Multi-Centric Quality Analysis of Oncological Data in FHIR

Area of Science:

• Oncology
• Health Informatics
• Data Science

Background:

• Reliable data is essential for impactful data-driven research.
• Complex data systems increase the risk of poor data quality (DQ), compromising research outcomes.
• Even standardized datasets, like the oncological standardized dataset (oBDS), require DQ assessment for research validity.

Purpose of the Study:

• To assess the data quality of the oncological standardized dataset mapped to FHIR resources.
• To evaluate DQ across six German university hospitals.
• To inform the integration of DQ assessment into the Real World Data Platform of the Bavarian Cancer Research Center (BZKF).

Main Methods:

• Automated data quality checks were implemented on FHIR resources.
• Pathling and Great Expectations tools were utilized for DQ evaluation.
• Key dimensions of data quality were assessed across multiple hospital data sources.

Main Results:

• Data quality was found to vary significantly across the six participating university hospitals.
• The assumption of uniform high data quality, even for standardized datasets, was challenged.
• Variability in DQ necessitates tailored approaches for data integration and utilization.

Conclusions:

• Standardized oncological datasets require rigorous DQ assessment prior to use in research.
• Automated DQ checks are effective in identifying variations in data quality across institutions.
• Integrating DQ assessments into real-world data platforms like the BZKF enhances data reliability for cancer research.