Our guest this week was Derek Ritz, Principal Consultant of ecGroup Inc, discussing his digital health work in low and middle income countries, Sub-Saharan Africa, the Middle East and Asia Pacific regions.
This was a fantastic and eye-opening discussion highlighting how Derek’s work overseas has influenced his work in the North American digital health industry, and most recently the hot topic of AI regulation in healthcare.
“There’s a few things in Canada that we’re astoundingly good at. We’re really good at telecommunications. We’re a big country with not very many people. It’s not that surprising. We’re the best hockey players in the world. As it turns out, Canadians are very, very good at digital health. You look at any initiatives, I’ve had an opportunity to work all over the world and we should be proud of the fact that Canadians have an excellent reputation internationally in digital health and that is well defended by absolutely everybody I’ve ever worked with on projects. person.” Derek Ritz
Watch the podcast video now, or download from any of your favourite podcast players.
Find Moving Digital Health on Apple Podcasts and Spotify, and subscribe to the MindSea newsletter to be notified about future episodes.
Read Transcript:
Reuben (00:03.85)
Welcome to the MindSea Podcast series Moving Digital Health. Our guest today is Derek Ritz, Principal Consultant at ecGroup, Inc. Derek advises private and public sector decision makers on healthcare strategy in Canada and internationally. Thank you for joining us today, Derek.
Derek Ritz (00:21.459)
Happy to be on the call.
Reuben (00:23.928)
Could you start by telling us a little bit about your background?
Derek Ritz (00:28.263)
Sure. I refer to myself as a recovering engineer. I’m in a 12-step program and coming along well, thank you very much. I did an engineering degree a little more than 40 years ago and worked in supply chain management and industrial engineering for the first 20 years or so. But I started working in digital health in about 2005, 2006.
Derek Ritz (00:57.011)
The last dozen years or so, almost all of the work, not all of it, but almost all of the work that I’ve done has been in low and middle income countries in Sub-Saharan Africa and in the Middle East and in the Asia Pacific region. And that’s been a very eye opening and interesting part of my career. And to be honest, I don’t think I could go back. It’s got good karma. I don’t think I realized earlier in my career just how much I needed karma in my compensation
Reuben (01:27.138)
Definitely. I think the healthcare industry as a whole is full of amazing people trying to do the right thing and have a positive impact on the world. So how did you go from engineering into digital?
Derek Ritz (01:45.199)
Interestingly, I completely by accident, I mean, I suppose as some things happen, it was just pure serendipity. The company that I had, I mean, when I say that I’m not a quite completely recovered engineer, I had a lot of recovery to do. I’ve got two global patents in cryptography and there’s really no geek like a crypto geek as any other crypto Greek will attest.
Reuben (02:07.694)
Thank you.
Derek Ritz (02:10.935)
But we had some patented technologies that were used for supply chain management and industrial engineering, but the exchange of supply chain documents. So authenticating a recipient and the sender and the recipient and doing non-repudiation of the exchanged documents. And I was making a presentation to the head of IT for a hospital network in Western Ontario. And the IT team that was supposed to be at the meeting were about 10 minutes late. So our supply chain software did and all of the things that made it interesting. And then the IT guy showed up. And so I tried to do a quick recap and I said, well, you know, we authenticate the sender and we authenticate the recipient. There’s non-repudiation of all of the exchange of the documents in between and it’s, you know, completely processable formats. And they said, oh, that’s fantastic. And I said, oh, thank you very much. And they said, you can do that with like patient data, right? I said, well, actually, I don’t know sure we could. I mean, we’ve worked with typically purchase order notices and shipping notices and invoices and the things that make supply chain management go, but I don’t know why we couldn’t do that. I said, oh, that’s a godsend because we’ve just come from a meeting with some of the people at Canada Health InfoWay. They were online with us and we have to try to find some way to do all of the things that you just described, but exchange patient information. And I just kind of filed that away and we went on and had the meeting that we were supposed to have about supply chain management.
Derek Ritz (03:40.807)
And when I got back to the office, I started doing a bit of research about all of this cool stuff that this new agency, Canada Health Infoway, was working on. And to be honest, it really sparked my imagination. And through various events that subsequently happened, I had an opportunity to cash out a little bit of the supply chain company that I was working at and start a new career for myself in digital health. And I haven’t looked back.
Reuben (04:09.69)
Excellent. And so what did the start of that career look like, you know, when you first started working in what was the state of the industry at that time?
Derek Ritz (04:22.103)
So a lot of things were for Canada, just getting rolling. And I say the for Canada part because there were other countries in the world that had an uptake of electronic medical records, for example, much more rapidly than we did. But in 2006, there were a couple of things that for me were very interesting and seemed to, seemed to, for me, make me feel like I was catching a wave.
Derek Ritz (04:51.579)
And the two things that happened in that same year, one, version two of the Canada Health Info Way, Canadian Digital Health Blueprint, the e-Health Blueprint was published. And that particular artifact has the distinction of having been followed by absolutely nobody, but referred to by absolutely everybody. And what it did was got all of the Canadian projects, and there was a number of them underway, whole whack of them that started in the years immediately after that. They all could refer to the blueprint and they all referred to the blueprint and could describe what the various puzzle pieces did in terms of that blueprint and how it described things. And then they could refer very deliberately to which part of the blueprint they weren’t following. And everybody was in that way knew what they were doing that was consistent with the overall health enterprise architecture for the country and what they were doing that was a little bit different. I’ll say in my opinion, not always for good reason. Usually there was a bit of real politics that was coming to play there. So that was one thing that happened and that was in 2006. And there were aspects of that work that I ended up involved in. The other thing that happened, I attended my first e-health conference that year.
And at that conference, there was a breakout session around establishing a health informatics professional credential in Canada. Many Canadian health informatics professionals have the CPMCA designation. And I do too. And I think I was part of the first group of two or three dozen that got the CPMCA. The meeting then, at the 2006 eHealth conference in Victoria was to say, well, how would we go about doing this? And it was four years later, maybe before the credential was made available. But I’m a registered professional engineer and have been for a long time. And one of the things that I found really fascinating is I figured, well, it’s 130 years ago or so now that it must have happened. But I imagine there would at some point in time have been a group of engineers that got together and said, look, in the interests of public safety, how are we going to establish codes of practice? How are we going to organize that we will make sure when engineers do their profession in Canada that they do it well and in so doing, defend the public good? And that was very much the conversation that we were having. And I absolutely felt like I was at the beginning of something big. I still believe, to be honest, I still believe that in the future, maybe not the very long future, maybe not too far in the future, we’ll see that health informatics is a regulated profession in Canada. And I say that with hope. I know that there’s a lot of people that think, oh, that would take all the fun out of it. But I think that regulating health informatics would be a sign of the maturity. It would show just how much we are coming to rely on health informatics as a core part of our healthcare delivery networks, as opposed to a nice to have or a maybe we have. I think it’s now a must have and it’s something that it will matter. It’ll be a key element of public safety that it’s done well. And one of the ways that you ensure that something is done well is that you ensure that there are certain credentials, certain capabilities and educational capabilities or experience capabilities for the people that perform those tasks. The ones that design and implement our digital health, I think should meet a certain level and that we could define what that level is and regulate it a little bit. I know many people say regulating takes all the fun out of it, but that’s what I like about it. Every time we drive over a bridge, we’re all very pleased that all the fun was taken out of the engineering of bridges. And pretty much they all stand up over the river for a hundred years in Canada or more.
Reuben (09:13.418)
Yes, agreed, agreed. So I guess my assumption was that most practicing people in healthcare informatics did have that designation. Is that not the case or like what kind of percentage do you think is lacking there?
Derek Ritz (09:33.127)
First of all, I don’t, I don’t know what the percentage is. I know that, uh, there’s been a pretty steady uptake of the CP Hymns CA designation in Canada. It’s a very well-respected designation. I, I don’t live in fear necessarily that the people who are working in digital health aren’t doing a good job. I think that one of the things that, um, that this kind of an approach would do is routinize that just basically say there’s.
Derek Ritz (10:02.111)
There’s a baseline and it’s, you don’t have to have anybody playing over their heads to have it go well. It’s gonna become a rudimentary and a baseline. I also, I mean, there’s a few things in Canada that we’re astoundingly good at. We’re really good at telecommunications. We’re a big country with not very many people. It’s not that surprising. We’re the best hockey players in the world. As it turns out, Canadians are very, very good at digital health. You look at any initiatives, I’ve had an opportunity to work all over the world and we should be proud of the fact that Canadians have an excellent reputation internationally in digital health and that is well defended by absolutely everybody I’ve ever worked with on projects. And so I don’t think it would be a big stretch for us to say this baseline is one of the things that we’re going to defend. And say we’re going to put in place the things that would make it routine and day in day out kind of a thing.
Reuben (11:05.994)
Right, and take the lead on that certification. You mentioned your experience internationally and working with health care leaders in many different countries. What are some of the challenges that health care leaders are facing?
Derek Ritz (11:27.423)
So in the, first of all, I’ll carve out the United States and I’ll talk a little bit about other places that aren’t there first. I live half the time in Europe and half the time here. And the things that we’re struggling with are pretty much the same things that the Europeans are struggling with. I lived for a while in Auckland. So the things that we’re struggling with are the same things that they’re struggling with in New Zealand and in Australia.
Reuben (11:36.046)
Fair enough.
Derek Ritz (11:56.403)
The OECD countries have got a need to make their digital health solutions bend the cost curve a little bit. There’s a growing group in, demographically we’ve all got a growing group of seniors. In terms of our burden of disease, they’re all pretty similar, and it’s largely driven by chronic disease and how you care for chronic disease.
Reuben (12:15.787)
Yes. Yeah.
Derek Ritz (12:25.627)
One of the truths of caring for people with chronic disease is you win or lose that battle in primary care because chronic disease is something you manage in primary care. And to be honest, when people are coming, you know, if you’re a diabetic whose blood sugar is crashed and you end up in a merge, that was, there were opportunities in primary care to have addressed that. We didn’t, and now this has become something that’s an acute care problem. So that’s, those tend to be the characteristics for the OECD countries.
And all of us are having to figure out how we will leverage our investments in digital health, largely public investments in digital health, to do a better job of bending the cost curve of managing an aging population who by the time you get over 65 in Canada, you will statistically have more than one chronic disease that you’re managing and probably for the rest of your life and we live a long time. So that could be another couple of decades.
That tends to characterize what are the challenges in OECD countries. I’ve done a lot of work in low and middle income countries. Depending on the country, they can have their most prevalent diseases would be infectious diseases. So there are parts of the world where I’ve worked that have a very high burden of HIV or of malaria or TB. And as their economies grow, that tends to shift from being a dominance of infectious diseases to being chronic diseases like we have to deal with. But they are having to deal often with being in low resource environments. And whereas for Canada, we want to bend the cost curve, we really can think, well, we want to get the same amount of health, but we don’t want to have to pay as much for it. So we want to divest the efficiency improvements.
In low and middle income countries, it’s the upside down of that. They want to invest the efficiency improvements because it’s a low resource environment. They don’t have enough doctors, nurses, beds, medicines. There tends to be scarcity. And so the goal of the digital health investments is to say, how can I get the most health production possible out of the not enough resources that I’ve got right now?
And as my economy grows and I have more resources, well, fine, I can deploy those towards health and the social determinants of health. But right now, it’s the upside down. I want to invest my efficiency improvements. I want to get as much healthcare as possible for this amount of these resources that I can bring to bear. So they’ve got a different motivation. And to be honest, where digital health can make a real difference, I almost feel it’s got a moral imperative to it.
Derek Ritz (15:14.207)
How could you not be implementing digital health when it can have such an impact? There was a study from about 2010 that simple SMS-based reminders for people that are HIV positive and are on antiretroviral medicines, simple SMS-based reminders would reduce loss to follow-up by half. Loss to follow-up is when you just get off your meds for whatever the reason might be. So when you can cut that in half, out of simple SMS reminders, how can you not do that? There was an op-ed piece in the Lancet, a digital health, it’s a medical journal, but there’s a digital health Lancet. And they just said, how could you not do that? When the cost effectiveness of this is so astoundingly high, how could you not? And I think that there are some situations where in low and middle income countries, digital health, if you’re going to go to scale with it, and actually say, yeah, I’m going to invest in this to go to scale. Well, how could you not? It can have such a huge impact. I’ve said quite deliberately that I separate the US market from this. And the reason is the US market has drivers that no one else has. It is very much. Even though CMS, their center for Medicaid and Medicare represents about 55% of the spend in the US care delivery network. And so it’s a heavy hitter when CMS wants to pay for something or how they pay for something can move the market a little bit. The real big drivers in the US market are very much economically driven. They’re market driven. And you behave differently when you’re driving profit than when you’re driving health
So in a publicly funded care delivery network, where you’re largely funding care out of the public purse, you wanna drive outcomes, you wanna get performance out of what you’re paying out of the public purse. The things that you will do to make more profit don’t always create more health. And so there are some perverse incentives that influence the US market in ways that don’t always serve it well.
Reuben (17:26.43)
Yeah.
Reuben (17:39.21)
Yeah, some definite contradictions there. And a very, very different systems. You cover kind of like, you know, three big buckets that have very different issues and motivators there.
Derek Ritz (17:57.375)
It’s interesting because so much of the work that I’ve done internationally has been funded by the Americans. The US government’s PEPFAR program, PEPFAR is an acronym for the President’s Emergency Plan for AIDS Relief. It is an astoundingly huge program in 53 countries. I don’t know what the current spend is, but while I was an advisor to them for about eight years, in those years it was about $7 billion US dollars per year. The program is single-handedly responsible for putting a knee on the growth curve of HIV on a global basis. Almost every UHC, universal health coverage initiative that I’ve been part of in low and middle income countries has been funded one way or another largely through the United States and yet the US remains the only country in the OECD that doesn’t have universal health coverage itself.
Reuben (18:56.51)
So you’ve been working with some of these countries long enough to really see the impacts of those programs over a span of time.
Derek Ritz (19:07.259)
I have, and to be honest, it’s, when I made the joke, I didn’t realize how much I needed karma in my compensation plan. To be honest, standing shoulder to shoulder with some of the teammates in these countries as they progress their work, it’s nothing short of inspiring. Pure and simple, it’s nothing short of inspiring. And there’s just people that I admire more than I could ever say. Who I’ve had the great good fortune to work.
Reuben (19:40.466)
That is amazing. Are there any other of those examples that you could share? Like you talked about, you know, the SMS having a big impact. Can you think of any other specific initiatives that you saw really making change?
Derek Ritz (19:59.803)
I had the great good fortune to work on a project in Tanzania some years ago. The funding for it was through the Gates Foundation. The major implementer was PATH, an NGO from Seattle that has offices all over, including in Tanzania. And this project was called the BID project, the Better Immunization Data
Derek Ritz (20:28.823)
one of the things that the notion behind it was that if we have better information about the immunizations that we’re doing of newborns, especially this was quite focused on the set of immunizations that would happen for children under in their first year of life. And it was absolutely amazing to me to work with these people on that program because
I can’t think of anything that has a bigger influence on population health outcomes than immunizing children. I had a chance some years later to work with UNICEF on this, and the data bears it out. When you start to immunize the babies in a country and you reach a threshold, so there’s a herd immunity number that you have to reach, but if you get into the low 90%, 91%, 92%,
of your children are being immunized against measles. They don’t get measles anymore. You reach herd immunity. So there’s these magic numbers that are, they differ from disease to disease, but as immunization programs start to break that 90% threshold, you see under five mortality just plummet in a country. And when the under five mortality plummets, there’s a step function in the life expectancy of a whole population.
Reuben (21:49.303)
Wow.
Derek Ritz (21:57.803)
And so being part of the early work in Tanzania around that effort was, it was just wonderful. And in the ensuing years to have seen the ministry take up the initiative and internalize it and leverage all of that digital health infrastructure that when we were putting it in place, we were doing some experimenting and we took one province in Northern Tanzania.
Arusha province and we did the work there initially and it’s now spread through the whole country and some of the digital health infrastructure patterns have been adopted with other care workflows around antinatal care upstream of the immunization or the integrated management of childhood in IMCI. I don’t remember what the last I stands for but you know these other programs have started to make use of the digital health. And the patterns that were established there, and really see Tanzania progress. And in the course of time, they’d started to do, it had been running long enough that they could do some analyses and see what kind of impact it was having. It was just terrific. And honestly, I believe to this day that if a country has to begin with something, begin with childhood immunizations.
Derek Ritz (23:22.847)
The digital health infrastructure that you would put in place for that is completely durable, reusable infrastructure. By the time you’ve got a client registry with all the babies in it, well, you could put their moms and their brothers and sisters and their grandpas and you could put other people in the client registry. So you have to have a client registry anyways, put babies in it to do the immunization, but it’s a completely durable digital health investment. Aspects of digital health infrastructure that you would put in place for immunization can be completely reused across other immunization programs like COVID and can be used for other care workflows. And on the strength of the health outcomes alone, you can warrant the investments for It’ll pay for it just that one use case will typically pay for all of the costs that you’ll incur to put that in place.
Reuben (24:22.034)
Yeah. Well, that’s a pretty strong argument. When you’re talking about the digital health infrastructure in place, are we just talking about patient records and the storage and transmission of that data? Maybe you could elaborate on that a little bit more.
Derek Ritz (24:43.947)
Sure, well, there’s an initiative that I had the great good fortune to be part of at launch called Open HIE and it’s the Open Health Information Exchange Initiative. ohie.org is their address. And it’s now 10 years old and it took the kind of the cartoon version of the Canada Health Invoi Blueprint. And then in the ensuing years has adopted a different set of transactions, digital health transactions than the ones we were using in Canada. Initially it was CDA based and now it’s using the HL7 FHIR specifications. But that infrastructure is basically an interoperability layer. And then in the cloud above this interoperability layer, you’ll find a client registry, a health worker registry, a facility registry, a terminology service, a shared health record repository, those sorts of things that were part of the Canada Health Infoway Blueprint. Below that interoperability layer, so on the ground, you’ll find electronic medical record solutions be they phone-based or, and by phone I mean like SMS-based or tablet-based or PC-based, you’ll find pharmacy systems, lab systems and so on. And that’s the care delivery network on the ground.
The digital health infrastructure, when I refer to it, I typically mean the data sharing infrastructure, the HIE. And for all of these countries, it’s same as our value proposition. It gives you a way to support continuity of care. And for chronic diseases or for antenatal care or for immunizations, that continuity of care is that’s the side of the bread, that’s the side of the bread the butter is on. So that’s where you really start to enjoy the benefits from a population base.
Reuben (26:15.661)
Okay.
Reuben (26:40.306)
It must be interesting for you seeing, being able to work with different countries who are in different phases of their healthcare infrastructure development and seeing the different speeds as well. I imagine, you know, with some of these developing nations, things actually move a lot faster than a country like Canada that is quite mature, but, you know, changes are maybe slower to happen.
Derek Ritz (27:11.147)
Yes and no. So yes, there tends to be, there tends to not be market forces that will sometimes impede progress. So where the Ministry of Health is behind something and they make a decision to leverage funding to progress, they can move pretty quickly because there tends to not be as much legacy digital health infrastructure in place. There tends not to be as many aspects of the care delivery network that are well-funded and who like it that way. And so you can progress sometimes pretty rapidly. The other challenges are that you’re in a low resource environment. So, you know, sometimes as we were progressing the work in Tanzania, you know, the electricity wasn’t always on and the internet wasn’t available everywhere.
And there are some things that really are different sets of problems than we have to address in Canada. But I will tell you this, one of the things that I read just in the last couple of weeks was an op-ed piece that pointed out that COVID put the lie to how long it takes to get everything done in Canada. I remember being at a conference in 2006.
And we were in the donut line between sessions. And we had been talking about telehealth. And the person in front of me just turned around and said, we are one pandemic away from having absolutely everything we need in about a half an hour.
Reuben (28:56.281)
Ha ha ha.
Derek Ritz (28:57.311)
And it was, of course, complete BS. It actually took 10 days. But after this was in 2006, in 2020, it took us just a matter of weeks to have billing codes for telehealth. And we had lived in hope that we would someday have that.
Reuben (29:03.266)
That’s right.
Derek Ritz (29:22.199)
Since well before 2006, but certainly that was the topic of this conference that I was attending in 2006. And as it turns out, it doesn’t take forever. It takes will. And I still think that Canadians have to come to realize that if we ever expressed our impatience, we would have all of the things necessary to address some of the challenges we have in our care delivery network almost immediately because there are no good reasons for the situations that we find ourselves in. There’s a list of reasons as long as your arm, but there are no good reasons. And it is almost entirely somebody got a horse in the race and that’s the way that we do it this way because to not do it this way is going to dis-intermediate somebody.
Derek Ritz (30:19.767)
And those somebodies have all sorts of passive aggressive ways that they can make sure something never progresses to address the foundational issues we need to address in our care delivery network. And other than that, I have no opinion on the matter at all.
Reuben (30:33.218)
Ha ha! Well, it’s certainly encouraging to know when things really need to get done that they do happen and can happen quickly, but at the same time discouraging that it takes along when there isn’t that huge push and incentive to get things done quickly.
Derek Ritz (31:00.887)
One of the things that I, one of the things that’s been a complete turnaround for me over these last dozen years, it’s interesting, I’ve had somebody say, oh, you must have learned, and even you’ve asked, you must have learned so much about this country or that country as you travel around and work in these places. And the truth is the country I’ve learned the most about is Canada. Because there’s an old saying that the mountain is always clearer from the plane. And there were so many things about our own context that I had.
Derek Ritz (31:30.167)
no appreciation of or no deep appreciation of until I had a chance to see other context or still had a chance to be seeing it at a distance. And one of the things I have come to believe is that we should all just be issued snorkels when you start to work in the Care Delivery Network in Canada, because we swim around in so much money and we just don’t realize it. And we, you know, it’s the almost knee-jerk reaction. We need more money. Oh, no, we don’t.
Derek Ritz (32:00.043)
We just need to start intelligently spending the astounding amount of money that we have.
Reuben (32:05.074)
Yeah. Well, I want to switch gears a little bit here. So I know you gave a talk at e-Health this year titled Do No Harm, Regulating AI-Based Digital Health Solutions. Now, I’d love to hear what some of your key takeaways and messages from that talk were.
Derek Ritz (32:28.431)
You’re making me laugh a little bit because I did refer to some, I did tell some of my friends that if you’re going to make a career ending talk, you may as well make it at the end of your career. And I was particularly in that talk calling out chat GPT. So first I should give some context.
Derek Ritz (32:58.231)
Working with standards communities. So I’ve been a delegate of Canada to the International Standards Organization, Technical Committee for Health Informatics, TC215, since 2007 or so. And the way that work is progressed is that, your head of delegation will, or within your delegations, people will be nominated to be the lead expert on certain work items. And over the years, I’ve been a lead expert on a whole bunch of them. In fact, it was how I started working low-income countries. I was nominated to be a lead expert on health enterprise architectures for low resource environments. And that was the first time I went to Africa and the first time I started to do some work there. But presently, myself and some teammates are nominated Canadian experts on a working group for AI, AI in healthcare. And I was one of the authors of an internal document to inform the technical committee on where some standardization efforts should be focused. I was one of the co-authors of the section on regulation. We have a way to regulate digital health products in Canada that meet the definition of software as a medical device. So there are times when, you know, software products will inform the course of care. And if that’s the case and meet certain criteria that are very well defined by Health Canada, then those software products have to have a class two medical device license. They are software that meets the definition of a medical device. And these are regulated software products. The talk that I was giving at the e-health conference was basically about, you know, how for AI based software that meets this, this definition of a medical device, how are we going to regulate it? And at a top level, there are two things that we need to be able to do that we don’t really have as tools in our toolbox right now in terms of regulation. One of those two things is we have to be able to establish efficacy. It’s kind of a dirty little secret, but right now when you get your Class 2 medical device license, you’re meeting a specification, ISO 13485, that is basically a flavor of ISO 9000. It’s a quality control, a quality management system audit. And quality management systems are around repeatability and consistency. That’s really all they’re around. So, you know, carefully said, your product could suck, but as long as it always sucked, plus or minus, you know, 2 or 3%, you’re able to achieve your certification. The expectation, of course, is that you’re gonna have something of high quality, and that it will be consistently of high quality. But it’s…
Reuben (35:48.454)
Right, but that’s not mandatory to get through. Right.
Derek Ritz (35:53.56)
We don’t test for efficacy. We test for repeatability and consistency in the manufacturer of the product. And we don’t do any attribute tests of the software itself to say whether or not what it’s doing is good. And I’ll put good in air quotes. But for AI-based software as a medical device, we are going to have to measure whether it’s good. And the reason for that is because the regulatory framework right now is very focused on pre-market. What are the things that you have to do in order to have your product able to enter the Canadian market? And then there’s an expectation that if you meet those, your product has entered the market and there you go. But with emergent properties come emergent risks. And one of the things that we all think is cool about AI and the machine learning aspects of it is that, hey, it’ll change, it’ll morph. And we always say that as if, and it’ll get better and better unless it doesn’t. Well, that drives us into the arms of having to do the two things. Have to do post-market monitoring of software products that are AI-based software products. And two, we have to be able to measure the efficacy of it.
Derek Ritz (37:19.535)
There’s been some useful work done by a joint group, the ITU, the International Telecommunications Union is a UN standards organization and WHO, the UN standards organization for health. Those two together have a working group, a focus group on AI and health. And they’ve developed a benchmarking pipeline that can answer the question, well, what is the efficacy of this particular AI based solution? And
There’s now some work underway within Health Canada and in the US FDA, for example, on how we’ll set up post-market surveillance of AI-based SAMD products and how will we test that efficacy? How will we tell that it’s getting better or it’s staying the same, but that it’s not becoming either useless or dangerous? There was a case study of a of an algorithm from a large EMR vendor that went from being great to being a coin toss over the course of an eight or nine year period. And we’re just lucky that that’s where it stopped. It could have gone from being great to being dangerous over that period. It just went to being innocuous. But we have to be able to measure those things. What I thought was interesting about ChatGPT is it has really energized a lot of excitement
Reuben (38:24.942)
Wow.
Derek Ritz (38:45.111)
And there have been all sorts of articles and, oh my God, the dewy eyed hype around chat GPT for healthcare has just been off the charts. And there were a lot of booths at the conference that were talking this up. And I just said, well, actually, number one, that’s not legal in Canada for these use cases that people are talking about with this great degree of excitement.
Reuben (38:47.372)
Yeah, I’m a little hyped.
Derek Ritz (39:15.183)
Just because the use case meets the criteria of a Class II medical device. And on its best day, this thing couldn’t meet the criteria of being certified as one of those. And I used a couple of examples that were trying to be a little bit humorous. And I was just saying, look, one of the last time you saw a want ad for, you know, wanted a very well-read,
Derek Ritz (39:43.995)
contextual awareness at all and no domain experience and a propensity for not only lying but then lying about the citations that it’s using to rely on the first lie. But must be good at limericks. And I thought, you know, you never see a Wontad that says that’s the person you’re looking for. And if you would never hire that person, why on earth would you consider using a bot that with those behaviors? In something important enough to health outcomes that it would meet the criteria of being regulated in Canada. And to make my point, I cut and paste from Entrepreneur Magazine, this little two paragraph snippet that said, chat GPT can be used for these things colon, and it sort of lists them. And I put it on the slide and everywhere it said, a healthcare use case, I put an air traffic control. And I said, you know, well, we can inform doctors to do this. And I said, all right, we can inform pilots to do that. And the air traffic controllers can use it for scheduling, or sorry, in the hospital, they can use it for scheduling this. And I said, the air traffic controllers, you can use it for scheduling takeoffs and landings. And you put the two on there. And when you looked at it and thought, well, if this hallucinatory thing is gonna be used for air traffic control, that’s just nuts. And as soon as you saw it that way, everybody realized that’s a nutty idea. But if it’s a nutty idea for planes, there’s only 800 people on a plane, how could that be a nutty idea for a plane and a good idea for diabetes care? And we have a million diabetics in Ontario. I hope it made the point that there is going to be a huge and important contribution from
Derek Ritz (41:41.439)
Generative AI that is based on transparent algorithms, not black box algorithms, and that doesn’t hallucinate. But that’s not what chat GPT is.
Reuben (41:52.99)
Yeah, yeah. And I think people, as you mentioned before, they just assume that it’s gonna get better really fast. So they say, oh yeah, well it has these flaws now, but it’s getting better so fast that, you know, those are gonna be fixed in the next, you know, so many months or year. But to your point, consistency is required here. Like you can’t assume that the change in the AI is always going to be an improvement.
Derek Ritz (42:31.719)
Well, and the way it will improve is it will learn from its failures. But that’s an interesting proposition right there. I mean, at what point in time do we sign up the Canadian population to be the guinea pigs that helps chat GPT learn from its failures? I struggle with something that when I was in engineering school, there was something I thought was interesting. From the 1750s to the 1950s, we built steam boilers that didn’t explode. And we came to understand that if you made the steel plate this thick and if you put the bolt patterns in that, you know, at that spacing, the steam boiler wouldn’t explode. We learned that because of all the ones that did explode, you know, with great loss of life and all this. But empirically, we learned how to make steam boilers that didn’t explode. And then for 200 years, we made use of them. It drove, you know, it drove the industrial revolution. It’s how we had railways.
You know, all sorts of things came from the fact that we could make steam boilers that didn’t explode, but we didn’t know why they didn’t explode until the 1950s when we had an electron microscope and we could see the lattice framework of steel and we knew, oh, okay, well, that’s why these ones don’t explode. And those ones did explode. So we, we got 200 years of benefit before we knew why it was. And I struggle with the idea that, well, what if, what if these large learning models are like that. And I guess the thing that we would have to be is completely transparent with a population that we were going to use these on and say, listen, some of these are going to explode. And we don’t know if it’s going to be you or not. And if we had sort of informed consent that everybody was okay with that. Then I suppose we could sort of say, we think the upside’s gonna be so much better than the downside that we’ll sign up for that. I just don’t see that being the way that we operate in our societies right now. We have an expectation of do no harm. And that was why I had that as the kind of catch phrase title for my talk. It is part of the whole gestalt in our care delivery network that it will do no harm.
Reuben (44:49.026)
Right.
Derek Ritz (44:54.907)
And the do no harm part means probably we won’t be able to say, eh, it’s going to hallucinate a lot, but we’ll figure it out over time. And in the meantime, when it hallucinates and if there’s, uh, you know, if it has negative impacts for people’s outcomes or starts, you know, we have very negative impacts and people die. I just don’t think as a society, we’d be signed up for that.
Reuben (45:21.702)
Mm-hmm. So your kind of thesis is let, you know, let chatGPT cut its teeth in other industries, where the stakes aren’t as high and then more when it’s proven itself, then that it’s ready for health care. And it’s in a more mature state is maybe a better time to start adopting that.
Derek Ritz (45:50.295)
There are healthcare use cases that aren’t around direct patient care, where I think it could make a contribution now. I think we also are using chat GPT as a synonym for generative AI, and it is one flavor only. There are generative AI projects where it wasn’t the wild, wild web that was used as the training data set. It’s a curated set of hundreds of thousands of medical journals. Well that would be something that would have a harder time hallucinating than chat GPT. Just because it’s trained differently. There are generative AI approaches that I think are going to lead to very useful and fantastically innovative products. And this notion of the transparent algorithms versus the black box algorithms.
For those of us that are in the regulatory AI regulation group at ISO, we’ve pretty much decided on that point that for healthcare purposes, we wouldn’t allow black box algorithms simply because you don’t need them. There’s such fantastic algorithms that are just as novel and innovative and wonderful that are from the family of transparent algorithms. And with those, you can inspect. You can see how did you come up with that answer. Recommendation, you can say, wow, that’s not something anybody expected. How did you come up with that? And you can see how. And it gives you the idea of, or it supports the idea that you could kind of do virtual randomized control trials and see why it is that certain recommendations are being made by the generative AI. I think the future is bright for those sorts of things, but there’s, you know, there’s serious people who have to do serious work before we can leverage this in healthcare. In the ways that we’re anticipating with chat GPT. And the hype is unhelpful to those serious people who have to do their serious work. Just because everybody gets all dewy eyed and frothy mouthed over the hype thing. And it creates an impatience for the careful work that has to be done for a safe generative AI that does no harm.
Reuben (48:12.014)
Mm-hmm. Well, some very good points there, Derek, and I think that’s a great place to wrap it up. So thank you so much for joining me on the podcast. I really appreciated all your insights and knowledge.
Derek Ritz (48:28.991)
Thanks so much, Reuben. My pleasure.
Reuben (48:32.106)
And thanks to everyone for listening to Moving Digital Health today. If you enjoyed this conversation, please go to movingdigitalhealth.com to subscribe to the MindSea newsletter and be notified about future episodes.
