SARDisms 06 - Marcus Baw

Transcript

Mariah [00:00:02]
Welcome to SARD’s podcast SARDisms, I'm Mariah Young, and today I've joined with Kevin Monk, managing director of SARD. We both love great technology coupled with great customer service. The main aim of SARD is to help improve the NHS, England's public health service. Healthcare and IT are ever changing and we are interested in the ways that we can help it evolve with the growing population. And this episode, we sat down with Marcus Baw, who is a locum GP and emergency physician who has a particular interest in health IT. Marcus is big on open-source health care and has some eye opening insights into the way that the NHS works. Welcome, Marcus. Tell us a little bit about you and your background.

Marcus [00:00:40]
So, yeah, I've got kind of weird health tech, informatics, Swiss Army knife, career type things. Call myself a general hactitioner-.

Mariah [00:00:50]
Oh I like that.

Marcus [00:00:51]
Which I invented, quite proud of that. But so I started off qualifying in medicine in 2000 and I did a whole load of, like did about 10 years of hospital medicine first. Anaesthesia, ICU and emergency medicine, mainly, and a bit of other things. And then I got to a point where I thought, I don't know if that's for me forever, and decided to have a go at General Practice. Moved down to General Practice and really liked it, and I kind of like the generality of it, I mean, in some ways, General Practice, still not general enough for me. So I sort of generalised even further.

Marcus [00:01:28]
Got interested in the tech side of things just because of the GP’s system. And in hospital, you're always told, 'oh yeah, you know, GPs have got these, you know, pretty good computer systems, you can do audits without having to get paper notes out, and you can do all this incredible stuff'. And it's true. You can do audits and you can do research and things like that without having to get paper notes out and do manual, like, research. But what I was struck by is how these systems in GP...they're all right. But they don't fly. They do not fly. They don't, you know, the expectations that we have in what was about, even 2010, you know, iphone had come out, but even aside from that, we were used to pretty good web applications and lots of good stuff that you could do on your computer.

Marcus [00:02:20]
And you were faced with this GP system, which was really very basic and just a sort of, you know, say administration platform, I guess, which is how they all started. And I got interested in that, joined a few kind of clubs, the primary health care specialist group, the British Computer Society, which is a kind of little group of people who were interested in GP systems, and the health informatics group of the Royal College of GP's. And I started that as like, as kind of, a wide eyed, kind of, observer in 2011 or something like that, and now I'm the Chair of that group.

Mariah [00:02:56]
Oh wow.

Marcus [00:02:56]
Which just kind of, randomly happened, I'm not quite even sure if I should be Chair of that group.. but I am! As kind of deadman's boots as it were, at the last minute. And I was like, well, I'll do it if no one else will. So there I am.

Marcus [00:03:10]
So it kind of just drifted around, not in an aimless way, in that some, I am interested in things and, you know, it sort of drives me to learn things. Along that axis, like, when I was in hospital medicine, I was doing emergency medicine, I saw a guy giving an anaesthetic and he just came down to, like, put someone to sleep because they were, you know needed to go to ICU, like, I have to know how that's done.

Mariah [00:03:34]
Yeah.

Marcus [00:03:34]
So, cue me going off for several years and becoming an anaesthetist and getting exam and all that stuff. And the same thing happened with tech, like, when I was in health tech as a kind of 'GP informatician', which is a word they use for it, which I hate, I hate the word Informatics but I sort of thought, well it's not really good enough to just stand around pointing at stuff and commenting on it, why don't I, you know, I need to learn how this stuff is put together. And so just gradually, you know, I've always been a bit geeky and a bit sort of computery, and I just taught myself to code.

Kevin [00:04:08]
Codecademy.

Marcus [00:04:09]
So I just sat on Codecademy for, like, you know, I used to put in like two, three hours a day. And learnt to code. Did Python, JavaScript, Ruby, and then went on a course called Doctors on Rails, which a mate of mine Ed Wallitt ran. And he sort of like broke the seal on being able to be able to do useful stuff. Like I came away from that weekend being able to write a Web Application.

Mariah [00:04:35]
Wow.

Marcus [00:04:35]
And then, of course, I'm, I'm still even now not someone I would consider like a great programmer.. there are many, many people who were thousands of times better than me. Um, but it's just something I had to do, had to learn how, how it all goes together. And I think that does give you a different insight as well, cos you start to get an idea of like what is a, what is a ten minute coding challenge, and what is totally intractable in, you know, space time, a problem that cannot be solved.

Mariah [00:05:04]
Wow.

Marcus [00:05:05]
You know, you get a different, you get an idea for that kind of difference and that kind of thing, so... And now I just do, I have a portfolio career, as they would call it. So just like which is code for, 'I have 17 very low paying jobs which I attempt to do simultaneously'.

Mariah [00:05:21]
Right. That's amazing.

Marcus [00:05:24]
Along the way, I think the way that picked up by the Kevin Monk radar on Twitter, was possibly my-.

Kevin [00:05:30]
You did, you did-.

Marcus [00:05:30]
Quite strongly outspoken views on open-source.

Mariah [00:05:33]
That's a big thing.

Marcus [00:05:36]
Yeah. It's a big thing for me. I mean, my introduction to open-source was my brother, who's a bit of a hippy, lives in a shed in Sweden, in a very cool way.

Mariah [00:05:44]
That's amazing!

Marcus [00:05:45]
And he showed me Ubuntu about, this is like, more than a decade ago. I was like, whoa, I did not realise that you could get like an entire operating system that comes with all this stuff as well. It's not just the barebones thing. Like really at the time, Windows was quite bare bones. You didn't get much with it. And this thing just comes with everything. And then everything it doesn't have you can just get also free and yeah, some of it's not as shiny as you're expecting from proprietary software, but then it's all free. And if there's something wrong with it, you can get involved and fix it, and suddenly that really piqued my interest. And over a period of a few years I got more into that, then started to think about medicine and open-source and like, at a time, of course, I'd been working with proprietary close sourced GP systems and all kinds of other software for health care and thinking, 'mm, where's the open-source stuff that it's parallel to this, you know'.

Marcus [00:06:41]
And whilst there is some, there are some good projects out there, in open-source, EMR's and EPR's and things like that, none of them are really mainstream. So there's, there is some stuff out there. But I started to really think more about actually medicine as an open-source, science and vocation, and how we might be losing it into the more proprietary software, so built into medicine, when you are trained, your expected to also train other people. It's part of the natural progression of medicine that you become, you know, at some point you're good enough that you've got knowledge that you can pass on to the next generation. And that's all open-source is as it were. We're expected to write this stuff down and we're expected to sort of pass it around and you know, yes, you have to pay for textbooks and things like that. But there's a very big difference.

Marcus [00:07:30]
You're still putting the knowledge out there. The contrary co-option is really nothing short of alchemy, it's how the, the alchemists used to operate. They work entirely in secret because they were worried that someone else would use their findings to beat them to the eventual goal, which was to eternal life and the Philosopher's Stone or whatever they were looking for.

Mariah [00:07:48]
Right.

Marcus [00:07:48]
But no scientific progress was made during that period, even though when you do look at some of the experiments they were doing, they were doing science, but they just didn't share it cos they were terrified someone else would steal it. And once we started openly sharing science and having journals and things like that, the rate of progress has gone exponentially faster, and now we've got things like PubMed, but we still have a problem with general information being locked in proprietary journals. But actually people working around that, both legally and illegally, you know we got open publications where you pay to publish but there's sci-hub as where you can just have it. That, to me, is an important political movement. That's something that has to change about science. And we will not get that by a consensual approach with the general publishers. We will have to disrupt them.

Kevin [00:08:34]
It so struck me as a really strange irony, that the scientific papers were one of the original intentions behind the World Wide Web.

Mariah [00:08:43]
Yeah.

Kevin [00:08:43]
It was built to share scientific knowledge from CERN, from the particle accelerator that was its original intent and the HTML and the Mark-Up language and everything that we share on the Web was about people sharing scientific knowledge, and the funny thing is, if you try and get a scientific paper now, most likely it's going to be a PDF document and not available on that thing it was built for.

Marcus [00:09:08]
Yeah.

Kevin [00:09:08]
I'm not an academic and I don't work in the scientific world, but that's the impression I get.

Marcus [00:09:13]
You're right. That is, it is a total irony. You know, that's happened. And I think disruption is the only way to change that, really, because the journals have had a cushy time for a long time. At one point they were providing value because they were actually the publication platform. They were providing the paper and the distribution. And now we don't need paper and the distribution is the Internet. So you're right, it should not need journals and proprietary platforms, but we kind of stuck with it. I have a high hope that actually what will eventually disrupt that will be the technology that's required inside a paper. So at the moment, you know, papers are still very dry prose and I find them really hard to read. I mean, I would do almost anything to avoid reading a scientific paper and I'm an actual doctor.

Marcus [00:09:53]
Because they're written almost it's like you could have boiled down what they wanted to say into a tweet, but they decide to pad it out into six pages of PDF and a whole load of diagrams. The diagrams are impenetrable. They persistence in sort of exposing you to all the most difficult parts of statistics without really sort of providing what you might call like syntactic sugar on the surface of that, that would make it easier to understand. And yet you could imagine how does the future of publication look like? Well, it could be Jupiter notebooks, which are almost like a Web page that allows you to run Python in a Web page. But you can share these notebooks and you can run dynamic code in them, you can run statistical code in them, so you could actually mix up small amounts of markdown, which I think is the, how the text is formatted in there, with passages of actual showing people the thing.. Cos I think that's one thing that PDF and most, most kind of journal article type things, are very poor at showing the thing.

Marcus [00:10:51]
And it's frustrating. I mean, especially, there are actual journals of health informatics, for example, and they will talk about EPRs and behaviour of clinicians in front of EPR and stuff like that, they will never, ever, ever include a screenshot of EPR. Or even better, a short video of the clinician interacting. You know, we've got all these tools that everyone else is using, you know, to do Tik Toks with...

Mariah [00:11:13]
Yeah.

Marcus [00:11:13]
And yet it's another irony that they just like obstinately refused to use them in the areas where they might be useful for like transmitting an idea, because that is the purpose of all communication isn't it, to transmit an idea from my mind...I have to convey it in some way, get it into somebody else's mind, and it strikes me that just acres of prose is the least effective way of doing that, whereas we've got so many more effective ways we should be using.. I mean podcasts are great.

Mariah [00:11:42]
Yeah.

Kevin [00:11:44]
Hear Hear. The other thing that always surprises me that's missing from, sort of, scientific papers, not like I spend a lot of time looking at scientific papers you know, I always consider myself more an engineer than a scientist, but the data, the data that underpins them. There's always a discussion around it and the interpretation of that data. I very rarely see all of the raw data that was used to process it. And yet I assume it exists in some format that would be really appropriate to, to disseminate.

Marcus [00:12:14]
Yeah, well, I think that you can imagine a kind of Jupiter notebook type thing coming with some data and you'd be able to like play with it yourself and maybe even find out new insights from it. The only limiting factor to that, of course, is that when you get sufficiently rich medical data of any kind, it's very difficult to keep it anonymized. And in fact, you really can't do that without aggregating it at some level. The date of birth that a person delivered a baby, like so that if I know that they, they delivered their first baby and I know the day they delivered their second baby, you can probably identify somebody uniquely from those two dates.

Marcus [00:12:52]
Even more so if you have a location. So if you know they delivered one baby in London, another one in Oxford, you now have got really tight data. And of course, if you have access to another data source like their Facebook feeds of lots of people, they're likely to have shared that life event as well. So you could, with enough data, actually not just say I know this person is unique within a data set, but actually work out their identity.

Marcus [00:13:18]
You could reverse engineer their identity, which is why we're so absolutely vociferous about the way in which NHS data is protected. Because we know there are great research insights and lots of planning you can do with data, with enough detail in it, or even anonymized pseudonymous data. Anonymization and pseudonymisation just as a concept do not exist when you have got rich data of that, of that kind. We just prevent anybody getting access to the kind of line-by-line records, because that's, I think that's quite dangerous.

Marcus [00:13:49]
What I would like to see is better use of the data we can share, because at the moment, you know, if you, there is loads of aggregated data, which is when you aggregate the data and filter out any sort of small numbers and rare conditions, and essentially you've got some really good useful data but it's not identifiable.

Kevin [00:14:08]
And how do you get that data?

Marcus [00:14:09]
Well, you end up getting it in an Excel spreadsheet..

Kevin [00:14:12]
Topical!

Marcus [00:14:13]
Oh yeah. I said the E word. It'd be rude not to go into Excel this week of all weeks.

Mariah [00:14:20]
So important!

Kevin [00:14:22]
We will get to excel.

Marcus [00:14:23]
So, yeah, the open data, you know, is there, but you have to trawl around to find it, and then when you find it, it's in, sort of, formats that don't really help you do anything useful or insightful with it, tends to be just like, here have an enormous blob of unintelligible data. And often you need inside knowledge to really understand how that data sets been put together and stuff. It's not, we don't have that kind of open data toolkit that, that we would really like to see. We don't have open data APIs, for example, where you can just get the data you actually need rather than what we tend to do is drown someone in the entire dataset and let them work out signal and noise, whereas we could give them much more of like declarative interface to it.

Marcus [00:15:10]
Like, you know, what would I like to get? What information do I want? Not here's all the information, see if you can find what you want. Open data could be doing a lot better. And there is still plenty of stuff we could learn from that by using that data better, without having to delve into the line by line, row by row, identifiable records of individual patients, which is really disclosive and potentially dangerous.

Kevin [00:15:38]
I didn't realise actually, I hadn't appreciated that you could reverse the identity. You could, you can work out the identity by going backwards through the data. There isn't something I'd ever really considered. I've been interested in the Vitamin D thing around, around Covid. And there was a paper released that basically suggested there was certainly a correlation with it, even controlling for comorbidities, essentially, if you were overweight, you had diabetes, et cetera. But you, you also had low Vitamin D levels, you were more susceptible to it. And the paper had very, very strong data suggesting that that was the case, but I couldn't actually get to any of the data. It was all their interpretation of it.

Kevin [00:16:22]
And Babak Javid, who's an epidemiologist, I think, a Virologist on Twitter showing him, and he was like, that's very interesting, but this, this result here surprises me. And I it, it surprises me to the extent where I don't, I can't really believe it or it requires more examination and I'm thinking well the data for this does exist. You know, if he could directly get to person A and so it wasn't necessarily identifiable, but him as a virologist or epidemiologist could go in and look at that data himself and interpret it himself.

Marcus [00:17:06]
Yeah. Well that is, you see, that is possible and actually quite topical as well. But this is something that I spend a lot of my working week emailing people about, and it's trusted research environments. So what you are talking about is, you want somewhere where you could, the data doesn't have to leave and you could just execute a query on that data, right? And you could just give the query to the trusted research environment and they'll do the query and they'll tell you what your results were. So you get to have access to the full data, but you don't actually get a copy of it. And that's the real worry that we have with this data, is that the de facto standard for a lot of this research is that big, I'm talking gigabytes of data, of identifiable data, being sort of shipped around the system to different research groups on, on trust basis, on on the trust basis that when they finish with it they'll delete it.

Marcus [00:18:04]
And so far so good. But, you know, as data gets more and more complex and bigger and there are more researchers, sooner or later they'll be an incident and a big escape of NHS data, identifiable NHS data, well potentially the identifiable stuff, would really impact the doctor patient relationship because it's us that put all that data in there, and it would look like we've blabbed. And the entire of medicine is founded on confidentiality agreements so how your data gets into Google or Palantir or Faculty or whatever. So, it has these serious negative impacts. There are trusted research environments which contain the data, and essentially it's like a run- time where you can say, I want to run this query on the data. I don't need access to the data for that, I just have to give you the query.

Marcus [00:18:58]
And normally in my, in my day to day work, I have to actually not name check open safely because I'm supposed to be completely impartial. But actually, Ben Goldacre’s stuff that he's been doing, the evidence-based medicine data labs, the UPN data labs in Oxford, is exactly this. So he noticed that whilst there are some trusted research environments out there, a lot of them are like somebody's personal research fiefdom, and only that research group get to execute their queries on that, on that data. And he has done some really good work to make, to essentially build an open access to trusted research environment. That's all built out of open-source software, and not only that, the actual queries have to be published on Github or in an open repository. So that gives you open oversight as well, like auditing of what queries people are doing.

Marcus [00:19:58]
So, for example, if, if there was some very privacy-invasive query or something that was libertary, whats the word, invasive of your civil liberties, then it would, it would be very obvious what that that query had been..published, so this OpenSAFELY system, will pull in the query from the Github page, and it's just sequel and you run the sequel against the just the research environment and the output of the query that they've asked for gets sent back to them. Now, that output is much less disclosive already, because you don't get access to line-by-line data, you just get access to the results of the query. But even that is reviewed and we ensure that there's suppression of small numbers and rare conditions, and things like that, so that what comes out of OpenSAFELY is... safe and very, very unlikely to result in data breach harm.

Marcus [00:20:58]
And that model just smashes it out of the park for me. It would solve the problem that you just said, which is, you know, if someone wanted to do this Vitamin D follow up study, they'd be able to say, well, we know that someone did query A and I can see that query on GitHub, so I can also build on the work they've already done, rather than having to start my query from scratch, I can actually see what they did and say, look, I want to take that, modify it slightly to do, to get a slightly different insight on this Vitamin D question, and resubmit it back to OpenSAFELY. And because it's all computable and all done automatically, it's really quick as well. So you can do it like in ten minutes or overnight for massive stuff..

Marcus [00:21:39]
Whereas this stuff used to take months and now it takes minutes. That is a step change. And so I'm not normally allowed to be a fanboy of OpenSAFELY but I can be on here, because I do think it's really, really great. And, you know, in general, we have to balance that with other trusted research environments, which NHS Digital have their own but like a lot of these things, they're just not as advanced in the way they built it.

Kevin [00:22:04]
That's why I try to steer clear of it, but I'm also interested in it as an engineer. I kind of like prodding things and going, 'why, is that, is that interesting? Does that work?'

Marcus [00:22:15]
I think the medicine's knowledge base is too closed. You know, even though I was saying before, it's open-source in a way, it's only open-source if you're a doctor in that, like, I still think it's too hard for people to get...trustable medical knowledge and to sort of educate themselves in medicine without having to go to medical school. Like, I think there's lots of people who could learn small parts of it. And we're constantly saying that we like patients to be more generally health literate, as it were. But then we don't really facilitate that. We still encourage everybody to sort of defer to an external body to check this and check this out.

Marcus [00:22:55]
And 111 and a chatbot, and like, it's always outsourcing your health checking, as it were, rather than educating people to be able to do a little bit of that themselves. And that's about how we present the data and information. And so, like, if you go to medical websites, there are two kinds of medical websites, one in appallingly impenetrable language, which is only readable by doctors, and the other version, which is very paternalistic, often in, written in simple language for patients. And I don't mean it's paternalistic in the way it's written, often the language is just written to cope with people's reading ages, it's just that at some point in every single patient guide it will say, OK, we've taken you as far as we can go here, now go and see your doctor. It's like a bailout.

Mariah [00:23:48]
Right.

Marcus [00:23:50]
You can understand why, because you could do a lot of harm by not encouraging people to seek medical advice at the right time, but at the same time, we're not bridging the gap between the highly technical version and the kind of patient version were we can't, we assume we can't teach you anything. Where are the stages in between that?

Marcus [00:24:10]
I mean, if you compare it with tech, I was able to teach myself to be an acceptable, you know.. good enough to be dangerous, kind of developer, for free. On the internet. And I'm still, still learning, just like I'm still learning medicine. We don't have that for medicine.

Kevin [00:24:30]
No, no. It's that hybrid. You're a very useful person to me because you're somebody who sits in the medical world, but you also have an appreciation of the technical world. And I think the technical world has always been very open in encouraging, you know, anyone can go on Codecademy and start to learn to code, you know, and I'd, I'd kind of like the opposite way round but I think it, it, there is a tendency to say, stay in your lane. Like, don't, don't get in, don't become an amateur medic. And I'm like, well I'm just interested and I can't help myself. I can't start...

Kevin [00:25:12]
And you do have expert patients. Friend of mine is a GP, she was telling me about the expert patient she gets in these days that come in to her surgery and they, they know some detail about their condition, and she kind of almost sends them out of the room and is quickly on Google like, what the hell are they talking about?

Kevin [00:25:33]
You know, because it's, they're just such an expert in their own condition. I had my own strange setup I got kidney stones and went to the doctor, and most...most times kidney stones it's X, right, but I had calcium oxalate stones, and...it was funny because I went on Google and I was looking up what, what, what foods have got oxalate in..

Marcus [00:25:59]
Rhubarb.

Kevin [00:25:59]
And it was basically my daily, it was my daily diet. So I was able to go to the doctor and go, I think this might be the problem, is I've been eating kale, spinach, you know, everything, seeds, everything that's got oxalate in it. I think this might be the problem doctor. And I was talking to a consultant urologist, and he said, in my head, there is one percent of the population that has this problem, that they eat so much spinach that they end up with kidney stones but I've never met one. You're the first..

Mariah [00:26:32]
Way to go.

Kevin [00:26:32]
You're the first person.

Marcus [00:26:33]
I love expert patients. I think, ah, you often learn from them. And I think it's really cool to see and I want to see, um, platforms that help both the patient and the doctor in that situation. So I think we talked about how the medical knowledge really isn't like democratised enough so that the patient can level up through their medical knowledge. And in stages they have to jump from like the patient version to the clinical version with no gap. And so I think that could be improved.

Marcus [00:27:04]
But also, you alluded to a point where, as the GP friend of yours is, like, thinking, I don't actually know about as much as the patient knows, what am I going to do? And I think that we need to move, like, move the conversation to a much more mature one in health care and say, you know, the model of medicine where you learn everything and memorise it, at medical school and you know you read journals and memorise everything. That is an outmoded model. Just because of the sheer volume of medical knowledge that there is to know, and the turn over time. You know, the half-life of anything you learn is now, you know, it used to be, if you learned it at medical school, the half-life of it was like 20 years, you know. After 20 years, it's half as useful.

Marcus [00:27:49]
Now that the fade is so much quicker, like something you learnt just this year could be overturned and changed by new knowledge in another two, three years or so and sometimes faster than that as well. I mean Covid, particularly, has been like every month... The half-life of Covid knowledge is, is changing. So we, we need to move to like, and this is, this is another tech, tech medicine crossover thought process, like just in time knowledge. So you know you've heard of just in time compilation where, you know, the compiler is just compiling the stuff it's going to need..as it needs it.

Marcus [00:28:26]
And whereas medical knowledge at the moment is like you compile everything at medical school, once you've got your binary block, that's you, starter forever, not quite like that, but I like the idea that we move the conversation to be much more accepting that your GP will know stuff and they are your guide through your illness and through the health care system. But they can't know everything. And so it's completely OK that they might Google stuff, or even better, you know I said about the GP systems do not fly, that's because the GP systems don't help you guide you through medicine as, as a GP, when you're in the middle of a consultation, you want the kind of sat nav of medicine, that helps you to navigate inside good quality knowledge resource, something that's going to help you right then and there, in front of the patient, what you call just in time knowledge.

Marcus [00:29:18]
So the patient is an expert patient and you're thinking, what can I add to this? And they obviously know a lot, but there is some, there's always going to be some context or just the objectivity that you're not the one with the disease. That is an important part of the medical consultation. And combining that with being able to like use your GP system to give you just in time knowledge about that rare condition, means you like supercharging the consultation. It's like machine assisted human. Phrase I use a lot. It's my counterattack against the the AI people. Because I am not an AI believer, to be honest. And, and I'm very wary of people saying, oh, you can use AI to solve medicine. And soon doctors will be obsolete. You know, there’s millions of arguments why that's just not true.

Marcus [00:30:07]
But, why would we need to jump from, like, a human using their wits and a computer system to AI, with nothing in between? Between those two extremes, there is a whole spectrum of machine-assisted hue. We give better and better machines to doctors so that they can do a better job. You know, whether that's like just in time knowledge in a normal consultation or a tricorder, like Star Trek, that can see inside the patient and give them more information about what's going on. There's a whole gradation of levels of machine-assisted human that make you better and better at medicine. Before we have to even get to the point of an AI. And it seems like no one seems to want to talk about the machine assisted human in between. They want to talk about, it's either self-care, with IOT devices or it's AI. and nothing in between.

Mariah [00:30:58]
Nothing. Yeah.

Kevin [00:31:00]
You know what, it freaks me out how much we, we have shared ideas about things.

Marcus [00:31:05]
Cool.

Kevin [00:31:07]
I wrote a blog post about artificial intelligence and rostering, and on our roster and we have this concept of centaur, Centaur Rostering. I don't know if you ever heard of Centaur chess, but essentially Centaur chess. This is chess competitions where the computer and a human chess player, play alongside each other and actually the human and computer can beat both the computer, on its own, and a human. So, the centaur approach to many things is actually a stronger way of doing things than just anything in isolation.

Marcus [00:31:42]
Machine assisted human yeah.

Kevin [00:31:44]
And the term centaur, comes from the idea that you've got the legs of this horse, you know, it's not a particularly smart animal, but it's got this big, powerful legs and it can get you so far. But you need the steering head of a human on top to, to make it go in the right direction. And when I think about a doctor in a kind of futuristic medical practice, I also think, you know, AI is often a marketing term and overblown. But it's because it's seen as these two opposite ends of, I either have a GP working, doing things traditionally, or I have this AI bot that does, that tries to do the GP's, er job, and I think well actually no, it's both, both those things.

Kevin [00:32:33]
I could imagine a surgery where someone comes in and the AI has seen their blood work, in, you know, its liver, their liver function, and can say, OK...they've obviously got a problem with their liver here. You know, they've got bilirubin through the roof, and the yellow, but it takes the doctor to see that, see the results in that AI system, and notice the brown paper bag with the bottle of whisky in the corner, and be able to put that in context and say, actually, there's a social problem here. There's some social prescribing, there's, they've got problems at home, you know, to see, to have that system that can work it out, but you go, oh, yes, yes.

Kevin [00:33:18]
And I can see the broader context because AI and technology is very bad. That's the in the broader context of things. It's why they're not funny. It's why computers aren't funny. It's why Data in Star Trek isn't funny. Is, is why every robot in every film isn't funny, is because they can't see the broad, the context of stuff, and comedy is about seeing lots of different things and pulling them together in ridiculous ways. And yes, that's what humans do really well, is they're broad beige thinkers as they see all of this stuff, and...consider it all...in separation and bring it back together. So I'm a big fan of-.

Marcus [00:34:00]
Yeah machine-assisted human.

Kevin [00:34:01]
Assisted humans. You know, the, the, the doctor with the big horse legs, to propel them forward.

Marcus [00:34:10]
I love the analogy actually. I've not come across centaur chess, but I do like the idea. It's another, like, way of assistance isn't it.

Kevin [00:34:16]
We call it the Centaur Rostering. So a big plug for our rostering system here, but might as well hey? It's, the idea is that the AI solves the roster for you, but often you've got someone overseeing it who can look and go, actually I know that person can't do such and such a day because they've got a surprise birthday party coming up, something that the computer couldn't know.

Marcus [00:34:42]
Yeah.

Kevin [00:34:43]
And so, you know, you've got that adaption.

Marcus [00:34:45]
The human, it's the human side of it. Been quite impressed actually by the stuff I've learnt about rostering and appraisal stuff, just by listening to your other podcasts, and actually the way, you know, business, that you've had like different people from different parts of the business, and like their insights been quite interesting, and took away... The guy who was talking about, um, We Are Lean and Agile.

Kevin [00:35:12]
Oh, yeah, Andy.

Mariah [00:35:13]
Andy.

Marcus [00:35:15]
That really spoke to me, because, like, one of the things that I kind of get very exercised about is like, why aren't things done better? You know, if only I could, we need a central, we need somebody to do it properly or like an organisation or, you know, the My Society of Medicine. I've thought of so many different ways of slicing this up. And he made me think in a completely different way because he said, well, hang on, I don't want- he actually said, I don't do consultancy, I'll run screaming from consultancy, because that tells me that I'm not doing my job well enough.

Marcus [00:35:48]
And he's inverted it all, and teaching the end person who needs to know how to do it, like how to be lean and agile and design processes that work better, and if they have to keep asking him for consultancy, then he's obviously not done the teaching well enough and goes back to them and does more teaching. And I thought, that's a much more scalable approach than what I'm trying to do with whatever it is. We want to upskill all these organisations like Public Health England, for example, to not need to use Excel for internal data transfers, which is what it's always, it's from one system to another. How do you do that? Well, if you go in as a consultant and fix it for them, when you go away, they're just as dependent as they were before, but if you train them up, then you can walk away and then not only that, but as they get better, they'll go and train some other people and you've created a ripple that goes out from a good idea.

Marcus [00:36:47]
And.... Then that started me thinking about Advanced Trauma Life Support, which is, feels like a bit of a crushing handbrake turn from where I've just been talking about. This guy that started the advanced trauma life support movement, and actually it all started because he crashed his light aircraft in a field in Iowa, in the middle of nowhere, and I think some of his family members actually died in that accident and the remaining survivors were really badly treated in terms of the medical care they got in a small local hospital, because small local hospitals were just not geared up for trauma and they saw, like, no trauma.

Marcus [00:37:31]
And so they didn't really, although they kind of knew what to do, what they didn't have, was a framework for thinking about it and some of the debatable decisions just taken away from them, and so this guy just said, right, well, I'm going to come up with a framework, he was a surgeon which helped I suppose, he came up with a framework for trauma care. And from there, you know, that's how we got this... You might have even seen like paramedics doing sort of like airway, breathing, circulation, ABCDE, assessments of critically ill people.

Marcus [00:38:02]
And that whole process really came about from what was popularised, let's say, from the ATLS movement and we now have advanced paedeatric life support, we have advanced cardiac life support, ACLS, all these things stem from the same idea that, like, actually you could spend a lot of time arguing what are the priorities for any given trauma patient, cos everyone's different, but actually in the argument, during the argument process, patient is dying, bleeding out in front of you. And so what this guy did, was just said, look, I'm going to come up with a good way and I'm going to write it all down so simply that you can memorise it and you can just do it, like, out of your spinal cord. It's a, it's a reflex.

Marcus [00:38:45]
You don't even need to think about airway, breathing, circulation, how to do all those things, how to fix things that go wrong all the way through. And you've got a stable patient now you've finished ABCDE, and they can now have more. That's when your thinking can come in, after the first, like, bit. So I was thinking, if you take the tech parallel of that, like where we're now saying to people like, look, you need to build better ways of not using Excel. Go figure. And of course, what do they do? They go away and just like arguing over the trauma care, you've got like three experts thinking, well, no, I'd definitely build it in Golang, and then the other person says, no, no, no, this is definitely a Neo4J problem, and you don't need relational whatever it, just shut up, shut up, shut up.

Marcus [00:39:31]
What we need is A good way and pick it, pick one and say we accept that there are thousands of other ways of doing it, but what we're going to tell you is A good way. That you can definitely get a success out of this, just like ATLS. Like you can argue that some patients need circulation dealing with before airway. Right. And that's quite common. But the framework is still airway, breathing, circulation, airway, breathing, circulation. Never changes. And now that, that you can debate about what we know, if that's the right thing to do, on a battlefield at ATLS, you see first, because that's a special case. The We Are Lean and Agile approach of, like, let's teach people something, and let's get them to propagate that outwards. Let's make it simple enough that it's not like you don't need a PhD in it, just like a two-day course.

Marcus [00:40:25]
And then let's give them a simple, good way of like, if you need to check move data from here to here, you need to do it via, like, we're going to make some, we're gonna take away some of the ambiguous decisions for you, or some of the debatables. And we're going to make arbitrary good decisions. Now other arbitrary good decisions are available. But these ones are the ones you know. You know? And like you say, right, use Python, you use a rest API, you document it using Swagger, you literally, these are the things you must do. And, and that way you take away all that debate that leads to six months or twelve months of delay or delivery failure. And you just say, look, well, this can be done now by a team over five days, because you know exactly what they're doing.

Marcus [00:41:14]
And the other thing that ACLS does, is that the anaesthetists, the surgeon, the off-peak surgeon, the A&E doctor, the nurses, they all do the same course. Exactly the same content, and they all get assessed in exactly the same way. And, and there's a good reason for that. It means that if they went on subtly different courses, like the anaesthetists did a special version of ATLS that was more about airway, and the surgeons did something that was more about circulation and abdominal injuries, then there would still be these arguments about what's the priority now. But because they've all done exactly the same course, everybody's on the same page. And so I thought, right, so if you had that, like, this is how to sort this Excel, let's, you know, we have to get rid of Excel interoperability, fudges, out of all our public organisations, and it isn't just Public Health England, it's, you know, it's every local authority. It's every NHS trust. It's every police service.

Marcus [00:42:10]
How do we get rid of it? Well we send everybody on the same course. The Procurement team, the HR team, the developers, the subject matter experts or commissions or whatever is that, you know, if they all knew the same priorities, then when it came to buying the capability to build that thing, you'd go.. our HR people would go, Yeah, I totally see why you need a developer, you can't do this with a business analyst, you need a developer. And, and then your, your finance team will go, yeah, OK, well, we, we can see how there's a short term cost but of course, we're not going to get fined by the ICO for half a million quid in a three years time. So, you know, that you can start to get everybody talking the same language. And even though there are still thousands of other ways you could have dealt with that problem, it's the collective action dilemma. Everybody's trying to work out the best way forward and no one wants to stick their neck out. That's what's actually stopping us solving this, because we all know that you could get rid of Excel in a very short space of time if everybody had a good way of doing it. But at the moment, Excel is their good way of doing it.

Mariah [00:43:18]
Exactly.

Marcus [00:43:18]
Unless you give them a better good way of doing it, they're going to carry on.

Kevin [00:43:22]
Well actually something that comes out of our shared heritage are our interesting in Ruby on Rails is it takes that approach to tech, right? It's one of the first things it will tell you on that framework. It's a, it's a, it's a coding framework but they seem to build Web applications, and one of the guiding principles of it is convention over configuration. So DHH who created that framework said, you know, if you go into to McDonald's and buy a Big Mac, you know what you're going to get in it. You know, the, the core ingredients, you don't go in and say, I want one beef patty, I'd like, you know, the bread in the middle, and then the gherkin on the top, and then I want 50 sesame seeds on top of that.. What you do, is you go in and you say, I'd like a Big Mac but I don't like gherkins, so take the gherkin out. And that, so you've got-.

Marcus [00:44:15]
Convention over configuration.

Kevin [00:44:17]
The convention over configuration so that you don't have to go through all of those default thought processes for, for everything. So if you've got a data problem, and how do you present data, here's our, as you said, here's our standard set of things. Make sure that you've got an API on it. Make sure it's documented in this way. Make sure you store it in a, in a relational database by default.

Marcus [00:44:42]
Take away some of those mind-numbing boilerplate decisions that aren't really that necessary. There are always people who will do it a different way, but that's fine, that's completely fine. But I think it's like, what is the one simple way that everybody knows, well it's Excel, and that's why it gets used. So if we gave them a new simple way, that was better, by a couple of orders of magnitude in all the different parameters of security, then they would start using that. Eventually. I love the opinionated framework type of idea. That's kind of it.

Kevin [00:45:17]
We're both quite passionate about getting open-source technology introduced in to the NHS and just the barriers around that.

Marcus [00:45:28]
That's it. I knew that there had been a lot of people talking about open-source barriers, and I just decided to completely sidestep that and say it's not about the barriers, it's about why you must do it. So, for example, you could say the same thing about drug development, like drugs, that if you left it to the pharmaceutical industry, all drugs would be completely proprietary, secret molecules, and you would not be allowed to know what the formula was. It would be a brand name and you would buy it from a drug company, and you buy it from them forever. Right. But at some point, medicine said, no, you must do it this way. And so I took that approach. That the approach that no, we are not going to allow medicine to become subsumed into a proprietary software establishment. Medicine is medicine. You happen to provide software for medicine. It's not the other way around. Right.

Marcus [00:46:19]
And we, we own you and we want to know how you build the software, we want to see that from a safety point of view, from an ethical point of view, from a developing countries access point of view, from a medical knowledge point of view, so I just took the, I'm not even going to go near barriers because there are loads of barriers, there's always barriers. You can say there's barriers to, like I mean, there barriers to why drugs, when, when the, the patent system was, was used to limit the length of time the drugs could be exclusive. That's a, that's a barrier. You know. Nobody really wanted to have to put in that legislation to force drug companies, but somebody had to do it.

Marcus [00:47:02]
And I think we're going to have to do the same in software. And at some point, it could just be like, if you're making medicine software, it has to be open-source, that's the end of the story. The slight complication we have in software, is software is not patentable, it's copyrightable. And then laws are very different. Copyright lasts for a lot longer. So whereas patents give you an exclusivity period and when the patent expires anyone can make your drug, which gives you a nice balance between rewarding drug companies for the R&D expenditure of developing that drug, and then marketing it, doing the research to prove it works, or not - cynical view, and then... And then at some point, everybody can make a similar drug.

Marcus [00:47:51]
With copyright, it lasts till something like, depending on jurisdiction, like 70 years after the death of the last living author. And of course, if, if the author is a corporation and their 'people', in most corporate law, then it never dies. And therefore, 70 years after the death is just a moot point. It's basically always going to be proprietary. So we have to find some way of making it better. And I actually don't think the legislation is solution here. I think you manage it by only buying software from people who are open-source. And you say, we're not gonna. If you won't play the game the way we want it played, you can go off and bankrupt yourself. But I think it's a, it's a slow burn that particular argument. It's going to take a while for me to win that one.

Mariah [00:48:41]
Yeah. If there's one thing that our listener should know, what would that be?

Marcus [00:48:47]
Gloves. Bizarrely. Which is that like, I'm a big fan of Ruby Rogues podcasts, which you might know about. Which is one of my favourite podcasts ever. And they used to do a picks at the end of, like, things that would change your life, that were really cool. And I climbed Scafell Pike, and I wore some lovely fingerless woolen gloves, and they changed, changed my life. Because I've got Reynaud’s syndrome, which means that generally on a cold, wet day, doesn't have to be that cold, just wet really, windy. My circulation to my fingers will completely go and my fingers are numb, and somehow numb and painful, even though you'd think numb would mean not painful. They're numb, painful, horrible feeling, and it won't go away until I get somewhere warm. And these gloves were, like, wow.

Marcus [00:49:30]
So, maybe that's my tip for Reynaud’s sufferers, or anyone with cold hands...

Kevin [00:49:35]
I love this, this is your message to the world, get fingerless gloves.

Marcus [00:49:42]
And, like, slightly less frivolously, I just think the thing that I've been thinking about a lot, given the Covid situation and how everyone's working, is how, you like, everyone thinks that they've done really well because they're doing meetings online. And I just think meetings must die. Meetings are the... If you're carrying on, doing meetings, but just doing them remotely, you've achieved nothing at all. Like apart from the tiny carbon footprint reduction maybe, and maybe save some time, whatever. But it's the whole thing of meetings that, you know, so much of the time is, is valueless. So much the time is, is not really productive for any of the people who are there, so much of what is discussed could have been done asynchronously better, because again, it comes back to the, sort of like, prose versus being able to show a multitude of different media types.

Marcus [00:50:43]
You know, if I've, if we discussed an item, every agenda item in a meeting could be a thread on a forum, and you could start those threads at any time, people can contribute to them at any time, we don't have to wait for the meeting. You might come to a conclusion and say, right, we've decided what we're gonna do about this, close the thread. Decision made. You also, now have...a record of everything that everybody said in relation to the issue, attributed to them time stamped. Everything is, it's like the best minutes you've ever seen.

Mariah [00:51:14]
Yeah.

Marcus [00:51:15]
And, and it's done. And it might not even have needed to get to the meeting. So you've just removed something from the meeting agenda. So by the time you get to the actual, say you decide that you are going to still have a meeting anyway, you've pre-discussed everything, everyone's familiar with the details of the issues, the agendas are not just a dead piece of paper they're a live document that is a list of these, you know, active issues. And you can also embed things like, I can put screenshots in there and illustrate what I actually mean, rather than just droning on in a meeting. You can put video in there. You can put links in there, you can, you can link to issues. So having discussed something, decided it three years ago, someone comes along and says, why are we doing it this way? I can go 'link', that's the thread. Of exactly why we decided to do it this way. However, maybe something's changed, maybe we reopen that, and we are managing organisational issues like Github issues, right, and now you're in the twenty, twenty first century. You know. Congratulating yourself because you're doing remote meetings is...hilarious. But all meetings must die. That's a much better...

Mariah [00:52:32]
Fantastic. I love it. Cool. Thank you so much for coming on.

Marcus [00:52:35]
Been a pleasure. It's been a pleasure. And I'll be back.

Mariah [00:52:38]
You will, you will be back. Thank you to all our listeners who tuned in to today's episode of SARDisms. We only covered the tip of the iceberg and look forward to having Marcus back in future episodes, which would include discussions around APIs being used in health care, projects that Marcus worked on that didn't go so well, but learnt a great deal about, as well as cryptography. We look forward to welcoming Marcus back soon. You can find out more about SARD by visiting SARDJV.CO.UK or send us a tweet on Twitter @SARDJV and use hashtag #SARDisms. Until next time. Have a great week.