Unity Health interview part 2 of 3
In this second interview, Michael Page (Director of AI Commercialization at Unity Health Toronto), shares his experiences of successful AI collaborations.
This is the second in a series (other parts: part 1 and part 3) of interviews with leaders from Unity Health Toronto, including Muhammad Mamdani (Vice-President of Data Science and Advanced Analytics), Michael Page (Director of AI Commercialization), and Derek Beaton (Director of Advanced Analytics), discussing the organization's pioneering efforts in integrating artificial intelligence into healthcare. They share their strategies, challenges, and successes in deploying AI-driven solutions that improve patient outcomes, enhance operational efficiency, and address real-world clinical problems. In collaboration with clinicians, researchers, and technology partners, Unity Health demonstrates how AI is reshaping the future of healthcare.
Read more about AI at Unity Health (unityhealth.to).
Michael Page, what role does collaboration with other healthcare organizations, academic institutions, or technology companies play for Unity Health?
- Collaboration is not just valuable—it’s essential for the broader adoption and integration of AI in healthcare. At Unity Health, a prime example of this is how our data science team works closely with clinicians to address real-world clinical challenges. This deep internal collaboration has been key to our success in deploying more than 50 AI and analytics tools into practice, delivering tangible benefits in patient care.
- While internal collaboration is critical, achieving greater health system transformation will require expanding these partnerships beyond the walls of individual institutions. Collaborating with other healthcare organizations, academic institutions, and technology companies allows us to pool expertise, and resources, which accelerates innovation and the spread of AI solutions.
- For instance, partnerships with academic institutions can drive groundbreaking research, while collaborations with technology companies can facilitate access to cutting-edge tools and platforms that may not yet exist in the healthcare sector. These collaborations help us push the boundaries of what AI can achieve and ensure we stay ahead of technological advancements.
- Moreover, collaboration across the health system through coalition building is critical for scaling AI solutions and achieving systemic change. While much of this happens informally through academic and professional networks, I envision a future where these efforts become more coordinated. Large-scale collaborations between healthcare providers, researchers, and industry players can help overcome challenges like data sharing, regulatory hurdles, and the clinical validation of AI tools.
- Looking ahead, partnerships will also play a key role in the commercialization of AI tools. Expanding our collaborations globally, such as with AI Sweden, enables us to not only enhance our own capabilities but also contribute to a more integrated and AI-driven global healthcare ecosystem. Ultimately, collaboration will drive the future of AI in healthcare—helping us turn innovative ideas into impactful, scalable solutions for patients around the world.
- While collaborations are essential to drive innovation, we carefully consider strategic collaborations that will add value not only to our respective organizations, but to society at large.
“We're fortunate that AI, along with research and innovation more broadly, is a high priority across the organization—from our Board of Directors to the executive team.”
How does Unity Health foster a culture of innovation around AI and data science, and what advice would you give to other healthcare organizations looking to adopt similar technologies?
- Fostering a culture of innovation around AI and data science at Unity Health begins with leadership. We're fortunate that AI, along with research and innovation more broadly, is a high priority across the organization—from our Board of Directors to the executive team. This top-down commitment ensures that everyone is engaged in and aligned with our AI strategy.
- At St. Michael’s Hospital, one of our key sites, we have a long-standing tradition of clinician-scientists who balance patient care with teaching and research. This culture of intellectual curiosity and the drive to continuously improve healthcare is deeply embedded within the organization. Our AI initiatives are a natural extension of this legacy, allowing us to use cutting-edge tools to tackle modern challenges.
- Success also plays a pivotal role in fostering a culture of innovation. For example, our AI-driven tool CHARTwatch, which has been shown to reduce unplanned mortality by over 26% according to recently published research in CMAJ, has been a major confidence booster across the hospital. Success stories like this create a "flywheel effect," where each achievement fuels more interest and momentum. This, in turn, inspires more colleagues to think creatively and collaborate with us on developing the next breakthrough—whether it's a tool for detecting traumatic brain injuries, predicting massive hemorrhages, or reducing the administrative burden on frontline clinicians so they can spend more time with patients.
- The only challenge we face is capacity. With so many exciting ideas coming from our colleagues, we currently have a two-year waitlist for new AI project intakes. I wish we had double or triple the resources to tackle them all!
Could you tell me about a recent AI project that you’re particularly satisfied with and explain how it has improved patient care or outcomes at Unity Health?
- We’re satisfied by everything we do! We have such a thorough process before we dive in that it helps ensure we have very high quality products for everything we do. Some of work, like CHARTWatch, is a fundamentally important tool. Recently we showed that CHARTWatch brought (and kept) ICU transfers and mortality (considerably) down since its deployment. And we’re able to see that because of how long it’s been in place
- Like noted, we don’t have any particular favorites. But we do have approaches across so many domains, each reflecting different types of improvements. There are three in particular we like to highlight.
Our ED RN assignment tool takes schedules, qualifications, and prior assignments of RNs in the ED into account, and will generate assignments (to zones or roles) for the staff. It’s something that takes a person a long time to do, but our optimization tool can make the assignments in about a minute and the lead nurse can make some edits in about another minute or two. It’s a significant time savings
We have a tool running in the hemodialysis unit that highlights patients at high risk for unplanned admissions (e.g., back to the emergency department). We like to highlight this tool for two reasons. The first is that it took a really long time to go from ideation to production. It’s a particularly challenging problem from data (extracted from dialysis machines) all the way to workflow integration and intervention. When we deployed, it felt particularly satisfying that we could bring it across the finish line. Even better is that we saw a substantial reduction in unplanned readmissions for high risk patients (25-30% drop).
Another set of projects have effectively turned into a suite of tools, or really a larger product and platform all in one. Right now we have two medical imaging models for all head CTs that come into the emergency department: one to identify intracranial hemorrhages and the other to flag cases at high risk for urgent neurosurgical intervention. These models send alerts to a large inter-departmental dashboard that shows all imaging orders in real time. We pulled these tools together and into a more harmonized platform (that we also built) with some new technologies and some older and very robust technologies in medical imaging. We’ve also built in some comprehensive and very advanced monitoring of the imaging models (via other data sources in different systems throughout the hospital). While largely centralized in the emergency department, this suite is one of our only major interdepartmental tools. It provides alerts and situational awareness across the emergency department, radiology/medical imaging, and neurosurgery. We’re really excited about the next step, which is to bring this to another one of our hospitals. We’ll make the jump from inter-departmental tools to organization-wide tools. We can notify neurosurgeons at St. Michael’s that there’s a case at St. Joseph’s emergency department.
What do you see as the most promising AI-driven healthcare solutions that Unity Health can offer to the broader healthcare community?
- We have numerous AI-driven healthcare solutions that we believe will add value to the healthcare system (link here), ranging from CHARTwatch – an early warning system that monitors our patients for clinical deterioration and has been shown to decrease unexpected mortality by 26% - to efficiency solutions such as or ED RN Assignment Solution – an assignment solution that allocates nurses to zones in the emergency department that has reduced human effort by over 80%.
- Unfortunately, Unity Health Toronto does not have the ability to deploy these solutions elsewhere given our limited resources and bandwidth, but we are working closely with private sector partners on efforts to scale these solutions globally.
- Though we can’t always directly provide our solutions, we can offer our experience and expertise. We build and maintain networks for knowledge sharing.
How do you ensure that the AI models developed are both accurate and clinically relevant for Unity Health?
- Our rigorous project intake process ensures we work on problems that are highly clinically relevant, as they are proposed by front-line clinicians and requires their full engagement through the development and deployment process. Our solutions are ‘developed by clinicians for clinicians’. While many clinicians may not have a deep understanding of AI, they partner with our AI team where they bring the clinical expertise and we bring the technical expertise. Sine our solutions are developed by our clinicians with our team, adoption is usually quite good.
- Our technical teams ensure model performance meets a certain standard by focusing on problem-specific performance metrics. We also take considerable effort in understanding the ‘status quo’ performance to ensure we have a sound understanding of what level of performance our AI solutions must attain in order to be useful. For example, our CHARTwatch solution is a prognostic solution that predicts patient deterioration (i.e. death or ICU transfer). The standard of care prior to the launch of CHARTwatch was clinician judgement. We collected over 3,000 clinician predictions and were able to quantify physician accuracy (e.g. physician were correct less than 1/3 of the time when they felt a patient would deteriorate). This sets the minimum threshold for AI solution performance in order to provide value.
- We also go one step further: We evaluate and monitor the impact of our solutions. For example, if a tool is meant to reduce unplanned readmissions to the emergency department, we continue to monitor unplanned emergency department visits after deployment. This tells us that tools are, or are not, still working.
What advice would you give to healthcare professionals who are hesitant about adopting AI in their practice?
- It's natural for healthcare professionals to have reservations about adopting AI, and we've encountered similar concerns within our own hospitals. These concerns typically come from a place of care and responsibility—they're rooted in a commitment to patient safety, ethics, and efficacy. And that’s a good thing. It shows that clinicians are thinking critically about the potential impact of new technologies on their patients.
- Our approach has been to engage directly with those concerns. It’s important to first assess current clinical performance and determine whether AI can truly add value. We actively include skeptics in our discussions because their perspectives challenge us to think more rigorously. We set clear, measurable goals, test our AI tools extensively, and only deploy them when there is strong evidence of benefit.
- For healthcare professionals hesitant about AI, my advice would be to see it as part of the continuous improvement process in care. AI isn't about replacing what works but enhancing it. Medicine has always evolved—adopting AI is just the next step in that evolution. By maintaining high standards and an open mind, we can integrate AI in ways that make healthcare safer, more efficient, and ultimately, better for patients.
Note! The content on this blog reflects my personal opinions and does not represent my employer. As the publisher, I am not responsible for the comments section. Each commenter is responsible for their own posts.