Clinical decision support and knowledge management are increasingly being shown to improve patient care. However, successful deployment depends on integration and interworking with each other and with more general clinical information sytems. It is rare that any one application is viable on its own. The goal of medical ontologies is that several independently developed medical records and several independently developed decision support and knowledge management systems should be able to be developed and work together without exponentially increasing effort devoted to translating the concepts and terms from one to the other and without imposing unacceptable burdens on either authors to select the concepts or users to enter them.
Any re-usable medical ontology will be both large and complicated. (Typical enumerated terminologies for medical applications contain from 20,000 to 250,000 concepts.) To be re-usable it must address the union of all of the potential uses at the finest level of granularity required. Inevitably, an ontology sufficient to address all these uses will be vastly over-complicated for most individual uses. Furthermore, the potential rigorous logical foundations for ontologies (description logics) are rigid and static whereas users require flexibility and dynamic development. The more expressive the logical foundations for such ontologies the less they correspond to users' intuitive views of their domain. GALEN's approach to this problem is a multi-component multi-layer architecture plus an authoring methodology which combines meta-authoring of application specific views by knowledge engineers with user-oriented tools for domain experts. This techniques has cut training time for domain experts from months to days and allowed distributed, loosely coupled development despite unresolvable differences between some centres.