JMIR Publications

ABSTRACT

Background:

Secondary investigations on digital health records such as electronic patient data provided by German medical data integration centers pave the way to improved future patient care. However, only little information is captured about the level of integrity, traceability, and quality of the (sensitive) data elements. From a technical point of view, the widely accepted FAIR principles for data stewardship demand enrichment of data with provenance-related metadata. One solution for health data research is the FHIR provenance resource, which enables the interoperable expression of metadata.

Objective:

This study aims to establish provenance traces underpinning data integrity, reliability and hence trust in Electronic Health Records (EHR), thereby increasing the accountability of the medical data integration center. We present the implementation a proof of concept provenance library integrating international standards as a first step.

Methods:

Provenance provides information about the readiness for the reuse of a data element and serves as data governance supplier. We followed a tailor-made roadmap for a provenance framework, investigated the data integration steps along the extraction, transform and load (ETL) phases and according to a maturity model, and we then deduced requirements for a provenance library. Based on this approach, we developed a provenance model with associated metadata, and we implemented a proof of concept provenance class. In addition, we integrated the international W3C provenance standard, mapped the resulting provenance records to the interoperable health care standard FHIR and provided several representation formats. Finally, we measured and evaluated provenance traces measurements.

Results:

This study implements for the first time integrated provenance traces on data element level in a German medical data integration center. We have developed and implemented a practical method that combines the strength of quality- and health standard guided (meta)data management practices. We measured satisfying pipeline execution times and achieved high-levels of accuracy, reliability and accountability of processed clinical routine data. These outcomes should serve as a motivation to develop further tools for evidence-based and reliable electronic health record provision for secondary use.

Conclusions:

The presented engineering strategy is generic and thus can be applied to a wide range of data sets and beyond clinical research. Our results improve the traceability of data elements along the data life cycle. Consequently, the system mitigates risks since data analysis without knowledge of the origin and quality of all data elements is futile. Provenance provides information about the readiness for the reuse of a data element and serves as data governance supplier. The developed provenance class has the potential to significantly reduce the gap between academia and industry in healthcare. Future provenance analysis stands to benefit from the provenance traces.