gms | German Medical Science

Background: One goal of the German Biobank Alliance (GBA) is to establish a federated search for biosamples across all participating biobanks [1]. All sites need to fill a local data storage as part of a so-called bridgehead, which communicates with the central search components. The GBA decided to use HL7® FHIR® for the local data repositories and created an implementation guide to guide local implementations and to ensure their compatibility

Methods: Items from the consented search dataset were mapped to adequate FHIR resource attributes. In cases of no fitting FHIR attribute, extensions were created. Where possible, existing conformance resources e.g. from the International Patient Summary were used. Existing biobank data sets like the Minimum Information About BIobank data Sharing (MIABIS) were taken into account and mappings were created [2]. Standard terminologies like LOINC and ICD-10 were preferred to creating custom codes. In a second step, the data model was extended to cover all items from the BBMRI-ERIC biobank directory in order to be compatible with the pre-existing European infrastructure [3].

Results: Eleven profiles, 18 extensions, eight value sets and nine code systems have been created as part of our Implementation guide so far. At the data models' center is the Specimen resource, which represents a biosample and is connected to the Patient resource representing a donor. All medical data, like conditions or smoking status, link to the patient.

The specimen can also, via a custom extension, be connected to a collection, which is a profiled Organization resource. Collections must contain their BBMRI-ID so their representation in the BBMRI-ERIC directory can be easily identified. Collections are also part of a biobank, which is another profiled Organization with a BBMRI-ID. A profiled OrganizationAffiliation resource can be used to express a biobanks' membership within a biobank network.

Profiles and documentation were made available to implementers on Github and the Simplifier platform and have been adopted by GBA member biobanks.

Conclusion: We developed a FHIR®-based data model for biobanks that covers both the GBA federated search and the BBMRI-ERIC directory. The current iteration of GBA's central search tool, the Sample Locator, uses the Clinical Quality Language (CQL) to query local FHIR® stores filled with resources conformant to our Implementation Guide. The addition of further disease-specific modules in cooperation with other projects using FHIR®, e.g. the German Consortium for Translational Cancer Research (DKTK) or the Medical Informatics Initiative are in progress.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.