AI and the Specialty Healthcare Opportunity
January 22, 2024
The sky’s the limit for AI in Canadian specialty medicine, though we need to build it from the ground up.
All of a sudden, AI is everywhere. It’s on our desktops, in our apps, on our lips at cocktail parties. It lurks inside our social media feeds, ready to mine and modify our habits. Not surprisingly, AI is also infiltrating the healthcare world, promising to transform medical decision-making and data analysis from the inside out.
For all the buzz around the technology, it bears noting that AI in healthcare actually dates back to the 1970s, when it helped doctors make decisions about which antibiotics to use for individual patients in hospital3 – an important consideration in the context of rising antibiotic resistance. Since that time, the use of AI in healthcare has done nothing but grow, with a rocket-fueled spurt in recent years.
Maxime Cohen, chief AI officer at the Montreal-based ELNA Medical Group, affirms that “there has been a huge explosion in AI activity recently.”14 While the mainstream media has focused its attention on ChatGPT, Cohen maintains that Hippocratic AI, the healthcare AI application underpinning an ambitious US project involving ELNA as the sole Canadian partner, “has outperformed ChatGPT on the majority of medical exams and certifications.” In the coming months, ELNA will also be piloting generative AI that “listens” to doctor-patient encounters and creates medical summaries, which “saves a lot of time for the doctor and can ultimately increase access to more patients.”
At the same time, AI is still in its infancy, much like the Internet in the 1990s. The World Wide Web of that time, while dazzling us with its novelty, was a little clunky and cumbersome – nothing like the lightning-speed, image-saturated cyberverse we take for granted today. The same applies to AI. ChatGPT, the world’s newest toy, has put AI into the hands of the “person on the street.” We’re having a great time playing with it, and forward-looking tech companies are heavily investing in its development, but – like the proverbial boulder rolling toward the edge of a cliff – it has yet to unleash its full kinetic energy.
We don’t quite know how AI will transform specialty healthcare, but by all indications we can expect a wild ride. Should we be excited, cautious – or perhaps a bit of both?
AI IN HEALTHCARE
To understand how AI can revolutionize healthcare, we first need to define it. Under the circumstances, it seems appropriate to give the job to ChatGPT, which defines AI as “computer systems that simulate human intelligence by learning from data, recognizing patterns, and making decisions.”1 AI performs tasks that have previously required human brain function and its algorithms evolve with experience, much as a human might learn from trial and error.Belying the cartoonish perception of AI as a humanoid made of wires, ready to take over the world, “AI does not automate thinking,” says Cassie Kozyrkov, CEO at Data Scientific and Google’s first Chief Decision Scientist.15 Rather, it automates tasks that “don’t require a high level of cognitive engagement, creativity, or critical thinking.” This may change over the years, of course. But for the time being, AI is best understood as a tool that removes the “complex drudgery” from your life, rather than a cognitive competitor intent on stealing your mind. By taking laborious tasks off your plate, AI frees you up do the cognitive heavy lifting.
Canada is emerging as a world player in the AI arena, starting with the launch of the Pan-Canadian Artificial Intelligence Strategy in 2017. The first initiative of its kind in the world, the Strategy received $125 million in funding, with the vision of developing a Canadian AI community.8 The effort has supported the launch of over 900 Canadian AI startups and over 50 multinational companies with AI R&D facilities in Canada, along with an annual crop of over 200 Master’s and PhD students graduating from Canada’s three national AI institutes.8
Within healthcare, AI presents almost limitless opportunities. “Several aspects of the healthcare system involve prediction, including diagnosis, treatment, administration and operations,” note the authors of a Brookings Institute article about AI adoption in healthcare.16 And that’s exactly where AI excels: making predictions on the basis of data.
“Within healthcare, AI presents almost limitless opportunities for all tasks that involve prediction, from diagnosis and treatment to administration and operations.”
– Brookings Institute article on AI in healthcare
By sidestepping the trial-and-error approach that still prevails in much of modern medicine, especially with complex diseases, AI can save valuable time and resources. For example, nearly half of patients don’t respond to the first antidepressant they are prescribed, and a quarter of those can’t tolerate their side effects.17 Based on the ever-growing dataset of antidepressant trials, AI could develop algorithms to predict an individual patient’s suitability for specific antidepressants – more than 35 are available in Canada at the moment – thus enabling a higher probability of prescribers selecting the right drug the first time around. In a similar vein, AI could help predict which drugs would work best based on a patient’s genetic profile and other characteristics.
It’s not a stretch to imagine a version of ChatGPT for diagnosis, another for assessing patients’ laboratory and imaging tests, and a third one for guiding treatment decisions. The capacity of AI to create and update algorithms for drug regimens could prove especially valuable in the world of specialty pharmaceuticals, where practice standards evolve with an ever-moving drug pipeline.
Back to reality: many healthcare facilities continue to use fax machines. There’s clearly a lot of catching up to do. So what needs to happen to make AI work in healthcare?
IS CANADIAN HEALTHCARE READY FOR AI?
Critical to harnessing the full power of AI in healthcare is an infrastructure of data assets and technical expertise, along with the vision to translate big AI ideas into healthcare realities. Let’s start with the data. In terms of raw material, there’s a lot to go around: the healthcare industry generates about 30% of the world’s total volume.5 The compound annual growth rate of healthcare data is expected to reach 36% by 2025, exceeding the expectations for manufacturing by 6% and for financial services by 10%.5
Canada exemplifies this abundance. “Across the country, practically every province and region has pretty nice data sets that many other countries would love to have,” says Muhammad Mamdani, Vice President of Data Science and Advanced Analytics at Unity Health Toronto.10 At the same time, the quality and completeness of the data still varies widely among clinics, hospitals, and jurisdictions. This lack of consistency makes it difficult to use the data for a common purpose, such as an AI-driven analysis and algorithm development.
The fact is, high-quality medical data is difficult to collect. As noted in the above-mentioned Brookings Institute article, “medical professionals often resent the data collection process when it interrupts their workflow, and the collected data is often incomplete.”16 This results in “data collection that is localized rather than integrated to document a patient’s medical history across his health care providers,” placing limits on useful AI applications.
Sean McBride, National Director of Commercial Operations at Bayshore Specialty Rx, agrees. “It all starts with data collection and refinement,” he says. This involves “upfront work to ensure the data has the appropriate rigour and to eliminate biases,” along with “KPIs to evaluate the data as you go along.” The next step is to “feed the data into AI-type systems and conduct pilot projects to see if the information adds value.” In other words, “just adding technology isn’t enough. It needs to be thoughtful and targeted toward delivering value to the patient.”
“Just adding technology isn’t enough. It needs to be thoughtful and targeted toward delivering value to the patient.”
– Sean McBride, Bayshore Specialty Rx
That’s where technical expertise comes in. To Canada’s credit, the country boasts three leading AI institutes: the Vector in Toronto, AMII in Alberta, and MILA in Montreal. Researchers at these institutes develop the processes that enable AI to work with raw data. In parallel with these efforts, companies like Bluedot and Signal 1 are working to translate basic AI research into practice, keeping their eye on the goal of improving patient care. While only a handful of initiatives have reached this “translation and application” stage to date, the future looks bright.
PUTTING IT ALL TOGETHER
A few visionary Canadian pharmaceutical companies are putting the key building blocks of healthcare AI together: data, expertise, and application. Leading the charge, Roche Canada has launched a centre of excellence called AIR [AI with Roche], devoted to improving healthcare through the discovery and application of AI research.19 With a mission to accelerate data insights to bring tangible benefits to patients, AIR covers all the bases in the AI lifecycle: conducting leading-edge research, facilitating partnerships, and mobilizing AI into practice, while prioritizing safety and ethics throughout the journey.Pfizer Canada, for its part, has partnered with Communitech, a Canadian technology supercharger that brings industry together with a national network of startups and scale-ups.20 In October 2023, the partners announced the three winners of a competition for startups working on technological innovations. Selected from an open call for proposals to address key challenges in healthcare, the winners included an Ontario-based company called PharmaGuide, whose patient portal will use AI to streamline drug coverage processes.20 The technology’s eventual goal is to help users predict coverage by insurers, thereby promoting transparency, informed decision-making, and better patient care.
LET’S NOT FORGET THE HUMANS
Healthcare AI holds promise on several fronts. It can improve individual health outcomes and the efficiency of healthcare delivery. It can lift much of the administrative burden from physicians and their staff – an especially important benefit for clinics that manage complex treatments like specialty pharmaceuticals – freeing up more time for patient care and reducing staff burnout. But it can’t do any of this on its own, especially in the complex and high-stakes area of healthcare: real humans must “hold AI’s hand” at every step of the way.
For those concerned about AI taking over healthcare, Dr. Winson Cheung, a Professor of Medicine at the University of Calgary and Principal Director of the Oncology Outcomes (O2) research program, offers a dose of reassurance. “AI is not going to solve every healthcare problem or replace the human side of medicine,” he says. “It’s another tool in the toolbox. AI algorithms facilitate decisions, but don’t make them. We need interpretations by a human. As an example, informed treatment decision-making by human experts that integrate patient preferences and input will continue to stay relevant.”
“AI algorithms facilitate decisions, but don’t make them. We need interpretations by a human.”
– Dr. Winson Cheung, University of Calgary
Zachary Stauber, chief strategy officer at ELNA Medical, concurs. “There’s always a human in the loop,” he says.17 “Nothing happens without being reviewed by a nurse or doctor.” AI systems have been known to make errors, and human oversight ensures they meet high accuracy standards.
There’s also the ethics component. Responsible AI requires balancing innovation with privacy protection, scrutiny of potential biases, and attention to health equity. For example, Canada does not systematically collect data on marginalized populations, leading to under-representation in national datasets,21 which could tilt AI algorithms toward majority populations. Like any disruptive technology, healthcare AI could also find a more receptive audience in younger and more tech-savvy patients. Could this put the grandmother in a nursing home or the elderly neighbours down the street at an healthcare disadvantage?
To mitigate against such biases, health experts across the country are having conversations what ethical AI should look like. In 2020, with funding from the Canadian Institute for Health Information (CIHI), a multidisciplinary team of Canadian scientists launched a pan-Canadian institute to explore the impact of AI on health equity and to train emerging public health and computer science researchers in ethical AI.21 Called Equitable AI for Public Health, the institute holds a conference every summer and invites applications from all over the country.
TAKING THE PULSE
With an aging population, changing patient expectations, and relentless innovation, demand for AI can only keep growing. On the flip side, end-users’ unfamiliarity and ethical concerns with the technology, paired with a shortage of qualified personnel, may put some brakes on uptake. To get a read on Canadian physicians’ awareness and use of AI, MD Analytics recently conducted a survey of 300 physicians, which found that22
• 93% are familiar with AI
• Only 21% currently use AI in their practices
• 60% are not comfortable using AI platforms
• 56% say they need to learn more about the platforms before using them.
As for patients, almost two-thirds of respondents to a recent GlobalData patient survey said they felt comfortable using AI in healthcare settings – but only if the technology was familiar to them.23 Absent this familiarity, their comfort level dropped to a mere 7%. Patients’ top hesitation about using AI in the clinic? The potential to reduce interaction with real humans.
Needless to say, regulating the technology will require juggling several priorities. With this challenge in view, the Canadian government has been working to fold AI legislation into its existing regulations. Enter Bill C-27, intended to modernize Canada’s Personal Information Protection and Electronic Documents Act and introduce new legislation to regulate the use of AI in Canada.24 Part 3 of the Bill, called the Artificial Intelligence and Data Act, sets out new measures to regulate interprovincial and international commerce in AI systems. If passed into law, the Act will place boundaries on AI development and prohibit the use of AI systems that could cause harm to individual Canadians.
By the same token, we need to understand where AI can achieve the greatest good in healthcare. As noted in a 2023 article for HIT Consultant Media, an award-winning digital platform covering healthcare innovation, the true value of AI lies in automating “the tasks that do not need a human touch, allowing clinicians and staff to focus on high-value interactions with patients.”25 Used appropriately, then, AI can actually improve the human element in healthcare delivery – which is what all of us ultimately want.
SO WHERE DO WE GO FROM HERE?
Let’s not get ahead of ourselves: to implement high-value healthcare AI in Canada, we first need to build a solid foundation of data, technology, and training. No less importantly, stakeholders in healthcare AI need to earn patients’ trust: a 2021 survey looking at patient apprehensions about healthcare AI identified concerns related to safety, treatment choice, data bias and security, and potential increases in healthcare costs.23 The authors of the survey report, published in Nature’s Digital Medicine, concluded that “patient acceptance of AI is contingent on mitigating these possible harms.”
To help healthcare organizations implement AI successfully, the Canada Health Infoway has created a Toolkit for Implementers of Artificial Intelligence in Health Care.26 The six-module toolkit offers best practices, tips and recommendations, case studies, and checklists to help organizations plan their activities and stay on track. Along similar lines, Alberta has launched a program called Enabling Better Health through Artificial Intelligence (AI-Better Health) to explore and break down the barriers to deploying AI in the province’s healthcare system.27
If all this seems daunting, we would do well to remember that every journey starts with one step. For healthcare AI, that step is data. So, let’s start by getting the data we need. High-quality data. Data that leaves no-one behind. Data that will enable AI to facilitate earlier and better diagnosis and treatment, thus fulfilling its greatest promise to patients: better health.
“Every journey starts with one step. For healthcare AI, that step is data.”
References
1. Query to ChatGPT. December 3, 2023. https://chat.openai.com
3. Nerkar S. A brief history of A.I. New York Times. December 5, 2023. https://www.nytimes.com/interactive/2023/12/05/business/artificial-intelligence-timeline.html?searchResultPosition=1
5. The healthcare data explosion. RBC Capital Markets. https://www.rbccm.com/en/gib/healthcare/episode/the_healthcare_data_explosion
8. The state of artificial intelligence research in Canada. HillNotes. March 8, 2023. https://hillnotes.ca/2023/03/08/the-state-of-artificial-intelligence-research-in-canada
10. How is AI transforming health care? Unity Health’s VP of Data Science and Advanced Analytics weighs in. Unity Health Toronto. July 26, 2023. https://unityhealth.to/2023/07/ai-transforming-health-care/
14. Zeidenberg J. Montreal's ELNA emerges as a healthcare AI powerhouse. Canadian Healthcare Technology. Oct. 31, 2023. https://www.canhealth.com/2023/10/31/montreals-elna-medical-emerges-as-a-healthcare-ai-powerhouse/
15. Kozyrkov C. Thunking vs thinking: whose job does AI automate? Medium. July 28, 2023. https://kozyrkov.medium.com/thunking-vs-thinking-whose-job-does-ai-automate-959e3585877b
16. Goldfarb A, Teodoridis F. Why is AI adoption in healthcare lagging? Brookings Institute. March 9, 2022. https://www.brookings.edu/articles/why-is-ai-adoption-in-health-care-lagging/
17. Bains C. Publicly funded genetic test for suitable antidepressants would save health-care costs: study. The Globe and Mail. Nov. 14, 2023.
18. Spatharou A et al. Transforming healthcare with AI: the impact on the workforce and organizations. McKinsey & Co. March 10, 2020. https://www.mckinsey.com/industries/healthcare/our-insights/transforming-healthcare-with-ai#/
19. AI with Roche. https://aiwithroche.com/
20. Pfizer Canada Healthcare Hub. https://www.pfizer.ca/en/science/pfizer-canada-healthcare-hub
21. Building equitable AI for public health. University of Manitoba News. January 17, 2020. https://news.umanitoba.ca/building-equitable-ai-for-public-health/
22. Turner D. AI: How receptive have physicians been so far? MD Analytics. Oct. 16, 2023. https://www.mdanalytics.com/en/survey-data/physicians-reception-to-ai/
23. Beer H. Need for patient education on AI in healthcare to build trust revealed in new survey. Hospital Healthcare Europe. Oct. 9, 2023. https://hospitalhealthcare.com/news/need-for-patient-education-on-ai-in-healthcare-to-build-trust-revealed-in-new-survey/
24. The landscape of AI regulation in Canada. Cassels. Sept. 12, 2023. https://cassels.com/insights/the-landscape-of-ai-regulation-in-canada
25. Jessel N. How AI Can Improve the Human Connections of Healthcare. HIT Consultant. May 18, 2023. https://hitconsultant.net/2023/05/18/ai-improve-human-connections-of-healthcare
26. Preparing the health care community for AI implementations. Canada Health Infoway. https://www.infoway-inforoute.ca/en/digital-health-initiatives/innovative-technologies/artificial-intelligence/toolkit-for-ai-implementers
27. Alberta Innovates is looking to enable better health through artificial intelligence. Alberta Innovates. Oct. 4, 2023. https://albertainnovates.ca/news/alberta-innovates-is-looking-to-enable-better-health-through-artificial-intelligence/