gms | German Medical Science

Background: One most serious economic risk is the successful recruitment when conducting a clinical trial, since recruitment period can lead to substantial delay [1]. The required data is often stored in different subsystems of the hospital information system, what makes the identification of suitable patients a long lasting process [2].

A clinical data warehouse (CDW) can reduce the effort of identifying patients for a clinical trial [3], [4], [5]. The Enterprise Clinical Research Data Warehouse (ECRDW) of the Hannover Medical School (MHH) is supporting such use cases [6].

The aim of the EHR4CR project* was to develop a federated platform to support patient recruitment in multi-center studies across hospitals in Europe [7], [8], [9], [10]. MHH was one of the first (of 24 connected hospitals) early phase adopters (Champion Programme** of the operational platform InSite™***).

Methods: Local data elements were assigned according to a core dataset [11] by mapping to terminology in the InSite platform (ICD, LOINC, etc.). Local laboratory coding system was mapped semi-automatically to LOINC by Custodix using available metadata from the ECRDW. After this the platform was populated with data from ECRDW.

A three-phase validation of the InSite platform followed by four participants from two study centres and several IT experts by i)translating study protocols into the query language used by InSite and using a study protocol selected with a pharmaceutical company; ii)validating the terminology mappings; and iii) manual based testing graphical user interface (GUI) functionalities.

Results: Evaluation showed that certain concepts of the core data sets (procedures, medication) were not available at MHH. The ECRDW provided demographics, visits, diagnoses and lab values. Protocols were translated and queried on InSite. Study protocols lacked medical nomenclatures, so translation could only be carried out by medical experts. Nevertheless a study protocol (NCT01874431****) was translated and executed by Custodix. Major and minor criteria were not provided. 264 potential candidates were suggested. Study centres could not consider the candidates for recruitment because major criteria were not matched. InSite uses ICD10-CM although a protocol required ICD10-GM specific I50.12*****. LOINC-mapping showed 521 (9.29%) of the total of 5,513 laboratory analyses were mapped by Custodix to 158 LOINC concepts corresponding to 112,569,672 laboratory tests performed in total, i.e. 70.77% of the total quantity (159,072,208 12/2006-07/2017). 24 (corresponding to 46 different local analyses) of 521 assigned LOINCs were not assigned correctly [12]. GUI-testing showed that logical exclusions are possible due to incorrect entries. Depending on the complexity of the query, the processing time was criticized by the end users.

Discussion: Using the ECRDW technology, it was possible to solve already known challenges in reusing clinical data [1], [6]. However, assigning correct LOINCs requires local laboratory experts. Evaluation has shown that InSite suits feasibility studies and explorative research with adequate protocols. Missing nomenclatures however unnecessarily complicated the transfer into queries. Furthermore, we believe that queries in local systems lead to reliable results only in dialog with local experts.

In our opinion, the decisive factor for the success of recruitment platforms is to ensure semantic interoperability by harmonizing clinical data with international standards [13].

*http://www.ehr4cr.eu/
**At the MHH the Champion Programme was financially supported by Sanofi S.A. and carried out in close cooperation with Custodix N.V.
***https://www.insiteplatform.com/
****https://clinicaltrials.gov/ct2/show/NCT01874431
*****http://www.icd-code.de/icd/code/I50.-.html

The authors declare that they have no competing interests.

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