A . L . Rector Clinical Terminology : Why Is it so Hard ?

For at least the last decade, problems of standardising medical language and terminology have been a major concern of Medical Informatics. Sittig placed achieving a common ‘controlled vocabulary’ at the top of his list of ‘grand challenges’ for medical informatics [1]. Major efforts have been mounted by the United States National Library of Medicine in its UMLS project, the UK National Health Service and its Centre for Coding and Classification, by SNOMED International, and by the GALEN programme of the European Community. Meanwhile, terminology standards have been a major effort for CEN Technical Committee 251 (CEN/TC251) [2] and ISO TC215 [3, 4]. There are major efforts on vocabulary within other standards efforts including HL7 [5], and the DICOM [6], and CorbaMed made one of its first Request For Proposals and standards the specification for a ‘Terminology Server’ [7]. In the US, the Kennedy-Kassebaum [8] amendment will mandate a national vocabulary early in a few years’ time, and the Group of Eight industrial nations efforts on health have also given prominence to problems of terminology [9]. In decision support, investigators such as Musen [10, 11] and Fox [12] have had considerable success in developing re-usable inference and problem solving methodologies (once thought to be the more difficult problem). However, re-using the ‘ontologies’ or terminologies needed for the re-usable problem solving methods has been much less successful and is now a major constraint on further progress. A large fraction of recent major meetings on medical informatics has been devoted to problems in the area of language and terminology. Most recently the Convergent Terms Project [13] and its successor SNOMED-RT have embarked on a multi-million dollar effort to produce a formal medical representation [14]. In the face of such concentrated effort, such a seemingly straightforward problem ought to have been solved. Judging by the scale of the continuing efforts, it has not been, and there remains scepticism as to whether the current efforts will succeed in providing the needed solutions. The difficulty is made more puzzling because special purpose solutions to small scale problems are easy – dozens of small special purpose vocabularies exist, most of which were developed relatively quickly. Given that major decisions are pending and major efforts coming to maturity, it is an appropriate time to step back and look at the reasons that medical terminologies are hard. More importantly, it is important to be sure we can answer the question: ‘How will we know if we have succeeded?’ This paper investigates the requirements and conflicts which make developing such a comprehensive re-usable terminology for patient-centred systems hard. In setting out these issues, this paper aims to broaden and deepen the debate on the structure of termiA. L. Rector Clinical Terminology: Why Is it so Hard?