Integration of clinical information across patient records: a comparison of mechanisms used to enforce semantic coherence.

Semantic coherence about clinical information is the bottleneck for true interoperability among applications in health telematics. Healthcare records are in principle made of statements about patient's health and activities performed, organized within attested transactions or messages. Various mechanisms were developed to optimally represent details of statements into a record system, creating de facto three subdivisions: 1) "containers" of clinical information, i.e., section headings, data elements in local records; segments and data fields in messages; 2) their "contents," i.e., coding systems and terminologies; and 3) "transaction context," i.e., circumstances related to document production and message exchange, typically represented in their headers. Details rely on a common semantic background and should therefore, be seen in a continuum; nevertheless, design methodologies and tools for the three subdivisions evolved independently and assignment of details to subdivisions is not predetermined by principles, but depends on implementation issues. Recent developments within the European Committee for Standardization (CEN/TC251/WG II) and in the European Project GALEN-IN-USE provide a new insight on semantics in healthcare. In order to guide harmonization of semantic aspects in the different series of standards--in information models, messages, document markup, terminology systems--we present here a comparison of the various mechanisms they use to enforce semantic coherence on clinical information.