VP, Biostatistics Ontada Chestnut Hill, United States
Background: There is growing acceptance of the value of using Real-World Data (RWD) and Real-World Evidence (RWE) to support pre- and post-approval regulatory decisions, as evidenced by the ongoing development of RWE regulatory guidance. These guidance centers around recommendations to ensure RWD and RWE quality.
Objectives: To describe data stewardship and its role in overcoming the quality pitfalls inherent in the generation of RWD and RWE.
Methods: Data stewardship in the context of healthcare refers to the processes and controls through which data from Real-World sources are systematically collected, transformed, de-identified (if applicable), curated, quality-tested and statistically analyzed. The goal is to produce information that can be reliably used in a myriad of medical research, regulatory evaluations and commercial decisions about medical interventions. Data stewardship is measured in two ways: data quality and methodological rigor. Data quality includes data linkage, completeness, reliability and accuracy, traceability, and interoperability. The methodological rigor includes ensuring fit-for-purpose, well-defined protocol with detailed epidemiological and statistical methods for addressing bias, confounding, missing data, internal and external validity, and transparency.
Results: The accuracy of linking patient records across data sources is higher if the data are from the same network or health system or if research permissions are available that allow linking records through patient’s protected health information (eg, social security numbers). An ideal starting point for collecting more complete and reliable data is to use data collection mechanisms specifically designed to capture information in line with conceptual definitions that reflect current medical or scientific thinking. An example is using input workflows that require data fields to be filled in a sequence that is intuitive to trained healthcare providers and administrative staff, with dropdown menus to populate fields wherever possible, to minimize data input burden while ensuring consistency in data entry. These design features, along with close coordination of practicing physicians inputting data at the point of care, are crucial to reducing data missingness and data errors. The quality of RWD is also impacted by changes to data collection methods over time. Close collaboration in the development of RWD collection instruments affords the ability to track granular changes in the interface used by practices. More examples will be presented.
Conclusions: When implementing strong data stewardship, RWD and RWE can play a more important role in driving research and successfully support new and supplemental indications.
