Most solutions in standardisation are focused on a specific Predefined Operational Domain (POD). Their standards relate to a set of (interoperable) applications and services used by a particular group or community of users identified either by constraints on their location (geographical areas, i.e. countries), on workflows, on health conditions, or on the general healthcare environment involved. In these solutions, standards are developed for healthcare informatics together with some additional implicit or explicit agreements, necessary for the appropriate level of semantic interoperability within a POD. These agreements are usually not scalable and coherent so that it becomes impossible to merge a posteriori the solutions from different PODs.

Problems related to the specifity of standards to PODs are particularly obvious in the field of terminologies and controlled vocabularies and ontologies. Terminology resources are often of uneven quality (to say the least); and many are marked by internal problems and idiosyncracies, which thwart not only the integration of data but also their exploitation for purposes of reasoning. Even a comparatively well-designed ontology such as the Gene Ontology (GO) incorporates several biases. Thus, because its architecture has grown in an ad hoc way, portions of the ontology differ in the granularity of coverage and use of relation types in ways which place limitations on the application of standard techniques for statistically-based reasoning.

The need has long been recognized for some sort of alignment of ontologies and terminologies in order to promote the reusability of the data annotated using their terms. Unfortunately, the dominant approach to the realization of this need, exemplified by the Unified Medical Language System (UMLS) of the National Library of Medicine, is one of retrospective mapping. Independently developed terminologies of varying quality levels are subjected by UMLS lexicographers to a process of cross-terminology mapping based on identification of putative synonymy relations between their terms.

The problem is that there are very few exact synonyms within a complex domain such as biomedicine, so that post hoc mapping has yielded results that fall short of the goal of integrating the bodies of separately coded semantic metadata in such a way as to make the underlying data sets available for common use. To do so in this manner, typically requires forgoing the fine-grained level of semantic specificity required by expert researchers within a specific scientific field.

There is, accordingly, an increasing trend on the part of the NIH, the FDA, and other bodies to promote reusability of data by consolidating standards for the acquisition, formulation, and processing of biomedical data which would work prospectively. This holds, too, for a series of initiatives under the broadly conceived umbrella of the W3C Consortium, above all the National Cancer Institute's caBIG project, which seek to promote reusability of data through a primarily syntactically based prospective regimentation of the ways this data is formulated and stored.

Similar initiatives are needed for a prospective regimentation and standardisation on the side of semantics.