One Step Away from Technology but One Step Towards Domain Experts—MDRBridge: A Template-Based ISO 11179-Compliant Metadata Processing Pipeline

Abstract

Summary Background Secondary use of routine medical data relies on a shared understanding of given information. This understanding is achieved through metadata and their interconnections, which can be stored in metadata repositories (MDRs). The necessity of an MDR is well understood, but the local work on metadata is a time-consuming and challenging process for domain experts. Objective To support the identification, collection, and provision of metadata in a predefined structured manner to foster consolidation. A particular focus is placed on user acceptance. Methods We propose a software pipeline MDRBridge as a practical intermediary for metadata capture and processing, based on MDRSheet, an ISO 11179–3 compliant template using popular spreadsheet software. It serves as a practical mediator for metadata acquisition and processing in a broader pipeline. Due to the different origins of the metadata, both manual entry and automatic extractions from application systems are supported. To enable the export of collected metadata into external MDRs, a mapping of ISO 11179 to Clinical Data Interchange Standards Consortium (CDISC) Operational Data Model (ODM) was developed. Results MDRSheet is embedded in the processing pipeline MDRBridge and delivers metadata in the CDISC ODM format for further use in MDRs. This approach is used to interactively unify core datasets, import existing standard datasets, and automatically extract all defined data elements from source systems. The involvement of clinical domain experts improved significantly due to minimal changes within their usual work routine. Conclusion A high degree of acceptance was achieved by adapting the working methods of clinical domain experts. The designed process is capable of transforming all relevant data elements according to the ISO 11179-3 format. MDRSheet is used as an intermediate format to present the information at a glance and to allow editing or supplementing by domain experts.

Understanding the Nature of Metadata: Systematic Review

“…One approach stated that improved tools would enable medical experts for data modeling and a direct quality validation [ 17 ]. The second approach was to divide the work: the domain experts could deliver the knowledge, and metadata professionals would compose metadata in consultation, resulting in excellent and reusable metadata [ 66 ].…”

Oncology on FHIR: A Data Model for Distributed Cancer Research

“…In addition, it supplements the FHIR profiles with extra information about the data elements, such as the cardinality, "must-support" markers and a human-readable description of the content. To take into account the habits of the user and domain experts and to help them understand the data model and the mapping between the specific data content of the FHIR resource, the DKTK data set (in its former implementation) and the ADT specification, the individual data elements and the mapping to FHIR were provided as an "MDRSheet" Excel spreadsheet [19]. In some cases, there were no corresponding data elements in the FHIR specification.…”

A Semi-Automated Term Harmonization Pipeline Applied to Pulmonary Arterial Hypertension Clinical Trials

“…The data from these trials were provided to us by the Cardiorenal Division of the Food and Drug Administration and are not publicly available. From the perspective of term harmonization, these studies were heterogeneous with respect to (1) what variables were collected, (2) how they were organized, (3) units of measurement, (4) use of slang or abbreviations, (5) which terminology standards or versions (if any) were applied in documenting the values for certain variables, (6) whether raw or analyzable datasets were available, and (7) the overall data model used to organize the respective database. More generally, these studies were also heterogeneous in many other ways including the PAH-specific background therapy (i.e., treatment naive vs. double or even triple therapy), time frame of studies, and definition of AE per study.…”

“…Several strategies, tools, and pipelines have been proposed to address data harmonization at different levels, phases of data collection and processing, and that target different data domains. Examples include global harmonization of detailed clinical models for clinical study data standards, 6 a pipeline to facilitate the collection of standardized medical metadata rather than deal with posthoc harmonization, 7 a flexible platform to facilitate the basic, high-level harmonization of individual patient data for meta-analysis relying on well-defined domain-specific data dictionaries, 8 and a variety of other efforts designed primarily to facilitate generalized data harmonization at the global level. 9,10 Data harmonization efforts like these seek to produce an integrated file with unified semantics for all features and feature values.…”