Open source software in the NHS

Growing awareness that the intelligent health record will only be possible using open source methods. Douglas Carnall reports from the OSHCA 2001 meeting.

Published in Linux User magazine, issue 15, October 2001. Download PDF version (99k)

"We are committed to open standards in a patient centred health service," said Nigel Bell, chief executive of the NHS information authority (NHSIA), "and we are aware that this cannot be 'architected' from the top."

When Whitehall mandarins speak, they are careful to nuance their statements. This one, made in any location, might be judged to give at least modest encouragement to those currently excited by the possibilities that open source software development offers to healthcare. That it was made as he opened the second international meeting of the Open Source Healthcare Alliance (OSHCA), suggests that the NHSIA would like to make this a reality.

Bell's first love (and first degree) was the study of ecology, a fortunate synchronicity with conference chair Joe Dal Molin, who also studied the subject. Citing Axelrod's "Evolution of co-operation" as he opened September's two day meeting with academics, suppliers, users, purchasers and developers of open source software in healthcare, Dal Molin pointed out that the pre-requisites of open source were similar to the pre-requisites of co-operation in the natural world. "You have to be able to find each other, and you have to be able to recognise one another." This meeting, attended by academics, suppliers, users, purchasers and developers from 12 countries across the globe made that a reality, as developers forged face to face relationships with names often known previously only by email. Attendance has swelled three-fold since the August 2000 founding meeting of OSHCA in Rome to more than 70 participants.

Speak to any open source developer in health, and you'll find a single obsession dominates all others: the obsession with open standards. The grail for all who work in this domain is the desire to enable intelligent records that can be shared widely to support patient care in healthcare organisations. But the problem, to say the least, is complex. Even specialists within a single area find it hard-going defining their own terms within a formalised structure; spread this out over numerous specialties and disciplines, and across international boundaries, and you have a problem of considerable difficulty indeed.

"What everyone wants is an intelligent health record," says Jeremy Rogers, a Manchester GP and computer scientist who describes himself as the 'benevolent dictator' within the OpenGalen project. "There's just too much data for us to cope without it." But implementing it is proving complex. OpenGalen's goal is to provide intelligent middleware: an ontology that expresses the inter-relationships between medical terms in a formal manner. According to Rogers, an intelligent medical record with have three main components: the terms themselves, the ontology that relates them, and an numerous rules and algorithms that express the practical management of patients once they have been so-described with the formal attributes.

And Rogers message is simple: open source development methods are simply the only imaginable way of achieving the goals that most doctors, stuck with primitive client-server software running on unstable Windows machines, can only dream of.

"A single central effort, where you co-ordinate the whole the process as a managed enterprise is an extremely complex undertaking that is just too risky for commerce to do. It's necessarily a distributed effort. And clinicians are going to want local customisation of decision support rules."

"Closed source gives you control, attribution (important for liability), and is commercially exploitable, but it's inflexible. You don't recognise other possibilities that you're not actually primarily interested in. But someone else might be, so you limit the ability of the thing to become all that it could be. You become isolated from prototyping because if you develop it closed source you can't develop a preferred supplier test bed, because that is supplier favouritism."

Open source development would change that. And he has support from the middle levels of information management within the NHS too. Julian Todd is a IT manager for the Dudley Group of Hospitals NHS Trust, responsible for IT support for the care of 400,000 patients each year, and almost 1000 in-patient beds in the West Midlands of England. He has long argued that the NHS should adopt an object-oriented approach to programming its information systems that enables the development of shared classes and libraries for the construction of local systems. Todd's argument with the status quo is that the NHS has insufficient internal programming expertise to fulfil its needs, and advocates a "technology transfer approach" in which centrally ordered components can be freely used by organisations, either with sufficient in-house expertise to adopt them for local use, or by buying in consultancy as new systems are commissioned. "I take open source as a given within my scheme. I also assume--though I have regular debates with clinicians about this--that it is possible to contemplate some generic process-based models of care."

And he is frustrated at the current progress--and expense--of software standardisation within the NHS. "It is extremely difficult to get suppliers to stick to even basic open standards. This causes me intense difficulty, because I've actually got to get all the stuff talking to each other. If the component base and applications actually had standards embedded into them, then that becomes much easier. The majority of software that we run in the service at the moment is proprietary and there are literally hundreds of often very small software suppliers using a whole range of development techniques. The consequence of that is the cost of complexity is much higher than it needs to be technically."

But grand schemes are easier to dream up than to implement, and Todd is humble about the difficulty of bootstrapping such a proposal. In his view, it can't be driven centrally: the risks are too high for any individual supplier, or even the government. But he thinks there is a winning argument for commercial suppliers, willing to adopt more service-oriented business models, that will help them de-risk the development of complex applications in which the need for safety is also paramount.

Despite the strength of recognition at the top of the NHS information heap (Nigel Bell), the hard-pressed middle (Todd) and from clear-sighted academics toiling away at the bit face (Rogers) it still seems there is some way to go before the NHS embraces the possibility of open source.

One major obstacle has been the practical difficulty of freely licensing software that has already been developed within the NHS. Ray Henry is the head of informatics for the Public Health Laboratory Service in Wales. Charged with the responsibility of providing a mission critical system to assist nurses in their collection of data essential for controlling hospital infection, he turned to review the proprietary market, but couldn't find any shrink-wrapped product to meet his needs. So the laboratory engaged a small development company to produce a custom solution.

Development was successful, and the product, InControl is being rolled out in Welsh hospitals over the last four years. The product is essentially a client which interfaces with local laboratory systems, enables hospital infection controls nurses to add clinical data, and provides management information for the Public Health Laboratory Service, which needs to keep tabs on new disease outbreaks in order to counter new trends in infection.

But Henry's problems started when the success of InControl started to become apparent. "We've got colleagues in England knocking on the door and saying, "Can we have this?"--and we want to give it to them." But it wasn't so simple. Henry's core business is not software development, and he certainly didn't want to be tied into supporting software in other institutions. He first sought an alliance with the original developers, asking them if they would be willing to market and support the product themselves--but they also felt that it was not their core business. And he was concerned at the risk that if he released the software to a third party company one possible fate for the code would be obscurity.

"You wonder about the future strategic development of the company. If we went to the wrong vendor, they might buy up the rights to our product, then put it on the shelf, sell their competing (inferior) product, and we couldn't do anything about it." Besides which, awkward questions might be asked by their funders--the Welsh National Assembly--about why they are selling it back to public organisations, when it's already been purchased and paid for by public money.

Henry engaged software consultant Tim Benson, who is interested in open source development methods to teach his organisation about the approach, and they made the decision to proceed with releasing the software under the GNU General Public License. He presented the idea to his board, who liked it, but insisted that they run it past the lab's business managers first.

At this point lawyers got involved. "They weren't too concerned about free distribution, but they were very concerned that the laboratory might open itself up to legal action if patients or consumers were harmed by use of the software. They were also concerned that we were scrupulous in ensuring that third party tools used in development allowed such distribution. (InControl has been developed for Win32 using Delphi 4). "So now what we are into now is writing our own licence. Solicitors in London are working on it. And the joke is, if they carry on working on it for much longer, it'll cost more than the software costs. And we can only distribute it to Welsh hospitals on the NHSnet."

A cynic might note that lawyers have got every incentive to create as much difficulty as possible when faced with such a request--the more spurious issues that are raised along the way, the greater the legal bill can eventually be.

Undoubtedly there are legal issues that must be resolved. The culture of software developers within healthcare is understandably cautious. The NHS seems not yet to have grasped that the essence of a free software implementation is that it enables individual developers and organisations access to high quality tools--but also the responsibility for ensuring that that they are correctly adapted to the purpose for which they are intended. Once such an evaluation has taken place, the developers add their value by giving their seal of approval to the application, something that it is entirely possible to do without a monopoly on source code. And those who understand the culture of open source development methods will naturally demur from the idea that development can reach a critical mass within a small area like Wales. Even if the application is released on the entire NHSNet it is unlikely to find the critical mass of developers within the NHS alone. Linux taught the world that free software needs to have a global perspective--life is too short to be reinventing the wheel over each national boundary.

Another more serious problem for open source developers is the growing trend, now spreading over from academic institutions into the NHS that requires NHS organisations to maximise intellectual property from the work that they fund. One developer joked that when lawyers use the term "intellectual property" they see only the word 'property' and don't understand 'intellectual.' But it is a serious point: if managers seek to harness intellectual property using outmoded proprietary software models instead of service-oriented exploitation they may find that the fruit they wish to harvest has withered on the vine. It is possible that the NHS has sufficient programming expertise to go it alone under some sort of internal licensing, but others are less sanguine.

Some hope that the NHS will look outwards from its own concerns and harness the global development community. One of the tenets of open source development is the need to release running code, and the NHS has the potential bring a lot to the party--not just InControl, but other flagship projects like GPASS (the Scottish general practice system), PRODIGY (general practice decision support system), as well as the Read codes. All are currently available under Crown copyright, but difficult to obtain, with no possibility of modification and incorporation of patches by the dominant developer. If the NHS Information Authority could bring material that is currently Crown Copyright into the free software community--perhaps by drafting a copyleft "Crown Public License" for the release of software developed using public funds, it could enable a flowering of software development for healthcare across the globe. Historically, closed source has failed to deliver the intelligent applications healthcare professionals need to support their work with patients. As Jeremy Rogers concluded: "It is not that it's free of charge; it's not that you can check to make sure that it really does what it says, it is that you really require people to be stretching and bending it to evolve it to where it can be."

Case study 1: Why pay for Oracle?

(Department of Clinical Computing, Newcastle)

As the head the Clinical Computing and Medical Device Evaluation of the Regional Medical Physics Department at Freeman Hospital in Newcastle upon Tyne, Andrew Sims started to use open source tools for software development because of the practical benefits. In an environment in which the expected service life of applications is 10-15 years, and which uses multiple platforms, it makes sense to use software that will not fall victim to planned obsolescence.

His Medical Physics department uses a variety of platforms (Win32, Unix/Linux/ Mac), he needs solutions which can be deployed across all of them. This is either done directly using Java or by recompiling C++ code across them. The zero purchase or annual maintenance costs go down well in a budget-constrained environment, as does their adherence to international standards.

The departmental clinical information system uses MySQL hosted on Solaris as the database management system, with forms and reports using Java deployed on PCs and Suns throughout the unit. As nuclear medicine image acquisition systems use Unix whilst NHS administration is PC based, the solution allows a single application to be deployed on both.

Says Sims, "MySQL has proved extremely robust, and fast so far; why pay for Oracle?"

A similar system, also based on MySQL supports the vascular ultrasound service. Naturally the department intranet runs Apache web server with PHP, and there are plans to extend this to clinical users to deliver patient images via the Trust network. "This is a very robust solution," says Sims. "We have also developed various small utilities to solve particular problems: HP2XX was integrated into an existing system we had developed to convert HPGL output to printer-specific code when cheap HPGL printers ceased to be readily available."

Case study 2: software fit for heroes

(VistA and the Department of Veterans Affairs)

Perhaps the most impressive free software system available for use in healthcare today is the VISTA collection of tools, made freely available under the US Freedom of Information legislation. Developed by programmers working within the Department of Veterans Affairs (VA), it comprises a comprehensive set of applications for every aspect of the administration of in-patient and out-patient care. In many ways the VA resembles the NHS, providing comprehensive healthcare not only for the servicemen and women of the US Armed Forces, but for their dependents as well. The VA has a budget of more than $20 billion, spent on 173 medical centers, 771 clinics, and 134 nursing homes, and 180,000 staff.

VISTA provides the software for a comprehensive information system to support this activity. Written in MUMPS (the Massachusetts General Hospital Utility Multi-Programming System), a language whose strengths are most apparent when building largescale databases with multiple simultaneous access, the system has been in continuous development since the early '80s, providing a well documented common services framework that permit permits painless migration across different platforms. A GPL'd version of the language (GT.M) is available at SourceForge.

VistA makes extensive use of integrated SMTP-based email for human or computer-to-computer communication.

Volunteers for its development efforts congregate at where full source code is available, and implementation has spread to in healthcare organisations in several countries other than the US including Finland and Germany.

November 2001. No sooner was this article written than the news that Nigel Bell has been replaced as Chief Executive of the NHSIA arrived. There is no evidence to support any anti-OS conspiracy theory, though the simultaneous announcement that the NHS is to invest tens of millions of pounds in corporate licensing for Microsoft Windows XP depressed some.

Copyright Douglas Carnall 2001. You may reproduce this page in any electronic medium provided this copyright notice is also maintained. This article commissioned by LinuxUser Magazine who have sole UK paper reproduction rights, and have made a nice value-added (illustrated, colour, formatted, sub-edited) PDF available.