With the progress of healthcare digitalization and the growing production of data, interoperability has become a major hurdle in all communities and for all usages, is it care, administrative support, research or public health. However, despite a profusion of standards, interoperability remains an unresolved challenge. In this thesis, a multi-dimensional approach for semantic interoperability on various types of data and for multiple use cases is presented. The proposed solution is based firstly on strong semantics by using compositional controlled vocabularies to create a computer-readable interlingua without enforcing a data model, it then restricts the representation complexity to a useful set of concepts encountered in practice and finally exploits the compositional capabilities of SNOMED CT to represent complex narrative data into post-coordinated SNOMED CT sentences. This approach defines a new, semantically interoperable landscape for clinical data that can leverage new opportunities proposed by the growth of personalized medicine.