Structured metadata for reproducible data quality assessments

Ensuring data quality is a fundamental challenge for research data infrastructures, as flawed data may compromise results, waste resources, and ultimately undermine trust. While basic checks, such as detecting out-of-range values or missing data, are commonly employed [1], assessing the fitness of data for complex analyses requires more comprehensive strategies. This includes identifying internal inconsistencies across variables, detecting multivariate outliers, and evaluating potential biases introduced by procedural variables such as measurement devices or examiners [2]. Yet, such advanced assessments are rarely performed in a reproducible or transparent manner. The R package dataquieR, developed within the NFDI4Health initiative, DFG- and BMBF-funded projects, and the EU-funded euCanSHare project, addresses this gap by using structured metadata to configure data quality checks through spreadsheets [3], [4]. These metadata files capture expectations and context, such as variable roles, value constraints, or relevant thresholds, which serve as input for the automated generation of detailed quality reports. The process of annotating data quality metadata in specifically structured files that can be made publicly accessible serves to make expectations and requirements for the data transparent and reusable. Using these metadata as main input for automated data quality reports renders individual results as well as complete data quality assessments reproducible. In this way, comprehensive data quality reports can be generated with dataquieR by a single function call, enabling domain experts with and without programming skills to implement reproducible workflows efficiently. Furthermore, the tool provides functionalities to create custom data quality reports, and can be incorporated into existing data management pipelines. dataquieR has already been linked to external standards or tools such as CDISC's ODM and REDCap, with further extensions (e.g., HL7 FHIR) under way. Its metadata-driven logic and infrastructure orientation make it applicable across disciplines, reflecting key principles discussed in recent European policy frameworks, such as transparency, standardization, and reusability [5]. As such, dataquieR contributes to the development of cross-domain research data infrastructures and data quality practices beyond its original field.