Use of Metadata-Driven Approaches for Data Harmonization in the Medical Domain: Scoping Review

Area of Science:

• Health Informatics
• Data Science
• Clinical Research

Background:

• Multisite clinical studies increasingly use real-world data (RWD) for real-world evidence (RWE).
• Data heterogeneity across clinics hinders unified analysis, necessitating Extract-Transform-Load (ETL) or Extract-Load-Transform (ELT) for data harmonization.
• Developing sustainable and time-efficient ETL/ELT processes is crucial for research data quality.

Purpose of the Study:

• Investigate generic ETL/ELT process development possibilities.
• Focus on low-complexity approaches utilizing descriptive and structural metadata.
• Explore metadata-driven (MDD) strategies for harmonizing health data.

Main Methods:

• Conducted a systematic literature review following PRISMA guidelines.
• Searched 4 major publication databases (PubMed, IEEE Explore, Web of Science, Biomed Center) from 2012-2022.
• Extracted and analyzed data from 33 included publications, categorizing approaches and tools.

Main Results:

• Included 33 publications categorized into 7 focus groups.
• Ontology-based and rule-based approaches were most frequently used across thematic categories.
• Diverse approaches and tools were identified for specific use cases in data harmonization.

Conclusions:

• Metadata-driven (MDD) approaches show promise for developing generalized ETL/ELT processes across various domains.
• Automating data transformation from Fast Healthcare Interoperability Resources (FHIR) to Observational Medical Outcomes Partnership (OMOP) Common Data Model using MDD is feasible.
• Selecting the optimal MDD approach and integration strategy for ETL/ELT remains a challenge, necessitating further evaluation.