Doctors want IT suppliers to adopt the Snomed lexicon of medical codes so that clinicians can iron out any flaws, but suppliers are reluctant to invest until the bugs have been eradicated. Tony Collins reports on the stalemate that could hold back the latest NHS IT roll-out
Anthony Nowlan, a medical doctor and a director of the NHS Information Authority, faces a classic Catch 22 problem.
He wants IT suppliers to adopt Snomed, a complex lexicon for describing the care and treatment of patients, so that thousands of doctors will start to use it and can report any flaws, allowing developers to make the product more robust.
Refined in this way through a national experiment, as if it were an NHS-wide drugs trial, Snomed could become a world standard for clinical coding systems. It would help patients by, for example, allowing doctors to check a mass of patient records and quickly identify what treatments work best with particular conditions.
The problem for Nowlan, and the future success of Snomed, is that some IT suppliers want the product's basic flaws removed before they will make a large financial outlay to adopt it in their systems. At worst, suppliers suspect that the Department of Health will promise to mandate the use of Snomed in the NHS but will quietly drop it if structural defects persist.
"The principal limiting factor [in the take-up of Snomed] is the development of systems," said Nowlan in an exclusive interview with Computer Weekly.
Healthcare specialists believe the problem is a serious one, for Snomed is a critical component in electronic health records which would allow GPs and hospitals to quickly exchange details of a patient's ailments and treatment. At present, paper-based records are often missing, incomplete or in transit.
"The paper-based system of records is a shambles," said IT manager Steve Ponting, who saw the problem caused by incomplete records when his mother was in hospital.
The prime minister Tony Blair has promised that everyone in the UK will have an electronic health record by 2005. But the successful take-up of Snomed is likely to rest in part on whether suppliers and doctors believe the department's claims that, after nearly four years' work (and seven years on its predecessors) key parts of Snomed are robust enough to use in service.
Nowlan is much respected, trusted and liked by his peers. However, the Department of Health has, in general, a poor record on delivering benefits to patients from national IT projects and has lacked openness when projects have fallen behind.
In the 1990s, the department's credibility suffered when its senior officials proclaimed as national successes projects that attracted strong criticism from clinicians, IT suppliers and the House of Commons Public Accounts Committee.
The projects included a clinical coding system, Read Codes version 3, on which taxpayers spent £32m. The Department of Health said in the mid-1990s that it would mandate the use of Read Codes 3. It never happened.
The Department of Health also proclaimed as a national success an IT programme, Hospital Information Support Systems, which was strongly criticised by MPs because it cost £106m but delivered only £3.3m in benefits.
"They [the department's officials] have told us many times in the past that something they have developed or been responsible for has been robust and we have found to our cost it wasn't. Why should we believe what they are telling us now?" said the director of a healthcare supplier.
The credibility of statements by the Department of Health on the success of its national projects may be further dented by its response to an article published by Computer Weekly two weeks ago that revealed the results of an evaluation of Snomed by about 40 reviewers. The "interim status report" said that most of the doctors, nurses and other health specialists who tested Snomed rejected it as unsuitable for use in the NHS.
An analysis of the official response to Computer Weekly's coverage of the report will give little comfort to those who would like in future to be able to trust the department's statements about the success of its national projects. The Department of Health also said in its statement:
"Snomed will not cost £100m to implement."
This gives the impression that Computer Weekly has got its sums wrong. However, a report commissioned by the NHS Centre for Coding and Classification, part of the Department of Health, estimated the cost of implementing Read Codes Version 3 - the predecessor of the Snomed, and arguably a simpler product - to be £100m. Also, the Public Accounts Committee estimated the cost of implementing Read Codes at "many times more" than its initial cost of £32m.
"The Snomed project cost $9m [£5.5m] over three years to combine the two previous terminology versions, Snomed RT and Read codes into a new product, Snomed CT [Clinical Terms]."
This comment by the department failed to mention that the taxpayer had already paid £32m for Read Codes.
"Despite the recognised difficulty of this sort of task, the Snomed CT project was completed on time and to budget."
Some IT specialists question the value of a product being completed on time and to budget in 2001 if it has not been adopted by the NHS by February 2003.
"This gives confidence in the ability of the NHS to successfully deliver complex national projects."
The department is now in its 11th year of developing a medical coding terminology for use in the wider NHS environment. Even if Snomed were an unqualified success, should this give confidence in the ability of the department to have success with far more complex national projects such as electronic booking systems and electronic patient records?
"The formative evaluation [the Snomed report as publicised in Computer Weekly] was carried out using the first evaluation release and subsequently two further releases have been issued with major improvements."
However, the department's officials told Computer Weekly they were keen to assess the full results of the tests on the first version. They felt the tests would highlight any structural defects in Snomed.
"It has never been the intention to launch [Snomed] in April 2003 for use across the whole NHS."
The department published a strategy document in January 2001 which said, "By March 2003 - migration to national standards for e-mail, browsing and office systems completed and all NHS staff with desktop access, and clinical information systems start to use the Snomed Clinical Terms."
"There are over 1.3 million concepts and relationships in Snomed CT and in an endeavour of this scale, for its first release, the number of issues  is remarkably low at this stage of a major project."
This referred to the Snomed report having identified only 836 flaws. But the department's statement failed to mention that the 836 flaws were found by only a tiny number of doctors who had tested Snomed - and most of them had not carried out a thorough evaluation.
Looked at in the round, the department's statement gives the impression that any serious criticism of Snomed is unjustified and that Computer Weekly is making mischief by publicising an interim report on an outdated version of the product.
But Snomed is only one of a series of projects being run centrally or regionally by the Department of Health. Its success, and to some extent the success of other national projects, will rest to some degree on whether the department can convince clinicians and IT suppliers that it is tackling criticisms, not deriding those who criticise.
If suppliers distrust what the department tells them, they will be reluctant to adopt Snomed in any form, which would anger clinicians, for all are agreed that the NHS needs to adopt electronic patient records as soon as possible.
What is Snomed?
The culmination of more than 10 years work in the NHS, Snomed uses clinical codes, a form of electronic shorthand, as a critical component in computer-based records on the care of patients. It allows doctors and nurses to record the problems of patients, how they were treated, and the outcome. The system ensures that data goes into patient records in a uniform way so that it can be retrieved quickly.