P229: Supporting Quality, Relevance, and Reliability Upon EHR-Sourced Data Accrual
Poster Presenter
Patrick Rodriguez
Policy Analyst
Duke-Margolis Institute For Health Policy (DMI) United States
Objectives
Present operational tools and best practices for regulatory fit-for-purpose electronic health record (EHR)-sourced real-world data and evidence (RWD/E); regarding data quality, relevance, and reliability; during the data accrual phase.
Method
From March to December 2024, we reviewed literature and regulatory use cases involving EHR-sourced data, held a private workshop (September 2024), and engaged RWE policy stakeholders. We identified and collated tools and best practices to support fit-for-purpose data accrual in EHRs.
Results
The research team reviewed academic literature, U.S. regulatory guidance, and RWD/E white papers. From these sources and the private workshop learnings, we identified nine tools that can be useful to support EHR-sourced data relevance, reliability, and quality assurance procedures,
and communicate those aspects to regulators and others. Best practices were also identified, with specific considerations given to operations intended to support relevance, reliability, and quality assurance for both prospectively and retrospectively accrued EHR-source data. Major themes for prospective accrual include clear, pre-defined documentation of data collection, provenance, and missingness; iterative quality assurance and quality checks; cross-team collaboration; and, if used, clear validation of any artificial intelligence and machine learning (AI/ML) algorithms. Retrospective best practices largely emphasize transparency from data vendors and other involved third-parties, and how proprietary restrictions and protections may impact data disclosure to U.S. regulators.
Conclusion
EHR-sourced data concerns a patient’s journey within a single health care setting or across multiple health systems, and holds vast potential to address research questions of regulatory concern regarding medical product safety and efficacy. Our findings can facilitate fit-for-purpose EHR-sourced data accrual and address fundamental quality, relevance, and reliability assurances needed by end-user decision-makers. The tools and best practices outlined herein can be a useful starting point among trial sponsors and data collaborators seeking to support EHR-sourced data relevance, reliability, and quality and communicate those aspects to regulators and others. Our findings additionally support fundamentals that are necessary to have in place to ensure EHR-sourced data is relevant, reliable, and of sufficient quality throughout and during later stages of the data lifecycle.
