Metadata-driven creation of data marts from an EAV-modeled clinical research database

Area of Science:

• Biomedical Informatics
• Data Science in Healthcare
• Clinical Data Management

Background:

• Generic clinical data management systems utilize the Entity-Attribute-Value (EAV) model for flexibility.
• EAV models allow recording arbitrary data parameters across numerous studies without schema modification.
• However, EAV data structures are not optimized for direct analytical processing, particularly for data marts.

Purpose of the Study:

• To describe an automated process for extracting and restructuring EAV-modeled clinical data.
• To demonstrate how study metadata can be leveraged to facilitate this data transformation.
• To enable efficient analytical processing of clinical study data stored in EAV formats.

Main Methods:

• Development of an automated data extraction and restructuring process.
• Utilizing study metadata (parameter descriptions and groupings) to guide the transformation.
• Exporting metadata alongside transformed data for enhanced human interpretation.

Main Results:

• Successful automation of the non-trivial process of converting EAV data for analytical use.
• Reduction in errors associated with manual or non-systematic data restructuring.
• Facilitation of human interpretation through accompanying metadata.

Conclusions:

• Automating the restructuring of EAV clinical data using metadata is feasible and efficient.
• This approach overcomes the limitations of EAV models for data mart analysis.
• The method enhances data usability and interpretability for clinical research.