I kicked off the afternoon with a whistlestop review of currently extant open source healthcare projects. A room change meant I had a few technical problems during my talk. I also discovered one of the dangers of trying to give an overview to experts: there's always someone knows more about you do about certain aspects of your talk. Still, I think it was a useful discussion. There's a handy table of links to the projects I looked at.
Chair: Welcome everyone to the NHS Session. It's enormously heartening that the NHS is actually having such a session: it is something that we were only dreaming of in ther OSCHA Rome last year, and now we are here, and the NHSIA for organising and sponsoring the meeting, and thankyou all for being here.
One thing that happened after OSCHA was the NHSIA had a board briefing on the topic of Open Source and its role in the NHS. There are several familiar faces from that meeting here, including Douglas, who at that time was an Associate Editor of the British Medical Journal. His talk is considering the role of Open Source in the NHS, and we thought that a good way to start is would be to think, what projects are there, and what do they do and what state are they in?
[I am fiddling with my Mac trying to prevent it from sleeping]
Douglas Carnall: I normally pride myself on not getting involved in this kind of behaviour, but the room changes, uh...
Chair: In the second part of the session there are 4 presentations on some of the open source projects that are running in the NHS and some of the problems--given that there is an open source solution--and supposing for a moment that it might be even be the best solution available--how do you actually manage to get that to happen in an organisation as large and complex as the NHS?
Douglas Carnall: Well, thank-you Adrian, I'm not sure if I can answer that question, but, but I will try to illuminate in my talk my understanding of where we are at with open source in healthcare in 2001.
Perhaps a little bit of background would be useful so that you know about me. I am currently working as a general practitioner in the East of London, and my background is in medical journalism, and I got interested in this last year when I wrote an article for Linux User about it. I think that article--you may have seen it--was quite seminal in terms of helping people to understand...
Douglas Carnall: [worried] I'm sorry, is there someone who could help me? It would be really great if I had my power supply, but that's something that I lost in the room transfer... is there anyone here who might be able to... it's in a rather scruffy bag at the front...
you'll recognise it... it's totally grungy... just think "grunge"
it's a security measure to have such a bag...
What I did for this talk today was to revisit all the people in my article, which focussed on, what use is being made of open source software in healthcare at the moment, and to try and expand on that, and also to talk a little about methodology. So basically my talk is based largely on searches made using Linuxmednews, which catalogues currently extant projects in open source healthcare, Yves Paindaveine's exceptional resource--probably another webpage that you are familiar with, my Linux User article telling us where we are at in 2000, and also a look around the web.
So, being asked to do a talk that is an overview is a bit of a poisoned chalice. I'd like to apologise in advance if I missed your project off. That in a way I think reflects the difficulty of being a situation where we are not quite self organising, we don't quite have community awareness. My impression is that there are a set of disparate projects out there, and though they all use open source software, I don't think that they necessarily think of themselves as a horizontal community across healthcare. They are in focussed areas using open source software quite happily but not necessarily thinking about the organisational implications of that.
So... isolated projects, disparate objectives, and we're based on anecdotal evidence. I'd love to show you some graphs and charts with lines going up like that, but I don't have them.
I've taken my classification from the Linuxmednews site: which is basically to have: EMR systems, infrastructure, bioinformatics, medical imaging, and also added another category, which I think is very significant, which is the publishing category, which is very important as we get the merging of content and software, for example in decision support software systems.
If we think back to five years ago, decision support was a paper based system: what changed a GP's practice? In theory, we read an article in the BrJGP or the BMJ, think "Oh well, I didn't know that, something's changed;" then we change our practice. For example, the BrJGP publised an article in September this year; Adrian will know; it's a big ritual in general practice that if someone comes in with earwax, they can't hear, we send them off to soften the wax with oil for a few days, then come back to the nurse for syringing. In fact, there a paper in the BrJGP in Sept 2001 showing that that's a complete waste of time and we should just get on with it.
That's the kind of thing you need to know, and obviously decision support implies the need to...
uum... OK I'm hoping this will be fixed shortly... I don't have the power supply... I do now...
...plugging it in...
[plugging it in]
yes please do plug it in. That will cure all our problems...
...will it reach there?... can we unplug that?
D: No that's the minidisk, so...
[Mac voice, Victoria]: "Alert. Are you sure you want to shut down your computer now?
D: No I don't.
Audience member 1: She sounds very nice!
Audience member 2: just wrong.
Douglas Carnall: Uuu. Were was I? So, you can see the outline there...
Brief namechecks only, but these are the significant systems that I found out there. There is GnuMed, ironically presenting next door as I speak, a paperless medical records system, licensed under the GNU General Public License, and using Python, PostgreSQL, which are all open source freely available software tools. It's at early stages, I don't think they've even got a demonstration record out there. I don't think there's even one example of the system working in practice...
But very promising I think...
Tk Family Practice again another EMR system, again licensed under the GPL again, using Tcl/Tk another open sourced freely available technology. I think that is one with an installation perhaps of a handrolled tool. And that's a very common experience, which is that you have a lot of people out there who basically wanted a solution and they sat down and they rolled their own. And for me the interesting thing is when you get into installation number two.
Audience member 3: Have you got any idea of how many installations there are?
Douglas Carnall: As far as I am aware there is 0 or 1 for GNUMed and 0 or 1 for Tk Family Practice.
Audience member 4: There might be one in France, in Bordeaux and Marseille installed.
Douglas Carnall: This is one general lesson that I have learned from preparing for this talk is that you could visit these people's websites and never know...
That's very important I think, and is one of the messages of my talk...
Can we just run through these? Vista, we've already heard about this morning and I don't want to talk about it any more. The licensing is not clear on Vista, it is public domain I think, but I'm not sure quite what that means... I assume it's some sort of BSD-like arrangement.
AM5: It's free
AM6: Send off 5 dollars or something, and they'll send you a CD with the whole lot on it. You can download it...
DC: Yes, they have a very good website on which the code is available.
AM7: It's available under the Freedom of Information act which puts it in the public domain.
AM8: Except for some crytographic software which you can't
AM8: Yes, there is a security hole that forbids you to plug in the hole the cryptographic module, so it's a bit more complicated.
D: OK, so this a very impressive project, and even if we're not quite sure about the licensing, we think it is basically free in the sense of free speech. It's written in MUMPS, an arcane development language... they do their thing. It's worth mentioning that they have more installations than just the two in Finland and Germany, they have more... they obviously have their organisation wide installation... it's interesting, the Americans are rampantly against socialism unless it's in the military
they've got an NHS, but basically it's for the military. The Veterans Administration isn't just veterans, it's all their families, and it's a significant healthcare organisation and the same thing applies in defence contracts--everyone has got to compete, except in defence where you can have statewide monopoly funding rather in the manner of the former Soviet Union..
anyway that's the Americans, let's not talk more about Vista because we know where they are...
Cosmos, I came across this... this is an example of a project that is like many sites.. they're on SourceForge, you can go and have a look at them.. they show up on a search for healthcare software or whatever.. but you go along, and there's absolutely no evidence--and absence of evidence is no evidence of absence--but no evidence on SourceForge, and this is perhaps a fault in the SourceForge design--of what use is actually being made of the code. You know, someone has kicked off a project that seems to use some variant of UML, it's licensed under the GPL, and I suppose the only way to get any more detailed evaluation is to talk and suck it and see, if this thing seems to be closely aligned to what you are doing.
OpenEMed, again, I think there are 0 or 1 installations of this. The object managment group's standards are probably something that you know better than I what they are. Some sort of middleware layer in software as I understand it...
AM10: That's the one from Los Alamos isn't it? I have a copy of that, and it is working in some places as a prototype and it produced a working system in the [inaudible] hospital in Denver, so it is a working system, but it is complex and difficult and I haven't got it working...
D: OK, so that's licensed under a BSD-style license. The price of a BSD license is that you need to have line of code that runs by in the boot messages crediting the original authors of the software but if you are willing to accept that limitation on your freedom then in many ways it's a freer license that the GPL because you can take code from BSD--Berkeley Systems Distribution--a variant of Unix-- back into the private domain if you so wish...
Arachne, again conceived conceived as a distributed clinical workstation, so an EMR system, under a BSD-like license. Again, that looks like it's a roll your own solution: I'm not sure how many organisations have implemented that...
Freemed, is a PHP system... they do have some screenshots of their system up on their website, licensed under the lesser GPL, which is a subtle distinction which means it can interoperate with other systems which you don't want to contaminate with the GPL. Written with standard browser-based interface tools. Again I think this is in an early stage. I saw a presentation, admittedly last year, where they showed some fairly basic EMR functions up and running. That was a handrolled product, Buchbinder? doing that in his Emergency Room somewhere in America.
FreePM, again, it's out there, it's in an early stage of development.
AM12: FreePM does give the appearance of being able to be installed and used; they have a company that offers support, though how many installations they actually have, again, I don't know. It's been reported as "finished" at least the first stages of it.
Littlefish, correct me if I'm wrong, but this is an older project, and an example of a situation in which the prime mover of the project has moved on and the reins have been taken over by another open source programmer, but I haven't seen any evidence of new development on the site since that happened, so quite what's happening there I don't know, but that definitely has been deployed in a number of rural primary care centres in Australia. A pragmatic choice of the Delphi database running on Windows and I think the underlying data structures are influenced by the GEHR project-- the Good Electronic Health Record
AM13: This is a good example of a project in which the developer decided to stop working for private reasons--for family reasons--and he handed over the management and all the code to the new benevolent dictator for the project--this guy happened to be in Belgium--and at that time his job allowed him to do it, and the same month that it officially happened the nomination in another job, to another job, which did not allow him too much time, so he handed over to a German guy, so at least, things continue.
D: I'm conscious of the time and there really are a lot of these so I want to...
MDSchedule: that's in a very primitive state as far as I can tell. There is very little activity on the website. Sometimes I wonder whether people set these up as kind of placeholders or set up with the best of intentions, but only a single developer or very few developers, and then you never quite know--it's not really possible to tell, where it's at.
Meditux, this is actually a functional project in a British hospital I think, a Java-based tool that compiles on all their platforms: they have a mixed environment and they were able to use open source software to integrate across the environment; again, single unit, single enthusiast, single department--I don't think there's any evidence of that having spread out into other locations.
So, let's get on
I was asked to mention Prodigy, by the way, all of these projects I have on a table with URLs and so on, so don't worry about scribbling them down, but I will put them on my website.
Prodigy is a computerised clinical decision support system developed here in the UK in Newcastle Department of General Practice and is the jewel in the Crown Copyright, if I say that. It is a very mature product that is integrated by commercial suppliers into their systems, and is freely available on their website. You can examine all of the content which is extensive, peer-reviewed information, for example, one example I looked at was otitis media, an infection of the middle ear in children, and it takes you, quite a nice interface with frames, through the logic of treating that child, prints a patient information leaflet that supports the choice you have made, advises you about which drugs are most cost-effective and most effective for the patient. That's the information though, there's no integration with the clinical suppliers as far as know. I suppose the next logical step is to automate that through so that it fills out the prescription form with the recommended.
AM14: I was speaking to someone at MedInfo yesterday who was telling me that they have got it integrated to the degree where someone can click through...
D: Well, there's embedding and there's embedding isn't there? In many ways the commercial suppliers are value-substracted suppliers in my opinion. For example the system at one of the practices where I work--I work as a locum so I have the opportunity to use a number of systems--has got system guidance, but they're unattributed, so I say, "Well, guidelines.. who wrote those? I don't care." But if I know it's a Prodigy guideline and they publish accounts of their peer-review process and the way that they assemble the information, and I know it's kosher, then I can use that information. If you strip that off, and pretend that it's part of the system, then that to me devalues the information in the system.
AM15: Do you know the terms of the license? I mean Crown Copyright isn't a software license. You could issue to the suppliers, but I don't know, is it GPL.
AM16: There's nothing in Prodigy that is open source. It would be economical with the truth to describe it as freely available. The only thing that is freely available with Prodigy is the actual data. There's a data diagram, here are all the files, it's up to you to integrate that. The browser that you use on their website, which is quite a nice way to demonstrate their data isn't actually of any practical use in GP systems.
AM17: There's a lot of this stuff is used to produce what is thrown over the wall which is quite definitely not open source.
DC: I'm not claiming this for open source. I'm mentioning because it is an example of a Crown Copyright application and as far as I am aware, they say on their website that it is crown copyright and they do let suppliers in to examine the source code, and they could do that. One of the messages of my talk, which I think you rather anticipate, is that we need to clarify whether Crown Copyright is the best way to distribute publicly funded software in the UK.
D: For me the licensing is the most important issue. Am I going to contribute Prodigy? No. Not until the issues clarify.
D: LOINC is a set of codes set up to enable the transmission of laboratory data. That is truely free: it's public domain. It's also been subsumed into SNOMED and in turn will be in SNOMED RT. Those things are not free software in any sense of the word. On their website, which talks about "come to see us with your requirements" with the implication that when you want to satisfy them, large cheques will be written.
OIO. Dr Andrew Ho is presenting here in one of the streams today or tomorrow... an attempt to be a middleware layer similar to the efforts going on in the GEHR to enable interoperability of different systems sharing information at the logical level across different platforms and computers. And the tools that they are using to do that are different standard Linux tools for browser-based applications, licensed under the GNU GPL so activity there.
Circare, a high profile project that is using open source tools to network a regional healthcare organisation in Canada, I think Ontario. I think Joe Dal Molin was involved in this project.
Then the GEHR which I hesitate to talk about with so many people who know much more about that me in the room, but a very long standing and worthy attempt to try and produce a data structure for a health care record that is platform independent. The Object model which is under development at the moment in Australia by a developer called Beale, is being written in Eiffel. Eiffel's author, unfortunately, seems to be actively hostile to the idea of open source software as far as I can tell.
Tom Beale I think is here. Yes, I mean the author of Eiffel itself, not the developers of the GEHR, who obviously are not hostile to the idea.
Notice that this system is being licensed under an open source license which is incompatible with the GNU GPL. We keep running
MIQUEST: Another example of a crown copyright application. It's aim is to enable the extraction of health information from different EMRs. For example it could extract the number of people with high blood pressure, and the number of people being treated for high blood pressure from GP systems across different suppliers' systems. That's been shown to work on eight different systems. I'm giving a talk to people who know much better than me what this is.
AM19: It certainly isn't open source: it's very expensive.
AM20: It's a requirement for accreditation of GP systems in this country.
D: LAMBDI: this is a project in America, basically a research project, which is going to try and connect physiological logic--what software people might like to call business logic--at the real hard end. It's going to measure your blood pressure, then fire a syringe driver full of vasopressor agent into the patient. Well. Good luck!
They're using Linux as the underlying basis of that, and they obviously have a fundamental confidence in the underlying stability of the operating system. That's in the early development stages. It's interesting that you have an anaesthetist who thinks that might be a good thing to do.
AM21: Is there any reason to suppose that this is any more dangerous that using Linux in other household appliances which after all have embedded systems that just go on running, why should Linux be any less secure that any other proprietary systems?
D: I'm not, I'm just saying, would you want to that at all? I'm just saying that if I were anaesthetised, I'd still like to have a human being looking at the monitor and looking at the syringe pump.
AM22: Given the mistakes that have made in hospitals over the past few years?
D: Sure, people make mistakes. Yes I have been anaesthetised.
AM22: Can I just ask... you're going through a whole list of very specifically medical applications... this almost concerns me as being all that's wrong with the way that the NHS looks at this. I mean what's wrong with sendmail, what's wrong with StarOffice. Just think how much the NHS would save by switching to Linux plus StarOffice instead of Windows 2000 and Word.
D: I completely take your point. My brief was to give an overview of applications for healthcare.
Chair: I think the standpoint from which we start here... certainly the organising group, is that we accept that the correct solution for email is sendmail, so there's really not much point in discussing it any great depth.
AM22: Well why not?
Chair: Absolutely. Office applications, yes. Open Office, StarOffice have become available and that's I would have thought that a business decision that could be taken by any health service organisation. Were we're really considering specific medical applications. You know as Glyn Moody was saying, first of all people said, "OK, it's produced a very good very stable operating system, but anyone could do that. You know they can't produce major applications."
But then it produced a webserver, it produced sendmail, and then the argument was, "Well yes, these a major special purpose applications, but that doesn't mean you can have the ordinary commercial tools, the desktop stuff."
And now we have StarOffice and so; and it's precisely addressing this next layer of specific domain application software that we are saying "Yes, well what else have the Romans done for us?" You know, can we zip through the specific medical applications, and then I rather think that our four later speakers in the "What can we do in the NHS?"
At least one of them is considering the ordinary business applications that keep the businesses of the NHS going.
DC: I apologise for trying your patience. I think I said at the start that this was the poisoned chalice, and that the overview has been done by a single person: me, and I don't pretend to be globally omniscient.
Medical Record DTD
Another place holder site that produces a medical record document type definition for use in XML systems. The worry is that someone has just said "Right we're doing this." A crucial part of evidence-based healthcare, and perhaps, a crucial part of evidence-based software should be that one should always write an introduction in which one states the search methods which you have done before you start to reinvent the wheel. You can criticise those search methods, that's entirely legitimate, but at least be explicit about where you have looked before you begin. A lot of these projects seem to have little history about them.
AM24. I don't know know about the previous project, but in America there's a consortium of doctors who were starting to get fed up with the system that was provided to them, so they said, "we want this, this and this" and with the help of funds from pharmaceutical companies and from Oracle and a few others, they started saying to a company, this is what we want, and this consortium is called Medbiquitous. We saw them, we came, they are expanding in Europe, and their goal is to have some medical type DTDs. And what is really refreshing is that this is a consortium of MDs. There is now IT involved.
AM25: Is that really relevant? Can we hand the floor back to the speaker?
D: Let me just flash through. One of things that I'm very relaxed about is to be a speaker who realises that the inadequacy of his talk is the most educational aspect of it
D: OpenGALEN. This is a definitely a project to namecheck. Prodigy are using part of their framework as a way of organising pharmacological data. Their motto is "Making the impossible very difficult"
a formal model of clinical terminology trying to translate human language into machine logic. So in way that's another piece of middleware in a sense.
There's a project that was developed several years ago as a way of parsing HL7 data and it seems to have gone to sleep. But that's all there and available under the GPL and presumably if the version of HL7 that you are using means that the current version is insufficient you could hire a programmer who would have a headstart with these tools rather than start from scratch.
Medzope which is an organisation which has been very recently set to sell and develop Zope content management format files. Zope is a webware... I think it's Zope Object Publishing Environment.... runs on Linux, runs on Windows as well, and it's how all those Slashdot style websites are put together. Some are Zope, others are PERL I think, but that's interesting because they are a commercial company who are explicitly setting up to release their product under the GPL. And they have a service-oriented business model.
In bioinformatics there are an extensive collection of tools for capturing and isolating physiological data, using Linux and a free software mathematical tools.
On and on it goes.
Now publishing. Now this is the area that I feel most strongly about. It's interesting to watch how the awareness of open source software has increased dramatically over the last two years, and that itself is feeding back into the publishing industry.
Increasingly questions are being asked. What do scientists get from publishers? The answer is ripped off. They give their papers for free, and then the publishers sell it back to their academic libraries, who have to somehow scrape together the budgets just to hold a proportion of scientific knowledge. Even Harvard can only afford a proportion. And we've got publishers now talking about making some of their content available to the developing world, but in fact if you go into any university library, for example where I studied in Sheffield, you know they will have only only a tiny fraction of the number of journals that you might potentially want to refer to as a medical student, or as a junior doctor, or as a researcher, and I think the open source world has provided a powerful lesson that is driving this forward.
And the public library of science thing... their deadline came and went on the 1st of September, which was to say to publishers that if you don't start to do as we say and make your content freely available on a time expiry model--in which the publisher gets the stuff exclusively for six months and then makes it available to all archive databases--then we're going to set up our own journals and run it in our own way. I'll think we'll increasingly see that happening. We've got eprints.org which is using free software tools to build repositories for scientific papers and the open archives initiatives--I think Steven Harnad the CogSci person down in Southampton is very important in this. They're actually looking at tools for searching archives automatically and harvesting metadata around papers so they generate automatic indexes of archives. And obviously that's a very significant thing--it's the difference between Napster and Gnutella--in that we have to ask, do we have a single awesome resource, or do we have distributed resources? And we will see. But these are very much ongoing projects with hundreds of contributors.
Let's just flash through medical imaging: there are lots of of projects out there. Radiologists are a techie bunch and they like to get their own tools together, and of course a proportion of them are available under free licenses.
So the problem that I have in 2001, as I had in 2000, is that we have isolated projects, with disparate objectives, with only anecdotal evidence about them. But you've already heard this morning how we have significant backing of some very large organisations.
No-one has mentioned yet Mac OS X which is based on BSD Unix and which has a free software kernel called Darwin. This makes it potentially the biggest Unix system distributor. I wouldn't buy OS X yet, but X.1 might be OK.
The EU and many countries in Latin America have made significant investments and the NHSIA are showing quite a lot of interest.
Companies. Just happen to be on the board of this company Medix UK which is using all open source infrastructure to build its site. It harvests the attention of doctors on the way into their site, getting them to fill in pharmaceutical marketing questionnaires and making money from gathering that data.
Pricom are the people who are building this medical extension to the Zope Content Management Framework. They have got their first paying customer in South Humberside Health Authority, and they've got their first open source collaborator in South West Devon, which is a doctor who independently contacted them and said, look we want to do the same thing, can we collobarate? Some people don't want to buy support, and some people do, and they're quite happy and optimistic about the future for their business.
We've got good recognition for open source in health care. The National Electronic Library for health which is a very budget constrained organisation have picked up on the idea of open source big time, and I get the view that they are trying to spread that across into academic publishing and free up the information on which we base our decisions as doctors.
We've got strange individuals who do in fact run open source software in their practices: David Bellamy is a guy who just uses Red Hat as a server and router and so on, Adrian's gone a lot further, but no doubt he'll fill you in on all that.
Andrew Sims expressed an interest in coming to the meeting: indeed perhaps he should have been the one giving this talk. He's perhaps a typical example of the kind of person who is using free software in the NHS at the moment: he's the head of a clinical computing and device evaluation department in Newcastle--the kind of place that provides radioisotopes for thyroid scans, and research prostheses--physics applied to medicine. I sent him an email: "Do you use open source source software?" And back comes: "O yes, the department information system is based on MySQL running on Solaris," which reflects the fact that he has a 10 - 15 year expectation of his hardware, and one of the great attractions for him for open source is that he can just update applications and recompile them as new platforms come along, and get away from the planned obsolescence of proprietary software. And I asked him why he thought Open Source software wasn't more prevalent, and he said "Microsoft has become the new IBM. You can't get fired for buying it." A point that was made at the session also this morning.
OK: this slide illustrates my own attempt to use GNU/Linux software. I think it's still very much in the technical domain. It's hard to use, it's hard to understand. Although if you want to customise it, and customisability is the great attraction of it, that's appealing to techies, but it's nightmare for ordinary users.
The main point that I wanted to make, is that many of these projects are largely individualistic. I think we need to do much more in terms of sales, marketing and PR. Those guys with the suits in IBM who do that are perhaps despised by technical people, but in fact they generate the desire for software, and I think the repository that is coming with the EU SPIRIT project will really help with this.
Let's just skip through this. That's what I'm working towards: it would be nice to have a top to bottom open source GP system; we need to clarify the issue of Crown Copyright: obviously it's better to work with an existing code base rather than start from scratch. A faculty of intranetteurs? I hope that doesn't sound too much like a disgraceful perversion. This is my idea for how we might harness the individualistic nature of current development and perhaps encourage people to share tools and content freely.
I think we've really had questions, haven't we? Shall we carry on with our four panel speakers, and then after tea we have a session for questions from the floor and general discussion. I'm actually pleased to see that people are coming up with sharp points, because it's terrible when you give a long talk and then people just sit there and then you go away.
|Copyright Douglas Carnall 2001. You may reproduce this page in any medium provided this copyright notice is also maintained. Transcription by Douglas Carnall: firstname.lastname@example.org