Evaluating data quality for blended data using a data quality framework.

Area of Science:

• Statistics
• Data Science
• Health Research

Background:

• The U.S. Federal Committee on Statistical Methodology (FCSM) released a Data Quality Framework in 2020.
• This framework organizes data quality into 11 dimensions across utility, objectivity, and integrity domains.
• Best practices for implementing this framework, especially for blended data, require further documentation.

Purpose of the Study:

• To evaluate the application of the FCSM Data Quality Framework for assessing blended data.
• To identify challenges, mitigations, and trade-offs in data quality assessments using real-world case studies.

Main Methods:

• Applied the FCSM Data Quality Framework to three health-research case studies involving blended data.
• Conducted data quality assessments for each dimension to identify threats and mitigation strategies.

Main Results:

• Data quality assessments are more complex in practice than initially expected.
• The importance of individual data quality dimensions varies depending on the intended data use.
• Subjectivity in assessments highlights the potential benefit of quantitative tools, though these are use-case dependent.
• Common trade-offs and mitigation strategies exist across different data quality dimensions.

Conclusions:

• Expert guidance and comprehensive documentation are essential for effective FCSM Framework implementation.
• A nuanced approach is needed, recognizing that not all data quality dimensions are equally critical for every application.
• Quantitative assessment tools can aid in explaining results but must be tailored to specific data uses.
• Understanding inter-dimensional trade-offs and common mitigation strategies is key to managing data quality effectively.