both practicing physicians. Every patient that we admit to hospital now, we are generating hundreds of thousands if not millions of data points just to provide care. So it’s the lab tests, it’s the MRIs, it’s the digital vital signs. That data is extraordinarily important for artificial intelligence and for clinical trials and we were seeing these other countries developing population-wide capacity essentially to get that data, to use it very quickly under really good governance, legal protections for broad innovation use cases. And that was not possible with the GEMINI model. And so VITAL was a proposal to kind of leapfrog to push forward to say, “We need this at scale for the country. We need to be as competitive or better with the best-in-class we’re seeing globally.”

John Stackhouse 00:09:13

So that gives us a good sense of the infrastructure and the data. So taking us into 2026. One of the many challenges here in Canada is that healthcare is provincial. And I think you were just with deputy ministers from the provinces walking through what this new national infrastructure can do. Great in theory, but we all know the challenges of getting anything across provincial borders, including data, including doctors, including other aspects of healthcare.

Fahad Razak 00:09:42

Yeah. So I mean, let me start with your first point, John, which is the governance and political structure of Canada has made this kind of collaboration difficult. So a reasonable question is why is today any different? And I think today is different for two key reasons. The first is that we are under enormous geopolitical pressure. The use of this kind of data at scale for the country is an important value proposition where Canada can be truly competitive. We have to harness as close as possible the data of 42 million people, again, responsibly with the highest levels of privacy protection, but at that scale. If we are able to do that, there is a competitive advantage that we have at scale and that is that we are a single-payer system from coast to coast. So we don’t lose people like the United States does who don’t have insurance in their datasets.

And we are the most diverse high income society on earth. And that means that if you do your science here, if you do your AI algorithm here, your clinical trial, it’s intrinsically better than if you do it in a homogeneous population than Denmark, for example. We look more like the world than any other country. So the algorithms are just fundamentally better if done here. The second important advance is what’s been the rise of federated analytics. The data of Alberta today and five years ago, we can’t just move that data into Ontario. That’s not allowed. That had been a barrier five years ago to doing this kind of analytics, let’s say developing an AI algorithm or running a clinical trial, but we have the analytic structures that can run and develop an AI algorithm across these provinces and territories without actually moving their data. The algorithm optimizes in each jurisdiction and it keeps optimizing and rotating until it converges, as an example.

So you have a technology solution, you have the geopolitical pressures, you have the recognition that no single jurisdiction is big enough to really be competitive and that you’re seeing the offshore of the opportunity. So what is physically left Canada? An early cancer screening trial has gone to the United Kingdom. The best new vaccine trials are going to Denmark. And I think over the next three to five years, you will see some of those trials and AI developments come back into Canada because the combination of the structure that VITAL offers and the competitive advantage of our population, population dataset, size, diversity means we become a best-in-class jurisdiction to do work.

John Stackhouse 00:11:56

That’s the national infrastructure story. Connect what has been fragmented, protect the data, and turn Canada’s health system into a platform for better care, faster research, and better tested AI. But for patients who are still waiting, the question may be more immediate. What can AI actually do for me right now? Mara Lederman points to one simple example.

Mara Lederman 00:12:18

With the rise of GenAI and agentic AI, we’re seeing a whole nother set of applications, many of which aren’t yet patient facing and many of which aren’t even physician facing that are leading to efficiencies. Many of your viewers, if any of them are over 50, have had a colonoscopy, right? It’s an unpleasant experience. We all know that. And if you’ve gone through it, you know that there’s a whole round of prep you need to do in the days leading up to your procedure. You get detailed instructions about various medications and foods you need to stop eating or taking up to a week before. If someone doesn’t follow their preparation, what happens? They show up at their appointment, they get asked if they did the preparation or if they stop their iron pills, they say, “Oops, I forgot to do that.” They get canceled, no one gets a scope, you don’t get anyone off the wait list and it’s just a wasted opportunity.

So one of the health systems in the US has built an AI agent that automates these pre-procedure phone calls. And what it does is it calls everybody a week before, it goes into their medical chart, it pulls their list of medications. It says, “Here are all the medications you’re on. These are the ones you need to stop. Here’s a reminder about your prep. Here’s a reminder about when you need to come.” And it can do this for everybody. And then after these phone calls happen, a different AI reads the transcripts and flags any that feel they need to be redone by an actual nurse and the data shows more people come prepared, fewer scopes are canceled, fewer unused opportunities. So from a quality perspective, it’s better for patients, from a staff perspective, they’re not standing there with no patient and from an access, you’re not wasting opportunities to move someone off the list. So simple example, and it’s not that high risk.

John Stackhouse 00:13:57

That example shows what AI could do at the edge of care. Fewer wasted appointments, better follow up, fewer avoidable returns to the emergency department, and more support after a patient leaves hospital. But in healthcare, promising is not enough. We need to know whether these tools actually work and whether they work for Canadian patients in Canadian hospitals across Canadian communities. That brings us back to VITAL.

Amol Verma 00:14:22

AI is at an exciting point in its inflection. We need to run clinical trials on AI tools and we need a capability to do that. And the capability required to do that is an integrated digital infrastructure because unlike a medication where you work through a pharmacy, you dispense the medication, there’s a structure and infrastructure already in place for all of that. What we don’t really have the infrastructure in place for is the digital dissemination of these AI technologies. The second important application of AI in the context of clinical trials is AI can make it much more efficient to run other kinds of clinical trials. Large scale clinical trials historically have been a very expensive proposition. The average cost of a phase three clinical trial, which is what you need to get regulatory approval to bring a new drug to market is 21 million US dollars for a clinical trial.

If we can make that more efficient, we can create an extraordinary opportunity to bring new technologies faster to Canadians and make these kinds of technologies available to more people and AI can help us do that. It can help us do that by one, making it easier to collect the data you need to tell if a clinical trial was effective or not. A lot of that data is already gathered in our healthcare system. So rather than phoning up a patient and saying, “Did you end up back in the hospital?” We can just gather that information routinely to measure the outcomes you need to, to study a clinical trial.

The second piece is it can make it easier to identify which patients are eligible for clinical trials. So a big cost in clinical trials is identifying eligible participants, recruiting them into a clinical trial. AI can help us look at all of the health record data and identify which patients might be eligible and then we can work through typical processes with patients, care teams, and otherwise to try to make those trials more available to people. So lots of opportunities for AI to improve clinical trials and for clinical trials to improve AI.

John Stackhouse 00:16:12

Trust underpins all of this. And as we were discussing earlier, trust generally has declined when it comes to AI, especially here in Canada as well as the United States. How are you thinking through the trust barrier in rolling this out?

Fahad Razak 00:16:27

Yeah, I think it’s a problem that Canada is actually in quite a deep deficit around, as you said, John. So we know from surveys that in fact, the Canadian population’s trust of AI and worry about AI is actually higher than many other countries. And I would say the first point to make is that there are good reasons to be worried about AI. And so I do think that we have to be very cognizant and respectful of these worries and say, what is our solutions? What are we specifically doing? So let me talk about a couple of them. The first is whose data is being used under what permissions? We have a precedent in this country. The kind of data that we are talking about has already been used for generations really in this country in highly respectful protected ways that protect individual rights that allow opt-outs if people don’t want to be part of these innovations, that have layers of protection to anonymize around who an individual is.

We’re talking about using these aggregated parameters to develop an AI algorithm in a way that is consistent with the historical accepted pattern of the use of this kind of data in the country. However, there are some particular nuances around AI. There are ways that AI can develop an algorithm that can unmask an identifier. I think we need to be upfront about recognizing what those algorithms potentially can do, making sure that those edge use cases, not the majority of AI, but those edge use cases are ones that we are particularly careful around because I do think that we’re starting with a very understandable level of distrust and we can’t allow that deficit to worsen. So the second is that again, we are the most diverse high-income society on earth. We know from other AI deployments that these algorithms will anchor and tailor and optimize very tightly to the population they’re developed in.

So if you have an AI algorithm developed in a Nordic country, which is 95% homogeneous and you try and deploy it in Brampton, Ontario, or in Northern Ontario, or in the none of it, the chance that it’s going to work as well or even approximate as well as what they initially claimed is probably pretty low and people will pick up on that. And I think we need to be very careful to ensure that these algorithms are being deployed in our population in a way that is fair, and unbiased, and actually shows benefit.

And the third important thing is I was quite struck that the government decided to lead their AI strategy with health. Even a year ago, health may not have been one of the main pillars of an AI strategy for the country. I think it is more than just symbolism. If you speak to the Geoff Hintons and the Richard Suttons and others, they will say that of all the areas where they are worried about AI, health is probably the most important sector where they think the benefit can far exceed the risk.

The ability to detect disease early, to provide more nuanced diagnosis, that is things that people can appreciate day-to-day in their lives. I had a cancer diagnosis detected before I otherwise would’ve because of an AI algorithm that was able to pick it up in the imaging processing of my mammography. So I think as people rightfully have concerns around AI, health could be a really important sector where we can show benefit early, but we have to do it in a way that’s respectful about privacy, about bias. So I do think it’s a very important area where people actually can see real benefits in their lives.

Part of it is also a broader consideration of risk and benefit and how we think about each sector. So let me just use an analogy. I’m in downtown Toronto. I get onto the gardener in my car and you’re in a metal box with a strip of cloth over your chest driving 100 kilometers an hour at each other. We do that every day and we tolerate the risk because we have a destination and not because it’s fun to be on that road. With health data, if the entire conversation is risk containment, you are essentially boiling down your speed limit to zero kilometers an hour. So these technologies have the potential to get us to better outcomes, to better diagnosis, to more efficiency, but it’s not risk-free.

I think we need to have a conversation about tolerating some level of risk, being very disciplined and transparent where we make mistakes, and iteratively getting the safety systems better. Cars today are infinitely more safe than they were 50 years ago. That’s because we’ve iteratively made them more and more safe. There are going to be accidents, disclosures, biases that emerge. Let’s talk about it. Let’s say this was a problem. Let’s make it better. And over the years, we will see a system that is more and more safe, that has more and more guardrails, but importantly gets you where you need to go.

John Stackhouse 00:20:45

The risk conversation matters, but there’s also a risk in standing still. If Canada does not build this capacity, trials go elsewhere, AI tools get trained somewhere else, and Canadian hospitals will have to import systems designed for other populations. And in the end, patients could just end up waiting more in a system already under strain. That’s why Mara Lederman argues this should be treated as a necessity.

Mara Lederman 00:21:10

First of all, I don’t think we should call it the Canadian opportunity. I think we should call it the Canadian necessity. We started this conversation by talking about the health system being in crisis and that we aren’t going to kind of build or hire our way out of it. As long as we call it the opportunity, it feels like pet projects and innovation and we should just say there’s an absolute necessity to use technology to make healthcare sustainable. At a high level, we need to push as much care as possible into the lowest cost setting where it can be delivered. It seems like a simple thing to say, but the default is to do most stuff in hospital. Think about all the care that became virtual during COVID that nobody thought was possible before, even virtual nursing.

I was just with a company last week who’s putting sensors and cameras in patient rooms. All the data’s computed on the edge so there’s no privacy issues, nothing is stored and they are cutting down on fall risk and adverse events that happen in a hospital room that otherwise humans would’ve had to supervise. A lot of US systems have virtual emergency departments where you first even just get triaged online before you even show up in the emergency department. We’re seeing the growth in the US of what’s called ASCs, ambulatory surgical centers. So how many surgeries can be done outside of an acute care hospital as opposed to an inpatient hospital?

AI is not the only answer, but more and more I think we have fallen into the same model of thinking: that most care is delivered in a doctor’s office and if not in a hospital. And the reality is, with the technology we have now, we can do more. And so if I were to focus our country on one thing first in terms of the opportunity, it is asking how we use technology and sort of associated changes in incentives and coordination to always ask, are we delivering this care in the lowest cost setting?

John Stackhouse 00:23:01

Mara, thank you for being back on Disruptors.

Mara Lederman 00:23:04

Thank you for having me. It was fun as always.

John Stackhouse 00:23:07

Care in the lowest cost, safe setting means different things in different places. You can start to imagine AI-assisted tech being deployed and connected across our country, allowing emergency rooms to regain capacity. But also in remote communities, it may mean not needing to travel hours or fly just to get care that could be monitored, managed, and supported closer to home. Healthcare really is the critical testing ground for AI and whether it can deliver meaningful benefits for everyone. What should Canadians expect to see over the next, let’s say, year?

Fahad Razak 00:23:44

I think the one-year outcomes are going to be the introduction of technologies that we’ll be testing and evaluating using this infrastructure. So you’ll see things like, are we going to use an AI prediction algorithm to, let’s say, detect heart disease, to detect viral outbreaks as they’re occurring? These are underway and being tested by our network right now. The reality is as much as we’re optimistic around AI, it’s not magic. So it’s going to require the evaluation and testing that we historically have done for a cholesterol pill, for a cancer therapy. We have to test and evaluate. Those tests are already underway. So if you’re in communities that VITAL stretches to in Alberta, and Ontario, and Quebec, for example, you’re going to see these algorithms start to be rolled out, but it’s under evaluation phases. Again, we want to make sure that both the access to these technologies extends out to Canadians, but it’s done in a way that’s safe and under evaluation.

The announcement last week by the federal government was an additional $ 100 million investment in the VITAL network. To our knowledge, this combined $ 200 million investment is the largest in Canadian history now in this kind of data innovation. That’s about getting it out to the rest of the country. So in the next year, I think you’ll see it extending out to these three provinces, 50% of the country. In the years that follow, year two, year three, you’ll see it getting out to more and more provinces and territories. Again, we don’t want to overpromise, but the idea is we’re going to choose the best of these algorithms. Get them out there, test them, evaluate them, get them out to people, but also watch very carefully to make sure that they’re working well.

John Stackhouse 00:25:10

Fahad, Amol, thank you for being on Disruptors.

Amol Verma 00:25:13

Thanks for having us, John.

Fahad Razak 00:25:14

Such a pleasure. Thanks, John.

John Stackhouse 00:25:17

There’s a lot of government money going into AI, but that doesn’t guarantee outcomes. It does change the scale of what’s possible. Canada has enormous pools of healthcare data. The challenge now is building the infrastructure, trust, and operating capacity to turn that data into better care. VITAL now has the mandate and the resources to extend to Alberta, Ontario, and Quebec within a year and to reach more provinces and territories after that. The national AI strategy put healthcare on center stage. This investment is the follow through. AI in the ER is our new reality and we’re about to see just what kind of difference it can make.

You’ve been listening to Disruptors, an RBC podcast. Follow us wherever you get your podcasts. And be sure to share this episode with someone thinking about the future of healthcare in Canada. That will help more people discover this kind of conversation.

For more RBC thought leadership on AI, healthcare, productivity, and Canada’s innovation economy, visit rbc.com/thoughtleadership. I’m John Stackhouse. Thanks for listening.