Rich metadata is required for a comprehensive description of research assets. We developed a metadata schema for clinical, epidemiological and public health research studies based on existing generic and domain-specific metadata vocabularies. It forms the basis for various search and data management services provided by the German National Research Data Infrastructure for Personal Health Data (NFDI4Health). Interoperability remains a challenge, as various health research standards are to be supported in the medium term. At the same time, embedding our infrastructure in national and international resources requires the use of overarching syntactic and semantic standards and vocabularies. In this paper we present a prototypical implementation in CEDAR Workbench. This not only provides a graphical web interface for collaboration and a possibility for form-based data entry for testing purposes. CEDAR also enables the use of standard vocabularies, annotation of concepts with medical terminologies, and a serialization in an RDF-JSON format.
NFDI4Health is an initiative to foster data sharing in the clinical and epidemiological research
community in Germany. To improve findability and reusability of structured health data from clinical
trials, epidemiological studies, disease registries, administrative health databases and public health
surveillance, a metadata schema (N4H MDS) was developed unifying these different types of research
studies particularly in the advent of COVID-19 [
The CEDAR Workbench [
All metadata elements of the N4H MDS version 0.8 (equivalent to version 1.0 after final consensus) were implemented in CEDAR. We divided the implementation into 109 simple data elements that can be reused and 14 more complex structures composed of simple data elements2. This allowed the schema to be mapped completely. As in many technical implementations, certain idiosyncrasies of the software had to be anticipated, such as unusual types of form fields. Although the MDS follows established vocabularies such as Dublin Core, DataCite, or the specifications of international study registries, machine interpretation of such embeddings requires dedicated references, which the MDS currently does not provide. The annotation of data elements with ontologies from BioPortal was therefore only investigated prototypically, but appears very promising. It was not possible to fully map the complex conditional conditions under which certain data elements become mandatory fields, must satisfy certain formats, or should not be filled in at all. However, this would also not be possible with other tools like REDCap without programming effort.
Collaborative development of metadata vocabularies with domain experts has suffered for many
years from limited support by intuitive tools. Mostly, the focus is on content work and the experts are
not willing to learn new software tools in parallel. As a result, Microsoft Excel is still a quasi-standard,
although its limitations in tracking changes, enforcing naming conventions and technical constraints,
and implementation in software or APIs are well known. The use of CEDAR can also only partially
resolve these conflicts. Further work should investigate the usability of the RDF serialization. Desirable
would be a plugin mechanism that would allow syntactic compatibility to the RDF variant of the HL7
FHIR standard, as this is expected to play a major role in health research in the future [
This work supported by the DFG grant no. 442326535 and WI 1605/10-2.