JMIR Publications

ABSTRACT

Background:

Interoperability is a well-known challenge in medical informatics. Trends in interoperability can be organized in three streams: knowledge, data and process. The relevant standards can be classified into three levels of complexity – serial, organization and meta-organization. Despite many initiatives and organizations, the interoperability challenge has not yet been resolved, and calls for innovative ways of addressing current deficiencies are still frequently heard.

Objective:

In this paper, we describe an innovative strategy for data interoperability, based on three distinct but complementary pillars, and its implementation within the framework of the Swiss Personalized Health Network (SPHN).

Methods:

Our strategy involves an ontology- and model-agnostic approach based on three pillars: (i) an integrative and usability-focused semantic framework that includes a set of concept definitions and terminology bindings to existing controlled vocabularies in accordance with the context of use, (ii) a description-based formalism for model-agnostic data storage and transport, and (iii) a purpose-specific transformation into specific data models depending on the use case.

Results:

The proposed interoperability strategy has been implemented in the context of the SPHN initiative to serve the data-sharing needs of multicenter research projects. The semantic framework was created by a dedicated working group, comprising data and semantics experts from Swiss university hospitals, in collaboration with the SPHN Data Coordination Center (DCC). The framework consists of a list of semantic concepts that capture the information required by each project and are mapped to international terminologies based on purpose of use. The Resource Description Framework (RDF) was used to formally describe the concepts and their relationships to a common schema that is used to generate instances for storing and transporting hospitals’ data. Finally, data transformers based on the SPARQL query language have been implemented to convert data from the RDF representation into various tabular representations (including common relational data models) that can be processed by analytics software for downstream analysis.

Conclusions:

The proposed strategy is not meant to replace existing standards, but to use them in a synergistic, purpose-specific way in order to enable semantic interoperability of health data at the interface between the healthcare, research and regulatory communities. Its wide adoption within the SPHN framework would lay the foundations for the creation of FAIR (findable, accessible, interoperable, reusable) graph-based data endpoints at Swiss university hospitals which could, in the long run, replace legacy data warehousing solutions based on relational data models, thus facilitating multicenter research for precision medicine in Switzerland and abroad.