A Data Element-Function Conceptual Model for Data Quality Checks

Area of Science:

• Health Informatics
• Data Science
• Natural Language Processing

Background:

• Existing data quality (DQ) checks are fragmented across heterogeneous formats, hindering comparison, categorization, and indexing.
• A standardized approach is needed to manage and understand the vast landscape of DQ checks.

Purpose of the Study:

• To develop and evaluate a data element-function conceptual model for classifying DQ checks.
• To explore the use of natural language processing (NLP) for automated knowledge acquisition from DQ check narratives.

Main Methods:

• A conceptual model defining "data element" and "function" was developed.
• NLP techniques were applied to extract data elements and functions from 172 Observational Health Data Sciences and Informatics (OHDSI) checks and 3,434 Kaiser Permanente Center for Effectiveness and Safety Research (CESR) checks.

Main Results:

• The model successfully classified all analyzed DQ checks.
• 751 unique data elements and 24 unique functions were extracted.
• Frequent data element-function pairings varied between OHDSI and CESR, highlighting differences in DQ check focus.

Conclusions:

• The data element-function model is effective for classifying DQ checks.
• NLP shows promise for scalable knowledge acquisition in this domain.
• Significant heterogeneity exists in DQ checks, reflecting variations in intrinsic data properties and use-case specific "fitness-for-use" requirements.