How can AI support human-centered healthcare?
Thank you for listening to the Ask AI Podcast. To get our latest episodes, AI news, discount codes for events, and links to free resources, sign up to our monthly newsletter by visiting askai. org and clicking the subscribe link. That's askai. org.
On today's episode of the Ask AI Podcast, we're taking it to healthcare. We sit down with Dr. Alexandra Greenhill, Founder, CEO, and Chief Medical Officer of Careteam Technologies. This is an incredible company and I can't wait to get into that but first it's an incredible person both a winner of a Medal of Service from Her Majesty, the late Queen Elizabeth II and Top 40 Under 40.
Dr.Alexandra Greenhill is re markable. What she talks about with Careteam is on the way to better and how they are the ways for the healthcare maze tune in. This episode is sponsored by Cinchy.
Cinchy is the enterprise Data Collaboration Platform that enables people, systems, and AI to co- produce intelligent, reusable Data Products in real-time. By eliminating data integration from IT delivery, Cinchy makes it virtually impossible for people or AI systems to violate Data Protection controls as you leverage AI technology for your business. If you are an enterprise IT leader looking to de risk and accelerate your AI journey, this is what you've been looking for.
Visit cinchy.com to learn more. You know when preparing to do this podcast and interview you you know we had met many years ago. It's great. Great to chat again.
I noticed we actually have a few things in common. One being winners of the Women's Network Top 100 Most Powerful Women in Canada. So congratulations.
That's amazing. And you were awarded by Queen Elizabeth II, a diamond Jubilee Medal of Service. And I was awarded the Queen's Young Leaders Award from Queen Elizabeth II. Congratulations on that too.
Thank you, thank you. That is an inspiring woman. Whatever one may want to say about monarchy, this is a woman leader that saw and did a lot. Yeah, absolutely. And, and she certainly towards the end of her life was really making efforts towards some of you know certainly the damages of colonialism that were inherited as she took over the empire. Right.
And so, I was part of that legacy to try and change and use the connections of those countries as an agent for good. So yeah she I agree she was an amazing woman to meet and really really inspiring. So congratulations to you and very amazing to meet and chat with someone who has such similar connections and such amazing experiences.
So very excited to chat with you today about artificial intelligence technology as it relates to health care. Why don't you tell us a little bit about Careteam for which you are the CEO and Chief Medical Officer? The big idea in healthcare is that a lot of the technology we use today has the letter R in it. So patient health record is a record of what happened. Electronic medical record is another record of what happened.
What one really needs is a guide to what should happen next. And so anytime you're diagnosed with a complex illness you want a map of what next? How do I get there? How do I ensure that everything happens the way it should happen to be able to successfully navigate the steps from symptom to diagnosis diagnosis to treatment and hopefully then remission ongoing life with a condition that is managed correctly. And so surprisingly in healthcare most things are focused on recording what just happened or has happened but there was nothing in place to help guide people especially if they had more than one health issue. And that's what Careteam does. It provides that expertise and guiding the person on the path to better health. Wow.
That is amazing. So to oversimplify and add a a quirky little way of describing it is keratin like a mystical magic ball where you can you know like a fortune teller see the future of what's coming next? Like you know without giving away all your trade secrets like what is your secret sauce? How does this mystical fortune teller what's coming next technology work? Oh very similar to a lot of the AI that is. It's seen as a black box but in fact it's a collaboration between humans and technology maximizing the benefits of each. The way we've designed this is to say there are humans, usually medical experts sometimes social workers or community organizations that know what the best path is.
And they just need to find a way of being able to share that at the right time and the right people so that they can take their expertise of the local environment and guide the patients across of it. So the you know if healthcare is a maze we are like the ways the app that tells you what's the shortest path? What's the best path to take for you right now on the way to better? But it implies humans because the local conditions are different. The way cancer is treated in one place versus another depends on scopes of practice. So what can each clinician do or not do? How many clinicians do you have? Where are they located? And so the last mile of healthcare is always different based on the local factors and our technology enables that collection of experts in that environment to better guide their patients. Wow. You said so many incredible nuggets there.
I love the ways for the healthcare maze. I'm literally envisioning like media and ads for you that's like on the way to better. We're the ways for the healthcare maze. I just I love it. It's so good. you know and as someone who has lived in Sault Ste.
Marie where I grew up my entire life then lived in London, Ontario going to school at Western and then living in Toronto briefly for a while as well. I have seen that you know that last mile piece that you talk about in the local factors and the differences and they are significant and it always makes me think of the Canada Health Act and universality and it seems to me like Careteam could be helping uphold some of those pillars of the Canada Health Act to ensure that Canadians really are getting you know as equal I say in air quotes with my fingers as equal as they can care by navigating them in the right way for their local environment. Would you say that you work to upload the Canada Health Act in that way? Well I mean uphold the right to better access the right to the right care regardless of where you're located or how savvy you are. And I've worked my entire career on that mission and one of our board members is Bill Tholl he was the CEO of the Canadian Medical Association, the Canadian Hospital Association but before that he was one of the people who wrote the Canada Health Act. And so you're hitting very close to home comment because we have a system that is ideal in theory but just because you don't live close to a city you don't have access to services. And so one of our partners and early adopters was for example the B.
C. Children's Hospital, the Type 1 Diabetes department. And what we're trying to do is to say kids were diagnosed with Type 1 Diabetes it's devastating usually to the family and very anxiety provoking because parents are unsure what to do. They're like if I feed this child in a certain way or if I don't get their medication the right way they can overdose on insulin and end up hypoglycemic and in a coma or they can underdose and end up hyperglycemic so high sugar and in a coma and or die. And therefore how do I handle this? And through our technology you can take this team of experts in downtown Vancouver and essentially make them available to specialists, family physicians, dieticians, nutritionists, patients, and family members in school systems across the entire province. And wouldn't you want that for your child when you're not sure what's happening and when it constantly has to adjust? Because you figure it out and then the child grows or becomes more active whether they switch school or sports.
And so that constant change is something that our system is equipped to handle. Wow. Wow. I mean this this is great.
This all sounds amazing. Like I love hearing you on your journey and seeing how far you've come. This is wonderful. But but let's take it back for a second. You know there are over 200, 000 new digital healthcare apps. So why did you start Careteam? Well how could you see this vision all the impact you're making now the ways for the maze of healthcare? Yeah.
Why did you start it? And how could you see this back then in such a crowded space? Wow. One of the first sort of by previous company with the first sort of telehealth solutions. And I saw then what everybody experienced now which is telehealth is great. It removes some visits because you can do them by video. And it doesn't really matter which platform they're more or less the same. But it wasn't a wow in terms of a virtual care experience. What was really missing
is the before and after and figuring out how do you connect all of these together? And so I did a year of travel. I went around the world and visited hospitals, emergency room departments, talked to patients in places where there was for profit only or mixed public and for profit or public systems all over Canada, US, Europe, China, talk to a number of places as well in Latin America and I realized that they're all struggling because the fundamentals are true. Someone's not well, someone's treating them, and someone pays.
But what has changed in the last couple of decades is we save patients that used to die. And they now survive for longer with more complex treatments. And they have more than one health issue. So a third of the people who are not well have six to seven different diagnoses. And they're all managed by a specialist or a team of specialists but they're not actually a team. They are sort of exist in their own reality.
And the patient then goes and downloads a number of apps from the app store. And none of these things talk to each other. And so in French you would call that a cacophony like a Tower of Babel or you know,collection of violins. There's many analogies.
It's chaos. And then I realized that wait a second. I started working. My work was as an emergency room doctor. I now I'm family physician and the job of an emergency room doctor or family physician is not to know everything but is to actually orchestrate everyone and say okay cardiologist, it's your turn to talk to the patient. Great.
Like let's send the patient x-rays. And when they come back the neurologist gets to talk about the brain. And so it's that sort of a okay coordination function that was missing and when we started building that out we realized that it's okay to coordinate but without a clear path forwards patients get anxious. And so we essentially put together this care plan for each person that's personalized to them and the ability for them to then pull different apps and devices and tools and people as they are needed along the journey in order to maximize the benefit. And so this is what Eric Schmidt calls a platform.
It's a tool that allows other tools to then shine because through us they can discover all these new apps and devices that can make their life easier and do a single sign on and send data to them but also pull data back so that everything is in one place. And then you can start benefiting from predictive analytics around the 360 of a patient of really what's happening to them as opposed to your tiny little view from your unique app perspective frontal front door to a hospital which collects just tiny little slivers of information I love that that holistic summary of what you could see. You know back when telehealth was emerging and I love the integration of the predictive analytics single sign on. Obviously now I work in cyber security and that's like very important to be able to to have those safeguards in place with data and different ways of accessing it. I love that.
You actually spent time internationally studying health care systems. That was actually 1 of my favorite classes in undergrad learning about Finland, Estonia, Taiwan, like all these other unique and different healthcare systems. So that's that's really really interesting. The problem is universal and the other element of international is as people travel, they leave loved ones behind. And so so many people today have family members who are not well elsewhere.
And so being able to access these records is complicated from not just a technical but also privacy reason. And so the way we built this is to allow the patient to share directly with whoever they need to, but on a role-based access so they can share everything with something. And so we are seeing now families being able to support their aging parents from afar.
And during COVID in particular you could be in the same city, you still couldn't visit but you could figure out oh your next appointment is with a cardiologist and and be there and provide information and do the actions and just. be more involved than you would have been otherwise. And that is such a peace of mind for so many people.
Yeah it's such a blessing. I mean I'm 4,000 kilometers away from my family, too. So I really get that.
That's such a such a wonderful peace of mind. and I also really think this is fascinating this what you said earlier in that response around patients that used to die are now living, but they're living with very complicated complicated diseases. Tons of care people on their teams all sorts of different carers and nobody's talking to each other and they can't get access to all this information. So it is, it is interesting that that shift is it's so true right? From longevity to quality of life. You got that. It's the quality of life that matters.
And a lot of people take what's being built in the valley and they say Oh we want maximum eyeballs times in the app. And in my experience patients don't want to be a patient 24/7. They want to be a spouse, a parent.
They want to go to work and have fun. And they want the technology to be there when they need it to tell them what the next action is. And so it's all about maximizing life with the support of a technical solution and same with physicians.
It's not about how long they spend. In fact our metric is how little they have to spend to make this work. Oh I love that. Like the reverse metrics in your app the the shortest amount of time to make those decisions to get people what they need rather than sucking them in and keeping those eyeballs on for as long as possible.
That's incredible. I also really like that you know not a patient 24/7 it makes me think of accessibility technologies and how Randy Marsden who used to work at Apple he was the one who taught me this another great Canadian doing amazing work in accessibility with keyboards. And now he's moved on to Nike. But I remember he was telling me you know think about it.
Like everyone is going to need some form of accessibility support in their lifetime whether it is. long-term in your lifetime whether it is temporary like breaking an arm breaking a leg whether it is aging induced the different comorbidities that come with aging and longevity that we just touched on that longevity and, quality of life balance. And so that that piece of not being a patient 24/7 and the holistic aspects of health also makes me think of the other side of the. You're not being a patient 24/7 but at some point in your life you are going to be a patient. No, it's not 24 seven, but it's going to happen to you and you're going to need and want these tools because inevitably it's you it's your kids, it's your parents like something in your life. So it sounds like this is super helpful.
Well it is and you know I have one of the best co founders. It's the second time I'm doing a business with him but Rob Atwell was the chair of the Neal's Choir Society that works on providing technology for people with deep disabilities and they have been pushing and it's still not done but there's been progress on bridging the accessibility gap. Because all of the technologies being introduced smartphones, tablets, computers, are never designed with people with accessibility needs in mind. And so it takes 12 to 18 months for then others to layer on accessibility features.
And their point is that, why? Why can't we just design this from the get go? To be useful to people of all abilities if we're truly inclusive in how we view the world. That's amazing. Well look this has been fascinating setting the healthcare stage, but we are an AI podcast so we have to get into that in the last five minutes here. You know talk to me about some of the you know predictive analytics that you have and how you can create these trends when everything is different you know different patients, different diseases, even different settings within the app. Like, talk to me about how that predictive analytics work and take us a little bit under the hood for the specific AI concepts that you are using.
Yeah. I'll swear you all to secrecy. No, I'm kidding. And so, right now, a lot of people are doing predictive analytics based on the disease or pain. moment in time. And this is almost like trying to guess from a snippet of a video of me moving exactly where I'm going to go or where I've come from.
And so we've designed models that allow us to look at multiple dimensions and figure out which ones of those dimensions matter more. And then provide that information to a human. So human in the loop AI to look at supporting the patients better. And what this really looks like is to name another great Canadian organization Body Brave in Hamilton, McMaster is dealing with people with eating disorders. During the pandemic their wait list tripled.
They went from 500 patients a month they could see with their team to over 3,000 on the wait list. And so they said, we can't train and hire that many specialized resources. What if we adopt care team and we start handling the entire population of patients but they could still only like mainly see 500 is. And what we did is we use the predictive analytics to figure out who are the 500 who needed the most help for that given time and allow the rest of them to self serve. And so an analogy that would be like in stores you could either wait for the cashier or if you feel well enough you can walk up to the self serve checkout and head out.
But there's somebody there to help you out in case you have trouble. And so by doing that they were able to literally six times increase their capacity to support patients. And their patient satisfaction went up because the ones who were getting seen are the ones who needed the appointment. Not the ones who it happened to be scheduled but they were now better and they didn't need it anymore, but they still wanted to say go and and so the sort of a readjustment that happened was just magic and taking that one approach allows you to know who actually needs help and just go talk to them and allows you to leave everybody else alone because they want to do self. Check in like in and out quickly. I've got it.
I'm doing fine. I'm back to doing my daily life and work and fun, and I don't need help. I'm okay. Yeah. And, you know, you mentioned some of the different dimensions beyond the disease. If you can't tell us what they are just be like, can't tell.
Yeah. But what are some of them, in addition to the disease that you take into consideration? Some of these additional dimensions if you can tell us. Well many people have designed AI around one of these parameters be it social determinants of health or agent stage or engagement on the app. It's really difficult to put together AI that looks at all of these dimensions as well as local factors. So it's not an AI for eating disorders but it's a AI for eating disorders in Ontario in that region of the country. And so, building that type of tool allows you to do what I'm starting to call micro AI, because it allows you to, give a tool to that clinical team that is relevant to them.
And so it's much better adapted to their needs and their protocols. Okay, there's two last things I want to touch on on the AI piece. One is going to be human in the loop, and one is going to be less micro AI. You know in this age of LLMs large language learning models like all these big bigger is better how much data do you have? You know look I worked in AI. I was building datasets. And I remember how much we could improve our accuracy yeah with volume but also with specific volume by having the right data and narrowly focusing on you know, we were working in, in legal the legal space and in contracts.
And it wasn't just all contracts. It was master service agreements. It wasn't even just all master service agreements. We went even narrower and said all the warranty clauses all the limitation of liability clauses like some of those more narrow clauses where contract managers negotiate the most. So we went and we built you know thousands and thousands of human in the loop data sets that were specific to these areas and it improves the accuracy tremendously.
Like, yeah, we were a small company. We didn't have billions of parameters and and I would love for you to speak to a little bit about the micro AI. The fact that when it's the right specific data. Even in the thousands, tens of thousands, hundreds of thousands, millions, like you can still get tremendous AI outputs doesn't have to be in the billions as long as it's narrow and specific and maybe speak to how you're sectioning that off focusing aligning the data you know to the extent that you can, and certainly touch on, please also the human in the loop aspect too, because that's so critical and people underestimate it or they're offshoring it which is also you know in a way predatory right? Where you have all these. Groups of people in different countries who don't have equity in the company. No, no, they don't have equity in the Silicon Valley based company, but they are the ones doing all the data labeling, particularly in the Philippines and countries like that.
You're seeing that a lot. So if you could touch on the micro AI piece and how bigger isn't necessarily always better, as long as it's specific and reaches a certain threshold of big enough, and then the human in the loop piece, I'd love to hear you touch on those. Well, I can recognize a fellow, you know, AI nerd here, so you are way ahead and it's gotten so much easier nowadays and faster, which is an important parameter. But, you know, to make it easy for people who are not as sort of in the tech space to understand is, it's funny to observe a medical student to me because it kind of reminds you of where you started, and I vividly remember that you go with, and you stand like three hours interviewing a patient and you come out and you think that you've covered everything and then you still don't know what the diagnosis is and the more experience attending, walks in, asks three questions and they got it. And so It's fascinating how that applies to A. I.
and so you know to give people a sense of that is you know the 3 questions that matter. Someone's coughing that really matter. Like, 1 of them is, you know, do you have fever? Yes or no? Is there blood in the sputum? Yes or no? And have you recently traveled? Yes or no? And then those 3 questions rapidly eliminate a number of Diagnosis, but, you know, the traveling 1 basically says, do you have a North American disease? Or did you contract something crazy? Like COVID 19, overseas? And so, you know, at the beginning of the pandemic, that was those would have been good questions to ask, but, it really depends on the experience of the person, asking the questions and also in, in, you know, Your interpretation of the answers because sometimes patients will be doctor pleasing and they'll say things like I'm not tired at all.
And you're like, oh, okay. Well, if you're not in pain and not tired on a scale of 1 to 10, what is your pain? Like? And they're like, oh, it's a 9 and and so you can't trust humans to give you the right answers. It depends how you ask the question to and so there's so many of these things that are important in the design of how you collect the data, not just label it but how you What questions do you ask? How do you ask them? In case of the eating disorder program if you ask people with eating disorder you know, what's your BMI, and they don't know what BMI stands for, or how to calculate it, or they're triggered by the question, because, you know, BMI is body mass index, and it reminds them of either how not to eat. Well they are or you know it's something that they're afraid of revealing. You suddenly cut off a number of people who simply not reply to your questionnaire and suddenly your data set is missing an important component And so, I, I see a future where you need more expertise to design these.
A lot of my medical students are being hired by Valley companies to do algorithms and protocols in their summer months, and this is literally somebody who spent their first year in a book learning the core concepts of medicine. And I worry about a lot of these algorithms, not understanding real life and how to ask the question, what questions matter. And so we are in the chaotic phase, in, in when I was researching patient facing AI apps for the book, Something like 80 plus of them don't even have privacy policies or terms of reference or any evidence of, any, I don't know, research. So it will say something like Harvard based research, mental health app.
And when you look at that, and it's something from 98 and you're like, how current is that? Or it's a mislabeling of the research proved that see Cognitive Behavioral Therapy could be useful for patients but not technology delivered or not that particular app. And so there's so much misinformation and hype that we need to proceed with caution. We need to ensure quality and we need to until we are absolutely sure that the machine gets it right. We need to ensure that there's a human in the loop to both direct the the algorithms to become better but to protect the patients in the meantime. Exactly.
What a, what a wonderful explanation talking about very complicated components with the micro AI concept and the humans in the loop. Thank you so much, Dr. Alexander Greenhill. Is there any one last thing that you would like to share with our listeners that we didn't touch on that you're just like, Man, people need to know this. I gotta say it. Anything, final words of wisdom.
both physicians and non physicians feel like, oh, AI has, the boat has sailed, it's too late for me to figure it out. On the contrary, the boat has gotten easier to get on. before you had to learn a lot about technology to be able to use it it's gotten to be much simpler. learn what questions to ask, including when people tell you, oh, it has AI And you're like, really? What kind exactly? you know, make sure that it was designed by people who knew what they were doing. Start using it.
Start engaging and learning. It, the, the gap of what you need to learn in order to be proficient and be able to use things has gotten to be slow, smaller. And so let's all become more comfortable and using these new things because they can make a positive difference if we let them if we choose to. What a wonderful note to end on. Well, thank you so very much. And that wraps up this episode of the Ask AI podcast, and be sure to stay tuned for more exciting episodes this season, ranging from everything from education, healthcare, and beyond.
Thanks. Bye. Thank you.
What an episode! Oh my goodness! Dr. Alexandra Greenhill, incredible, right? Like how we have so many records of what happened in healthcare but not what should happen next. This this future predictive here's what should happen next given all these factors about you and about whatever disease or experience you're having in the healthcare system. I feel like we're on our way to better already. A quick reminder to our listeners that the Ask AI newsletter is the best way to stay updated on key AI news, event promo codes and the latest episodes of our podcast Thanks again, Dr. Alexandra Greenhill from Careteam for joining us.
Thanks for listening to this Ask AI podcast. This episode was edited by James Fajardo. Original music was provided by Mike Letourneau. The series producer was Chris McLellan.
To view the episode transcript, stream the video version, and get helpful links, check out the episode post in the Ask AI blog.