OBJECTIVES:

In 2018, Cure SMA launched the Clinical Data Registry (CDR), a real-world data (RWD) source to improve spinal muscular atrophy (SMA) standards of care. RWD included electronic medical record (EMR)-sourced data linked to an SMA-specific electronic case report form (eCRF). The CDR is comprised of data from ~800 patients with SMA receiving care from 19 US care centers. Documentation patterns, EMR systems, and data transfer method vary across data-contributing sites, which may impact data quality. We sought to explore gaps in CDR processes to understand steps needed to improve data quality and to increase the confidence of future CDR analysis designs to support various use cases.

METHODS:

We created a customized framework that included criteria that an EMR-sourced registry should have in place. Content from existing RWD frameworks, such as AHRQ, the Learning Health Network, and other published resources were incorporated. We assessed if the CDR met each framework criterion. Finally, external RWD experts were consulted to add criterion, prioritize gaps, and provide recommendations for timing/implementation.

RESULTS:

We identified 44 criteria within four categories, which included documentation/supplementary resources, data quality, fit for purpose data, and interoperability. While many criteria were met, we also identified five high priority improvement areas, including 1) establishing systematic quality checks (including missingness/completeness); 2) data documentation (including data management/provenance); 3) meta-data priorities; 4) publishing operational definitions; and 5) harmonization of data types and nomenclature.

CONCLUSIONS:

The creation of any novel dataset with data sourced from multiple sources/pathways will always be met with unavoidable challenges, but an analysis of the system and processes surrounding the dataset can help reduce risks and improve quality and accuracy. This analysis highlights the importance of having the right processes in place to ensure the data is high quality, reliable, and appropriate for the research question of interest to minimize bias and maximize impact.