Article for Linux User

This article forms part of a continuing personal project to document and illuminate developments in open source software for healthcare. This version is planned for publication in the British magazine Linux User whose circulation includes 1500 IT professionals working within healthcare IT. All feedback gratefully received:

Open source software in healthcare

Healthcare is information intensive, but most of the business at the sharp end--the interaction between professional and patient--is still documented on paper. Douglas Carnall asks whether the current business models for software help or a hinder application development in this domain, and gives an overview of currently active projects

The history of healthcare system development in the UK has not always been a happy. High profile cases of duplication and waste make it an area of considerable managerial and political nervousness. And when the managers get nervous, the bureaucracy burgeons. Extensive commissioning processes like PRINCE2, designed to reduce the risk that the systems developed lapse into irrelevance, can be so cumbersome that using them runs the risk that nothing gets started either.

Although there is much talk of "customer need" within the new look NHS Information Authority, the system is remains rigidly hierarchical, narrowly focussed, and based on outdated business models. NHS organisations must buy approved commercial systems, will have little control over their data, and none whatsoever over the source code of the applications they run.

Gradually, though, an appreciation of what open source might offer the health informatics domain is developing. Open source systems are being used in the market, though with less hype than when a hospital buys another big system from one of the big commercial suppliers.

Sheffield GP David Bellamy is typical: he runs the market leading EMIS database system for recording clinical data in his practice. When he needed a server and firewall for his upgraded network 18 months ago, he installed Redhat 5.0 on a Pentium 75 box he was retiring from personal use at home, to act as a server and router.

"It started out as a personal hobby, but I could immediately see the potential for it to be useful to the practice. Reliability is superb - the system has been rebooted about 3 or 4 times in the last year. I rebooted it to install some software for the NHS net, and the time before that my practice manager turned it off by accident. You do have to be a bit of an anorak to get to use the system, but once you are familiar with it, it's OK. I can administer the system from anywhere on the net, including three miles away in the branch surgery."

General practice systems are quite advanced, and the NHS imposes standards on connection to its private NHSNet: a private network maintained by BT which runs a mixture of X.400 and standard TCP/IP applications. Connection to the NHSnet has been a major bone of contention because of professional concerns about confidentiality, widespread doubt about the technical approach adopted, and perhaps most importantly until recently, an unrealistic per message pricing structure which was uncompetitive with standard dial-up or leased lined connections from an ISP. Linux stands up well within this difficult environment though.

Dr Bellamy again: "BT and the local Health Authority know my system very well and have not tried to change a thing. I allocated all the IP numbers to fit with our strategy when we put in the new network in 18 months ago we had support from EMIS (GP computer supplier) and HA engineers when we connected to the NHS net 4 months ago."

With competition between a (reducing) number of major suppliers, the market for GP software his is not a big: there are around 30,000 GPs in about 8,000 practices in the UK. Most GPs will have about 2 support staff, so this is potentially only about 100,000 users. There are few surprises in this rather conservative area: most suppliers provide some sort of proprietary database: the largest single GP system supplier (EMIS) has over 3,000 users running MUMPS-based software (Massachusetts General Hospital Utility Multi-Programming System) (now InterSystems 'M'). Work in Windows APIs is also common: other big players in the GP market, which is generally reckoned to be more advanced than the hospitals include Torex Medical and VAMP Vision, and in Scotland, all general practitioners have been issued with GPASS, a primitive but adequate standard system that runs on Pentium 75 boxes linked by ISDN.

All this is a long way from a vision of the future in which practices enjoy complete control over both their software environment and their data using open source paradigms. In such a small and specialised area as healthcare computing, single or two-developer outfits are common, and vendor lock can come, not just from failure of the supplier to agree a reasonable pricing structure or make required changes to the functionality of the software, but from straightforward business failure, retirement or career change.

This was the situation that faced the Walton NHS Trust five years ago, when the supplier of their HISS (Hospital Information System) CHC went belly up, leaving them facing the need to completely recommission their system, or rescue source code from the ashes. Fortunately, their agreement with the software house had included lodgement of the source code with a thid party, and that software ran on HP-UX. With some tickling (you can read a compelling account of how it was done, with no documentation or support at this ported over to a Linux box: at a cost substantially less that the original asking price of 160,000 for the HP9000/H30 mini computer it ran on originally.

According to the software manager who co-ordinated the rescue, Neil Spencer-Jones, now a consultant for escrow specialists NCC Global such scenarios are not uncommon:

"In niche markets such as specialised healthcare applications software houses regularly fail and leave their customers in the lurch: and not everyone will have third-party escrow written into the contract. The open source model is obviously attractive for this reason, but writing medical systems is such an intensive process that I can't personally see the GPL happening for applications software any time soon. We've had a lot of enquiries from Eastern Europe asking whether we could release the code for our system, but unfortunately the license precluded us from releasing source to third parties."

Third party escrow is one way to reduce the risks of vendor lock, but going straight to open source seems even more likely to be a desirable course of action. While it is certainly true that the most system vendors for healthcare have dealt using traditional business models, there is a surging interest in open source development models from many parts of healthcare, and the new tools that the internet enable is joining them up and helping them to organise.

One of the recognised perquisites of a successful open source project is the establishment or discovery of a code base. According to Spencer-Jones as many as 50% of legacy healthcare systems were written in some form of Unix, and the Posix-compliance of Linux generally means that porting them is reasonably straightforward.

"Unfortunately, you'll find most of the application software's owners don't see the benefits of open source, and still harbour the illusion that there is some value to be extracted from holding the copyright of the software." says Spencer-Jones. "In fact, with small locally developed applications this is very unlikely to be the case, though they may find a lease of life if they were released to a wider audience under less restrictive licenses."

Possible alternatives to the radicalism of the GPL might include Crown Copyright which would enable free usage within the NHS and potential exploitation of the licenses abroad (the model the NHS Centre for Clinical Classification (CCC) has decided to follow). Quite how profitable this will ever be is open to question, and the future of that particular project is far from certain. Its recent merger with the American SNOMED system may provide a critical mass, but languages are most readily adopted when they are free, and the licensing arrangements may prove prohibitive. It is a customary Anglo-Saxon habit to laugh at the Academie Francais as it attempts to legislate for the language--so why will allow our own institutions to do the same for our technical language--and attempt to charge us for the privilege? Current systems mean that the taxpayer is paying twice: once as the system is developed, and a second time as publicly funded organisations buy licenses. This makes little sense when the marginal costs of sharing code is very nearly nil, and the potential benefits of widespread adoption are great in terms of capture of mindshare and a critical mass of developers and users.

Still, although the government has yet to grasp the open source idea with any conviction, a worldwide group of committed and talented developers is accumulating. The best place to find a listing of open source projects currently extant is from the links page of LinuxMedNews, a slashdot-like site run by Ignacio Valdes. Valdes has just returned from a project installing recycled machines running Linux in a mission hospital in Guatemala, which he will administer from his Houston base using ssh. The machines will run the GPL'd FreeMed Software, which is written in the web database guru's favourite, PHP.

The flagship project is the Good Electronic Health Record: an international attempt to develop open standards and an API for the interchange of records between different systems. Originally funded by the EU, the code base is open sourced and is under active development by an Australian team. The lead developer Thomas Beale of Deep Thought Informatics is developing a kernel that will enable exchange of data between GEHR compliant applications. The project makes use of the object-oriented language Eiffel, and will aim to support records compliant with any of the major standards (HL7, CEN 251), and will enable the ready construction of archetypes which can be adapted for local usage (for example, the requirement that NHS GP systems connecting to the NHS net comply with RF4+).

Although not strictly open source, the National Library of Medicine's UMLS (Unified Medical Language System) is publicly available, and importantly, is increasingly widely used by ordinary doctors as they search the Medline database: its MeSH terms are a subset of the UMLS.

Another project that attempts to extract meaning from codes for medical use is the OpenGalen project presided over by Professor Alan Rector of the Department of Computer Science at Manchester University. Another medical semantic mapping system, it is notoriously difficult to use (the Galen slogan is: "Making the impossible very difficult") but the fact that it is now freely available will enable third parties to tackle that difficulty and, more importantly, extend and develop the language in a way that would not be possible in other nomenclature systems that are more rigidly controlled by license holders such as the UMLS or SNOMED/Read.

With some honourable exceptions, most healthcare professionals are still impossibly technophobic. This is partly because the software currently written for them has failed dismally to address real clinical needs, and because the difficulty of addressing those clinical needs is immense. It also reflects the time it takes to learn to use software well, and a reluctance to train in "non-clinical" areas.

The user interface should belong to the user. What doctors really need is the right to assemble and maintain their own custom suite of tools, for use whereever they might happen to be on the network. Doctors will use familiar interfaces, and have access to familar knowledge sources which they have been continually in contact with since their early days at medical school. As the resources relevant for their current job and locality change, they will pick up new resources and drop old ones, so that a gradually revised core of applications remains at their finger tips. Those standard interfaces between knowledge sources, tools of communication, and the databased records of the patients will enable the presentation of new material, and the revision of the old to proceed in real time in the consultation, enabling doctors to be a better guide and teacher for those whom they serve.

Good software forms seamless connections; as George Orwell said of prose, the best is like a window pane: transparent. The obscurity of commercial binaries will always be an obstacle to good quality intercommunication between systems. Good quality communication comes from open standards and systems, and in healthcare, good communication is too important to be remain proprietary. Software developers should remain confident that there will always be work for the future in discovering, providing, adapting applications for organisations, and training people to use them. This, rather than the sharp-suited gouging of Bill Gates wannabees, should become the predominant business model for software in the British NHS. Software engineering will become a profession more like medicine and the law: in which practitioners earn a fair hourly reward for their experience at interpreting, evaluating and applying knowledge from a specialised domain to the benefit of their clients. Currently prevalent models based on copyrights, patents and licenses restrict the sharing and development of medical knowledge, which is certainly counterproductive and arguably unethical. Open source is the future: all we have to do is built it.

Further reading:

Raymond ES. The cathedral and the bazaar. Sebastapol, CA:O'Reilly and Associates,1999 (also available at

DiBona C, Ockman S, Stone M. Open sources: voices from the open source revolution. Sebastapol, CA:O'Reilly and Associates,1999

Key websites: like site with links to most significant open source based projects in the world of health care

Key mailing lists

The open health mailing list is a community of developers of applications in the healthcare domain dedicated to the open source ideals, and using open source tools to do their work. Traffic is moderate--up to 10 posts a day.


archive available at:

Linux in intensive care

If you walk into any intensive care unit in the country, you'll see an array of high tech equipment with digital displays: and nurse writing the output from all of these displays on a large paper chart. VITAL is a new standard that enables all the machines that go beep to talk to each other, and the hospital information system. Paul Woolman is the Glasgow academic who has recieved EU funding to develop the standard, and has chosen Linux as the operating system at the heart of the system. "Linux is very reliable, and obviously in this setting that's vital. The fact that we'll later be able to embed the system within EPROMs as part of the device's hardware also made Linux a sensible choice. We'll be open sourcing our work in this area: our Spanish partner, hardware manufacturer RGB has to leverage the power of open source to be able to compete with the big boys like Siemens and Hewlett-Packard."

Dr Woolman has also runs Linux as his default webserver since 1993.

For further information see:

Medical informatics in the NHS

A middle aged man slumps forward and collapses unconscious in the street. A passerby with a mobile phone calls an ambulance, a doctor rushes to his side in the emergency department of the hospital. Without access to a patient's healthcare record diagnosis can be difficult, yet paper systems can be difficult to access in an emergency. Knowing that the patient was previously fit, or is known to have diabetes, or epilepsy, or a heroin habit, can be crucial in reaching the right diagnosis quickly, and instituting possibly lifesaving treatment.

Such scenarios are commonplace, but devising solutions to such informational problems is far from easy. It is estimated that around one third of doctors and nurses' time is spent accessing, interpreting, and recording information about patients alone, whether on paper or on computer. Each has their problems, but the NHS is now committed to the development of a fully electronic patient record [1]: with the potential to provide instant access to patient records from anywhere within the NHS.

Additionally, the pressure on doctors to remain up to date in their speciality is well-recognised and enormous. The number of relevant articles in even a single speciality such as Urology is beyond the potential of any single doctor to keep up to date: the Medline database compiled by the National Library of Medicine in the USA indexes approximately 2 million new articles in 3,500 journals each year: reading even a fraction of them would mean that the doctor literally had no time to do anything else, least of all see patients. Getting the right knowledge to the right person at the right time is a major challenge. [2]

Ironically, the early hubris of artificial intelligence developers that they could automate the knowledge and decision making powers of a doctor have if anything slowed the development of useful clinical systems, alienating doctors with an attitude that they were trying to replace them rather than support them. The rather more mundane task of supporting day to day practice activity was not the sexiest area for developers, and widespread enthusiasm for computer technology among ordinary clinicians has not really developed until the advent of the internet in the latter half of the last decade. With limited spare time in their busy professional lives, most doctors are still at the bottom of a very long learning curve in the use of information technology.

Most information systems built to support healthcare to date have focussed heavily on the administrative functions such as recording the patient's name and date of birth, and the fact that they have been in hospital before, in order to index paper records. Hospital laboratories have a long tradition of computerisation: many handle thousands of specimens a day using automated analysers--whose final end product is a paper report.

Specialised sub-component systems that help manage conditions in which doctors have to make a large number of routine checks such as diabetes are also common. Traditionally such systems have either been built by individual enthusiasts, or by commercial companies with proprietary systems. Such systems may serve local needs well, but have limited applicability away from its local context, or have prohibitive costs attached to them.

1 Burns F. Information for health. Wetherby: Department of Health, 1998.

2 Gray JAM. Where is the chief knowledge officer? BMJ 1998 317:832

All feedback gratefully received. Mail me at

Version 1.0

Posted on 6 June 2000.