AI’s potential in medicine has rarely felt more important than now. But is Australia’s regulation helping us fully realise that potential? How can Australia become a medical AI innovator in a COVID-reset world?

To mark the 2020 Australia-Israel Innovation Summit, medical AI industry leader Emma Hossack explored these questions and more with Gilbert + Tobin lawyer Anna Belgiorno-Nettis. Emma is the CEO of the Medical Software Industry Association, an Australian Digital Health Agency Director, and past CEO of healthcare software provider Extensia.

To find out more about regulating medical software in Australia, read the MSIA’s submissions to the Therapeutic Goods Administration in May 2020 on Software-based Product Regulation and in September 2020 on Fit for Purpose Regulation.



Anna Belgiorno-Nettis (00:09):

Around the world AI's potential in medicine has rarely felt more important than now. But is Australia's regulation, helping us fully realize that potential? How can Australia become a medical AI innovator in a COVID reset world? Hello, I'm Anna Belgiorno-Nettis, a lawyer at Gilbert and Tobin. Today, we'll be exploring those questions and more, in a podcast marking the 2020 Australia Israel Innovation Summit. I'm joined by one of Australia's leaders in this space, Emma Hossack. Emma is the CEO of the Medical Software Industry of Australia. She's also an Australian Digital Health Agency director, and she's the past CEO of healthcare software provider, Extensia. Emma, it is so wonderful to get to talk to you today.

Emma Hossack (01:05):

My pleasure to talk to you.

Anna Belgiorno-Nettis (01:07):

So I wanted to start off with your arrival into this industry, you know full well that medical software is booming at the moment and that was before COVID, let alone now. You had the foresight to join that rollercoaster pretty early on. When you made that jump from your private practice job as a corporate lawyer, what made you choose the tech industry and specifically health related tech?

Emma Hossack (01:41):

Well, to be perfectly honest, it was a little bit serendipitous, but not entirely. So at that time, I was practicing law, but I was also considering doing some post graduate work and then a commercial research collaborative out of the University of Queensland, became available, and that company was Extensia. And that was essentially a version, in fact, some people would say, and did say and still do say, it was what the My Health Record should be because it's a web based platform, which allows people genuinely to share real time information in a way that suits them. Be that in itemized format, through whatever technology, in any way that suited them. And I just thought that was really exciting and I thought it was probably way behind the times when you look to what was happening in the finance industry. And so I then got very excited by the whole concept of ownership and management of information and what more valuable information is there, then health information.

Emma Hossack (02:49):

And yet most people in Australia would not have heard of the medical software industry, which holds managers, collect transfers, manipulate all the health information in Australia through it's member systems. And yet everyone will have heard of the Australia Bankers Association because that manages money, but there's no question which is more valuable. So it's really interesting and I thought it was a terrible indictment on industry that it was so little known and yet achieved so much in Australia and had such potential to increase, not just, I suppose, the expanding of Australia owned healthcare, because although we all have issues with health system, we still sit up at number one for healthcare outcomes, according to the Commonwealth Fund. So obviously way above that USA and above the UK, so on and so forth and we come at number two in efficiency behind the UK. But we obviously have got a very long way to go still and I wanted to have a part of that, I thought it was a real mission.

Anna Belgiorno-Nettis (03:54):

Mmm. And in a sense, you've even got a greater part now in that mission, from going to the CEO of Extensia, to now the CEO of the Medical Software Industry of Australia. So you took on that role early last year and what were your priorities stepping up into that position for the industry?

Emma Hossack (04:18):

Oh, number one was to shore up all of the support and make transparent to all of the stakeholders, and there are many, there's probably over 75 that we deal with. Make transparent to then the actual work that we do and the fact that unlike, I guess, some of the preconceptions people have about tech industry and software in particular, we're actually not a bunch of vendors wanting to chuck something over the fence to our customers and just hope it works. We're generally people who have come from backgrounds as clinicians, be that nursing, doctors, specialists, people who have actually worked in law, people who are engineers, a whole range of different professions and FinTech, a lot of people come out of banking as well because they actually see the potential to make things a lot better and they are working really hard to do that.

Emma Hossack (05:16):

Okay their businesses need to be making them viable clearly, but there's an awful lot of pro bono work being done by our industry as well. So my mission was to make sure that our reputation was actually as good as it should be, with all those stakeholders and particularly the ones that have massive impact on industry. Like obviously the government, the departments, and those assets like the Pharmacy Guild of Australia, the AMA, RSCGP, all those bodies which are effectively sending our clients, make sure that they are all acutely aware of what we did and where we could, we work with those guys because they've been around a lot longer than we have. And have a great [inaudible 00:06:07] too, with things like the seven community pharmacy agreement, of course, which goes for five years at a cost of multiple billions of dollars. So it's good to work with people who are involved in those projects.

Anna Belgiorno-Nettis (06:19):

And you can really see, just even in the work that I've seen, how involved and engaged you and the Medical Software industry of Australia have been in a lot of the government initiatives that have come forward. So to turn a little bit more onto that topic of today's podcast, medical AI regulation, if we look at regulating medical AI, a lot of frameworks globally have gone about putting it into the regulatory framework for medical devices. But Emma, from your perspective, at a principle level, is this a bit of a square peg round hole problem, is trying to regulate medical AI through the lens of medical devices, the right way to go about it?

Emma Hossack (07:19):

No. And you said it really well. I think the square peg round hole thing is absolutely appropriate. We are software and I had this discussion with one of your founding partners, Peter Landon, who found the whole thing quite confounding. I hope he doesn't mind me saying that. But software is not a thing, it's not actually a tangible thing. And when the TGA refers to devices, it's got a long description which doesn't in any way refer to software, it refers to things, to tangible items and I think it might use the word articles. Now, software, just to kick off, isn't even in part of that definition. And yet I guess the TGA has found itself in an invidious position because on the one side, it will have, quite rightly so, great concern about making sure that consumers are safe from wicked software, software that could actually do harm.

Emma Hossack (08:30):

Now, particularly in mental health and that comes through over the last 12 months, there's been a lot of concern about mental health apps, because there's an explosion of those. And obviously you're looking at a vulnerable market and you're looking at genuine harm that will could come about from someone who has perhaps actually listen to a malicious app and follow instructions to do terrible things. So understand that, but then you go, okay, well, the Therapeutic Goods Administration, what can they do about that anyway, you've got to look at them in a global context because even if they do their absolute hardest, work really hard to get that right, I'm still, as someone looking to buy those apps, they will get them from overseas. And if the TGA does a whole bunch of regulation on Australian apps, those apps are going to incur a lot more expense to go on the market. And that means that there's going to be obviously market forces at play and people are not going to buy the Australian ones. They're going to go to the ones that I can get on the net for free, which has probably very limited oversight.

Emma Hossack (09:33):

So they've got themselves in that difficult position and they're coming at it from a old fashioned hardware perspective too. So it's really difficult for them to understand how software works. And to be fair to the TPA, they do actually admit they don't want to look under the hood and they're not software experts. So they're not software experts and yet they're being asked to regulate this really complicated software environment, where every day there's changes to software, there's changes to, not just policy, but how it's coded and changes to the latest way that archetypes can be created, the latest way how algorithms are deployed and of course, every day there's changes to health science. So it's a hell of an area to try and regulate and to try and do it with existing methods that have been around for regulating devices that have hardware that could actually cause me an electric shock if I put it into my body, could cause gave damage. To apply that to software, which actually doesn't do that, is, I think, very much square peg round hole. Absolutely agree with you.

Anna Belgiorno-Nettis (10:49):

And that point that you mentioned just there, on the fact that it's changing, that can almost be seen as the fundamental element of AI, that it's got this machine learning element. And that often is what distinguishes it from any standard IT system that generates or stores data, the fact that it can learn, so it's even more similar to human intelligence, but at the same time, that can sometimes be the trickiest part to regulate. How do you regulate something that changes, especially when it's changing in a way that you don't quite understand, or you can't quite interrogate? There are even some developers that are turning that machine learning function off, just so they can fit regulation, but then goodness knows what benefits we're missing if that happens. So what are your thoughts around regulating to address that machine learning element of AI?

Emma Hossack (11:51):

That's a key point and what you're saying is absolutely right. So when our initial consultation began with the TGA about it, you could hear a pin drop in the room full of about 150 companies, large and small. The reason for that is that all of them are trying really hard and succeeding as well, to make themselves more innovative, more efficient. And make the products that they're giving, the tool that they're giving to their customers, much safer and able to do cool things as far as research, as far as saving lives, as far as improving procedures. You could hear a pin drop because we all knew that to actually take advantage of all of that, regulation was going to be very blunt hammer. And there was a quite one of our members provided in some other literature and it was, we need to re-imagine software regulation in ways that specifically align with the highly iterative and adaptable nature of its development.

Emma Hossack (13:01):

And that's right. So by recognizing health software and AI, with regulation things that have been around for 20 years, is really not fit for purpose and we have made that point. You asked before about other places in the world, because we're not the first person, first company, first country to be dealing with this issue. Having to understand, we've come to the conclusion as an industry, that Canada has nailed it, their getting right into hardware, we've got blunt hammers and nails. But honestly, Canada took a principle-based approach and instead of actually putting it through regulation, which means that you've got to define quite tightly what you're talking about, as opposed to principles and guidance. They've enabled the industry to innovate, they've provided clinician oversight so that where algorithms can be used safely.

Emma Hossack (14:04):

Let's just say, you're in emergency and you're a specialist emergency clinician, and you want to get some information really fast and you've been in that area for a while. So there's that sixth sense that all clinicians have, which saves so many lives, despite all the barriers that we will put in front of them. You're in that position and if you've got an algorithm that can throw up a whole range of different possible pathways for you, you'll go, "Oh cool." Everyone is not perfect, despite their best intentions and endeavours, so having a tool like that to consult you, is actually not going to make you think, "Oh gosh, that's interesting. I think I'll put a catheter in that patient who's losing a lot of blood."

Emma Hossack (14:53):

It's not going to be bizarre stuff like, which is what you'd be led to believe if you read some of the commentary about what the TGA are concerned about. It's actually going to sharpen up that practitioner. So a number of our members are actually in that area and they're in our industry because I think that they see how important it is when you've got those split second decisions to be made, that you're actually at your best. And the only way you can be at your best, is if you've got all the information to hand and better still, if it's evidence based information, not just something that you learned in medical school six years ago, that's no longer current, relevant or even very helpful. So I think that if we look at the ever changing nature of software and you look at the very rigid framework of legislation, and this will go through as regulation obviously, which will be read word by word and then applied to software which evolves every day and which you want to evolve every day.

Emma Hossack (15:51):

And there's a perverse, I suppose, result in what will happen in practice, because you mentioned before some companies might avoid or turn off their AI capability. Well, not only that Anna, they might decide not to go down that route because it's going to add millions of dollars to their cost space and their clients quite simply won't buy that. So we're talking about dumbing down software at the very time in the world that we need to have the nimbleness, safest, most efficient systems ever, with COVID and that lack of ability for people to get virtual paired, worried that they should get it now. All of those things make the position that we've carried since pre COVID times on this, even more relevant. And as they're like, I guess, with our industry, you see the response with programs like electronic prescribing, where it became really important for people who couldn't leave their home and were going to use telehealth to get their prescriptions dispensed and delivered to them.

Emma Hossack (17:00):

Well, that was able to be done really fast, safely but fast, because there's regulation in place through practice, through professional bodies, through ISO, through community standards [inaudible 00:17:16] seven. But there's not, at the moment, there's not a law saying that thou shalt not have any code that we haven't approved in building the TGA and Canberra. So we were able to do that and deploy it safely, so that you were having electronic scripts being delivered in Melbourne, so just out of Melbourne in Anglesey, as early as May this year. Now with the TGA, they would take at least six to 12 months to look at the changes that would have been required to implement that, so that alone says what?

Anna Belgiorno-Nettis (17:50):

Yeah, let's jump into that TGA topic and particularly the amendments that happened last year. So the Therapeutic Goods Administration, the TGA, as you well know Emma, is responsible for regulating medical devices in Australia. And they're clearly front of mind for the industry, the software industry at the moment because of the regulatory amendments that brought in classifying medical devices with software, from a largely regulated, self-regulated class one, to much more specific classification rules for software related medical devices. And these amendments tend to mirror a much more prescriptive regulation, rather than an approach where the regulator has more discretion in determining what is self regulated and what is not self regulated. You've already mentioned Emma, that you're a fan of the Canadian approach. Could you compare that Canadian approach a little bit more to these amendments that have come through and what you particularly like with how they've gone about it there?

Emma Hossack (19:04):

Yes, I suppose in summary, what we really liked about the Canadian approach is that they looked at it as a whole problem and they didn't see information and knowledge as actually dangerous per se, which is almost how we feel the TGA is hearing information, they're hearing it as something dangerous that needs to be managed carefully. Whereas we're actually taking the view that health information is very precious and needs to be shared judiciously, but as much as possible to avoid people dying and to avoid unnecessary inefficiency. So they took a very different approach because they looked at that something like, okay, how are we going to enable the best healthcare in the world, but also avoid there being any unintended consequences, so have they seen impact on clinicians and patients. And so what they did was they said, "Well, how does this work now, how is the industry working with clinicians now?

Emma Hossack (20:13):

And when I say clinicians, I'm mean hospital, I mean all manner of organizations providing, even lifestyle sounds really casual, like how to wear your hair one day. I mean, when I say lifestyle, advice about how to manage your alcohol or your drugs or your cigarette habit, those kind of things. We have members in our association like Hello Sunday Morning, to help people get better at managing those things. So it's pretty important stuff. So the Canadians looked at this as an existing workflow and how it's being managed and said, "Okay, we really think that we're not the ones to be laying down the rules about how software should be used, particularly given we're not software experts." But the clinicians, they have been using software for 20 years and they know what capability is there and they've got this oversight and ability to go, "Okay, that's really useful and that's great to get that pop up warning me about that drug, but I know in this case that because of the circumstances of the patient, but it is safe / not safe." So they use their discretion.

Emma Hossack (21:24):

So our argument, which is what the Canadians have adopted, is that where there is sufficient clinical oversight, then that is the principle that should be applied, not a regulation written in Canberra, in isolation, about a body of work which the organization doesn't understand, professionally doesn't understand because it's not software, it's not their area and yet they're trying to write regulations that will box it in, in a way that they feel that they can regulate it. So it beggars belief really, when at the same time, Australia is desperately in need of becoming more innovative and productive. We've slipped several pegs down in the last two years alone, in those areas. And perverse as well because we've got a health minister who's incredibly progressive and supportive of using digital means to make life easier for Australians, all of them, the consumers and the people practicing clinical services in every environment, from hospitals to aged care, GP's and specialists.

Emma Hossack (22:34):

So I guess that's where were quite perplexed about the way that the TGA seem to approach it. We understand why they are concerned, but we've made the point that when you've got that clinical oversight, that's a very compelling principle and ignoring that in favor of a blunt legislative instrument, doesn't appear to us to have a balanced result. As I said before, there's not even one Australian dying a year because of software, in fact, many would say that hundreds of thousands are being saved by it because of the way it's able to make events and healthcare delivery so transparent to the people who need to know. That's probably why we like the Canadian approach, it's guidance based and it's got that principle of clinical oversight.

Anna Belgiorno-Nettis (23:33):

The guidance based and the clinical oversight points, how can they potentially feed into where the TGA is up to on its consultation process at the moment? So they're looking at what software devices should be exempted from this new regulation and how have you and the medical software industry gone about consulting on that and trying to feed in some of the points that the Canadian model has?

Emma Hossack (24:07):

Well, we've had at least two meetings, some other software company direct to TGA showing them how software works as well, but we've had round tables with the TGA and explained the position, followed up by papers, which I'm more than happy for you to make available on your website afterwards. There's two major one's, one in May here and one in September, both of which make our position quite clear and suggest to the TGA clear and straightforward way forward. What appears to be the case now is that despite intense arguments to, I suppose, try and uncover the mysteries of software to the TKA and ways forward adopted by other countries, some countries do [inaudible 00:25:04]. So it's not going to be something that is absolutely perfect for Australia, that's just a fact. We've got different payment systems, we've got different practices, we've got a whole range of federal rules and regulations that other countries might not have either.

Emma Hossack (25:17):

So I'm not saying that we could just come pick something from Canada, it'd be perfect, just the general approach, which will be good. Look, as far as the TGA and where they go on now, what they do is they put together advice for the minister now, to recommend that he adopts on the following terms. Now we've put in a position which pretty much summarizes and provide evidence for what I've said online, that we'd like to see it being principle based, would like to see it being fit for purpose so that all of the software, which are clearly not intended to cause harm, administered assistance systems, which are recording electronically health records, all of those, there's a lot of lists, radiology and the other company and pathology systems, some of them are basically replacing really what paper does.

Emma Hossack (26:16):

Those kind of systems should be carved out per se and then there's the more gray area. And that's where I think the real value on the economic side of it comes in, when you said before, premiers might be looking at skipping some of those benefits because of the cost. So where we are interested is to see what boundaries TGA is going to apply. And unfortunately, although we asked them, we said, "Okay, these are the really gray areas where we want to work closely with you to actually nut out the terms of the boundaries that you talk about" but they just say that they'll have boundaries and they'll sort those boundaries out. And that's really concerning to our industry because obviously we don't seem to be now engaged in any draft legislation or draft advice to the minister. Whereas I actually think that that is key, because the organizations that are going to be affected by this and then have that flow on to every Australian who interacts with the health care system, it's our members who manage all of that.

Emma Hossack (27:25):

And I think to put something into the minister without a final check with us, because it might not even be fit for purpose on them. There just may be areas where they had not understood how an algorithm works and there are many algorithms. So they're so far from danger and so far from anything that the TGA is trying to regulate, but it's beyond belief that they will throw them into a bucket, if you like, of possible danger. So yeah, you need the clinicians to answer clinical questions at the point of care and sometimes that would be impossible if they aren't able to avail themselves of the tools that they're used to using. They're used to using them, most of them at least 10 perhaps 20 years. And to dumb everything down at this point, [inaudible 00:28:17] sensitive time in a hundred days for Australia, is of great concern to us. And we're hopeful that the minister will be taking that overarching view of healthcare, technology, safety and efficiency, and applying his very good judgment to that.

Anna Belgiorno-Nettis (28:40):

Mmm. And as you say, the flow on effects are massive, both for business and health and our economy and Australians inside and outside this industry. And we've never really felt the importance of health, I think, more than now globally. So thinking about the opportunity cost, but also the benefit, the potential that we might have if we get this regulation right, what opportunities do you see for Australia in this medical AI space, as we try very hard to climb out of this COVID crisis?

Emma Hossack (29:21):

We are probably in pole position in a number of places. There's the digital health, Commonwealth Research Collaborative, which I think is arguably where the funds have been many equivalent per capita in the world. And that is collaborative funded jointly by the government and industry since 2018. And that is set up specific place to leverage the software products that are being developed or are currently developed and need to be commercialized. And we've got a population of 25 odd million people, which is a perfect population size. A lot of countries use us as tests down because it's a really good size to have as a test bed for mass deployments, apart from anything else. So we've got a population that's a really good size, we've got a government and hope we keep the health minister who understands the value of data and managing health using technology as well, we've got that in place, which is a real plus, not many people get one of those in place.

Emma Hossack (30:32):

So we've got a minister understands. We've got funding in place through the digital health CRC. We've gotten the agency, the Australian Digital Health Agency, which I must disclose that I'm a director of, but I don't see it as a conflict because the aim of that agency is to make Australia have one of the most vibrant digital health software industries in the world. And so we've got that working in our favor. So that's three big positives and then we've got one of the most creative and I suppose, successful health software environments in the world as well, per capita. We've been having digital health employed through primary health care now, for on 20 years, I think 97% of general practitioners use clinical information systems to help them manage their information, but a really good solid base actually.

Emma Hossack (31:32):

A lot of countries are still using paper and I'm not saying that we don't either, like your hospitals, regional hospitals that use paper. But is you've got 25 patients, you're not going to put in a massive system that costs millions either. So it's horses for courses a bit and on the whole, Australia has a really great base. During the bush fires, digital health was able to allow people who had lost their scripts in the fires, to get their medicines safely dispensed from chemists because of the way all of the dispensing software was over Christmas changed to allow them to do that safely and securely and still get PBS payments. So in a matter of five or six days, all of that was changed to enable Australians to have that continuous dispense of their drugs. Looking at the way the industry is able to operate so nimbly and so fast, in conjunction with government being the Department of Health, as well as the Services Australia, you then go, "What's the problem TGA?"

Emma Hossack (32:33):

We've got the Australian Commission of Safety and Quality in Healthcare, they are there to specifically look at protecting consumers and making sure that the quality and safety is the highest possible level. They haven't come to us with any issues about the safety or otherwise, of our software. You've got the Royal Australian College of GP, Specialists, Surgeons, all of those bodies, but all of them use software and haven't raised a murmur and they will support what we're doing. They've put quotes in our submission saying, "For goodness sake, help the software industry, try not to wrap it up in more red tape."

Emma Hossack (33:19):

So I guess, for Australia and for where we are now with post COVID, I would just be imploring, not just the health minister, but also the business, the finance minister, the technology ministers, all of them, to support industry in what is really all about basically, all about being nimble and all about helping all of Australians to manage health more effectively and give them the autonomy, give them the transparency that they need, of their health data. Which will not be possible without using AI, without using algorithms and of course, without having the over sight of their clinicians. So that's a bit of a rant, but it seems so obvious.

Anna Belgiorno-Nettis (34:09):

An important rant and an exciting rant in a sense, for what we might be able to achieve for all of us, both health and economy wise. And as you say, all eyes are looking at the industry at the moment, so it's a really crucial conversation and rant to have. Thank you so much Emma, for taking the time to talk to us.

Emma Hossack (34:35):

Absolute pleasure Anna, good luck with the summit.