An institution should be set up to coordinate efforts, runs a documentation centre and ensures consensus and coherency among the stakeholders through education and promotion.

For messages such as the laboratory results or the admission of a patient, the essential operational context is naturally registered by data elements, regarding for example the location, the motivation, the professionals involved.

The interpretation of clinical data depends on the semantic background, i.e. on certain other clinical data, specific for that particular patient and that particular task. Therefore a more complex approach must be developed, taking into account also the semantic background.

Several organizations (e.g. medical societies, hospitals) are developing data sets for general usages or for precise decisions within a specific clinical pathway e.g. in the specialised networks of oncology or diabetes treatment or research. Often the interpretation of a data element strongly depends on the values of the other items in the data set.

A clinical data set is a set of entries stored, shared or presented as a unit within clinical applications, messages and electronic health records. An entry may be in turn a data set.

A clinical data set is conceived for a particular user community within predefined operational contexts, implicitly or explicitly declared by the organizations which are responsible for the data set. Data sets are developed and maintained by those organizations according their own quality procedures. Application designers may refer to registered data sets to facilitate cooperation between healthcare operators.

In ideal circumstances, clinical data sets are stored in a clinical metadata registry, governed by rules provided by standardization bodies. It is a function of a clinical metadata registry to provide definitions on data sets and their components (together with suitable documentation on the intentions and limitations of their usage) and assure stability of the system. However, a metadata registry will not be responsible for any specific content. For, the stored content remains under the control of issuing organizations. By the description of a clinical pathway that expresses the clinical environment in which a document is shared, the evolution of a document can be made explicit across the involved parties.

The clinical pathways express the real workflows in which interactions between health professionals and patients take place. They define the use cases, the triggers, and which clinical data set needs to be shared for each interaction.

More efforts are desirable in the next future, complementary to the deployment of PODs based on IHE-XDS or similar architectures and to the jurisdictional infrastructures based on HL7 v3, to define which clinical data should be shared by different healthcare professionals to perform different parts of a common task, i.e. by professionals that assume the various roles in the care of a particular patient.

Semantic interoperability in a clinical setting could be facilitated by a standardization of multiple complementary components:

• data elements, i.e. the names of fields in databases or messages, or the XML labels in structured documents; the data elements should be registered in a comprehensive data dictionary (metadata dictionary), and mapped to a Reference Information Model, in order to assure their coherence. See recommendation 8.
• value sets, i.e. particular enumerated lists of entries that provide the appropriate values for the content of each (coded) data element in its context;
• clinical data sets, i.e. the predefined set of data elements that are needed by a professional to perform a particular task in a shared clinical pathway;
• clinical pathways, i.e. a description of the tasks performed by different professionals to cooperate on the care of a particular patient.

Common care plans should thus be agreed upon and standardized, either formalized or not into the information system itself, in order to define the reciprocal information needs on a case by case basis. A generalization is possible for the most relevant use cases (i.e. decision nodes in significant clinical pathways).

The current challenge is to standardize the information needs of healthcare professionals about the clinical data required for decision support and for task management.

In this context, the kinds of data elements (i.e. which subset of clinical data are necessary in a particular situation) are more difficult to characterize, and the utility of this subset depends strongly on the clinical situation involved.

Depending on the particular circumstances, health care professionals and patients may need various patterns of usage of clinical information.

Conceivable simple solutions could be found for specific domains in a relatively short time, thus leveraging the experience cumulated with the most predictable settings, with a relevant impact on the quality of care and on the cost containment.