On November 29, 2021, Veradigm submitted comments on the recently issued Food and Drug Administration’s Draft Guidance: Real-World Data: Assessing Electronic Health Records and Medical Claims Data To Support Regulatory Decision Making for Drug and Biological Products.
In our response, we laud the FDA’s commitment to evaluating the use of electronic health record (EHR) and/or medical claims data to support regulatory decisions on product effectiveness and safety. At the same time, we note that it is critical to analyze the quality of a dataset before using it in medical research, to ensure that the research results are reproducible. For this reason, we recommend opening a dialog on ways to improve the completeness and quality of Real-World Data (RWD) for research use.
Alignment on standards is one step that can further improve data quality. Specifically, there is a need for standards on traceability and for maintaining the provenance of data recorded in an EHR. Transparency about data that has escaped or has been excluded from use is important as well, as these can affect confidence in a study’s findings.
Data quality can also be enhanced by creating guidance regarding data sources. For instance, differentiation between EHR and Claims data in the Guidance document is essential, as these data types are not interchangeable but have differing strengths and weaknesses. Guidance related to data sources can also help ensure the use of heterogeneous EHR data, to avoid bias, rather than data that is not sufficiently representative for research. Finally, we recommend guidance on linking separate data sets, which must be performed prior to de-identifying data while still maintaining patient privacy.
The FDA’s draft guidance on the use of EHR and/or medical claims data to support clinical studies and facilitate regulatory decision-making, combined with corresponding innovations from the source providers of RWD to identify key considerations and minimal standards for data curation for these purposes, will introduce a future in which the biopharmaceutical industry uses RWD to support regulatory decision making and other types of research. In this future, innovations will be able to reach those who need them more quickly and efficiently. Veradigm supports this vision and is eager to begin working with industry partners to make it a reality.
The full text of our comments follow. If you have questions, concerns, or comments, please contact Leigh Buchell, Vice President, Policy & Government Affairs, Allscripts at firstname.lastname@example.org.
To Whom It May Concern,
I am pleased to submit our comments to the United States Food and Drug Administration’s Draft Guidance: Real-World Data: Assessing Electronic Health Records and Medical Claims Data To Support Regulatory Decision Making for Drug and Biological Products.
Allscripts, with health IT solutions for ambulatory and acute care settings, is relied upon by a large network of providers — physicians in thousands of different physician practices, as well as 2,400 hospitals, surgery centers, rural care clinics and other care delivery locations. Through our Veradigm business unit, we serve as a large source of de-identified EHR real-world data to life sciences clients (150+m patient records), which may be used in clinical studies to support regulatory decision-making and other research projects. Our unique position as both a worldwide provider of EHR and other health IT software to physicians, as well as a source of de-identified real-world EHR-sourced patient data for research, enables us to provide a unique, informed perspective on the types of research questions that are suitable for EHR data, the development and validation of variable definitions using EHR data, and the provenance, quality and also potential drawbacks of using EHR data for research.
Allscripts supports and encourages the FDA’s commitment to evaluating the use of EHRs and / or medical claims data in clinical studies to support a regulatory decision on effectiveness or safety, in order to help to bring innovations faster and more efficiently to the patients who need them. In addition, we would welcome a dialogue on the topic of how to improve the capture and quality of real-world (EHR) data for research.
The intended audience for this Guidance is the Research Community. Allscripts also recommends a future, companion Guidance document targeted to source providers of real-world data, identifying key considerations and minimum standards for curation and documentation necessary to support regulatory submissions.
We echo the EHR Association’s comments that alignment on a clear set of standards to ease the sharing of real-world data with the research environment would be helpful, with a focus on common vocabulary, structuredness, and relevant granularity in the source systems. We suggest that targeted workshops would be the appropriate path to convene expert stakeholders to collaborate on key topics that enable such alignment, and experts from our Veradigm business would be pleased to participate. (See section VI detailed comments below)
Like the EHR Association, we appreciate the FDA’s acknowledgement that real-world data provides tremendous opportunities to accelerate and enhance drug and biological product submissions, even where we acknowledge that the data comes from sources that may not be as complete and stringent as otherwise required for clinical trials. While clinical trial-quality data is not the expectation, increased focus on these advanced uses of the data can serve to ultimately move the needle on expectations around data quality and completeness upstream in the clinical documentation process. Alignment on standards is one aspect, as mentioned above, that can further enable the quality of data.
Two particular examples that would benefit from alignment are the need for traceability and for maintaining the provenance of data received by an EHR for the delivery of care. We suggest that alignment on data requirements in the context of ONC’s USCDI standard, which includes data provenance as a required data element, provides an opportunity to ensure consistency of approaches. This is relevant from initial data capture through the various systems that said data may flow before it is subsequently used for specific research studies.
We also recommend that guidance be released to address the need for transparency about data that is escaped or excluded from use as part of data cleaning, normalization, or harmonization processes, as this can impact confidence in a study’s findings.
While this Guidance document provides high level examples of differences between Claims and EHR data, Allscripts notes that it often uses EHR and Claims data somewhat interchangeably throughout the rest of the document. EHR and Claims data are not interchangeable. As the FDA has noted, they are captured for different purposes with different strengths and weaknesses that are worth distinguishing in further detail, in order to better delineate appropriate use. In addition, Allscripts notes that EHR data is heterogeneous. Independent physician practices provide a significant percentage of outpatient care (and RWD) in the U.S. Heterogeneity of care within EHR data is an important consideration – data from one or a few health systems may not be representative enough for research.
Again to echo comments submitted by the EHR Association, we note that it is necessary to ensure that evidence is based on relevant real-world data and that it must be complete for the patient cohort at hand. Patient matching and linking challenges must be addressed, especially as patients have most frequently been seen by multiple providers and one cumulative patient record does not exist with a single provider. Depending on whether a study involves identifiable data or de-identified/pseudonymized/tokenized data introduces different considerations.
When identifiable data is used, ostensibly the patient has provided the necessary consent; further, there typically is enough data available on the consenting patient to enable both record linking and further data normalization when data for the same patient is pulled from different data sources. However, when data is de-identified/ pseudonymized/tokenized for the study (which is the more likely scenario for studies subject to the proposed guidance), record completeness will be a challenge, given that multiple data sources need to be aggregated. This underscores the need to be able to link data across data sources, while at the same time introducing the challenges of linking data sources once that data leaves the EHR as de-identified/ pseudonymized/tokenized. If the necessary linking is not achieved before the data leaves the provider’s stewardship and has been de-identified/pseudonymized/tokenized, appropriate consent must be in place. We therefore suggest that appropriate guidance must be provided on how to manage this consent process in the context of a clearly defined privacy framework.
Allscripts is appreciative of the opportunity to provide feedback to FDA regarding this topic and welcomes the opportunity to speak further about any of our feedback or suggestions.
Leigh C. Burchell
Vice President, Policy & Government Affairs Allscripts