AMIA Cautions FDA on Use of EHR Data for Clinical Investigations

In May, the U.S. Food & Drug Administration (FDA) issued draft guidance for industry on the use of electronic health record data in clinical investigations such as controlled clinical trials. In a July 18 written response, the American Medical Informatics Association applauded the approach but cautioned the FDA against assuming that most EHRs are readily configurable for clinical investigations, even among more advanced institutions.

AMIA stated that it fully supports FDA’s willingness to consider EHR data as a potential source for FDA-regulated clinical research, but it noted that data collected and used in EHRs are intended to support the care of individual patients, rather than providing research-quality data.

For instance, blood pressure measurements in research studies usually have a strict protocol for how values should be taken, whereas in EHRs, most blood pressures are recorded in the same field no matter how, when, where or by whom they were recorded. “This lack of standardization will be a tremendous challenge for high-quality, rigorous research. Further, ensuring data integrity and tracking data provenance in clinical settings is incredibly complicated because multiple, authorized individuals contribute to the EHR and the specificity of audit logs varies widely,” AMIA wrote. “If FDA is interested in which data populated the EDC system or eCRF and where they originated along the continuum of care, the answer could prove extremely difficult and burdensome.”

AMIA encourages FDA to suggest sponsors look toward enterprise data warehouses or translational data warehouses, which may have better semantic interoperability and data integrity compared to sources that remain in the EHR default data model.

AMIA called attention to several other issues, including:

ONC’s Health IT Certification Program. The FDA guidance referenced ONC’s Certification program and suggests certified EHR technology gives “FDA confidence during inspections that the EHR data is reliable…” But AMIA notes that ONC’s program is not designed to ensure data “reliability” or integrity. “We see this guidance overstating the ability of certification to deliver clinical investigation-quality data as a byproduct of care delivery. Further, our members note a general lack of consensus on IT standards used by private-sector actors to render clinical data useful for research, which adds to the variability of data quality and potential approaches meant to mitigate data integrity deficiencies

Audit Trails. FDA’s guidance references “Audit Trails” suggesting that there should be “adequate methods to monitor, track, and document all changes made to information in the EHR pertaining to the conduct of the clinical investigation. AMIA said that while it agrees this would be an important capability to preserve reproducibility and detect any tampering, it questions the capacity of current EHRs to deliver on these kinds of audit logs.

Use of Genomic Data. AMIA foresees a time when genomic data are stored in EHRs as unstructured data that may have relevance to FDA. It said FDA should consider this and related use cases and address potential compliance issues, especially as we move into the precision medicine era.