Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Research Data, Analytics & AI · Impact of electronic health records

The Impact of Electronic Health Records on Research: Data Quality, Regulation and Common Data Models

Electronic health records (EHRs) have made routine care a source of secondary data. This is both an opportunity and a limitation: the data were not collected to answer a given research question, so their appropriate use requires detailed knowledge of how they were collected. This article examines the implications for data quality, regulation and analysis.

Pubrica AcademyWritten by a data-analytics reviewer Reviewed bya clinical data manager Published10 October 2026 Last checked10 October 2026 Reading time10 minutes Sources5 peer-reviewed

An electronic health record (EHR) is the digital record maintained by clinicians during patient care, including diagnoses, orders, notes and results, and it is designed for billing and continuity of care rather than research. When the record is reused as a research data source, each property that made it useful for care (free text, local coding practices, fields left blank because they were not clinically relevant) becomes a property for which the analysis must account.

Learning objectives

  • Why EHR data are described as a secondary use of clinical documentation, and what that implies for validity.
  • The data-quality dimensions — completeness, correctness, concordance, plausibility and currency — used to judge whether an EHR dataset is fit for a specific study question.
  • What regulators now expect when EHR-derived “real-world data” are used to support a regulatory decision.
  • Why common data models exist, and what they do and do not resolve.

Common misconceptions

  • “A greater quantity of data outweighs better data quality.” A large EHR extract with systematic missingness or miscoding produces a confident and precise but incorrect answer; sample size does not correct bias.
  • “An EHR field means what its label states.” The same field is populated differently across sites, EHR vendors and even clinicians within a single site; the label is a starting point for verification, not a guarantee of meaning.
  • “Structured data are research-ready, whereas free text is noise.” Much clinically important information (severity, intent, reasoning) is recorded only in notes, and excluding notes discards information that structured fields never captured.

Differences between EHR data and data collected for research

A trial case report form collects exactly the information the protocol requires, in a fixed order, from a trained rater. An EHR record differs fundamentally: it accumulates whatever each clinician, billing administrator and system interface entered, in whatever order a 12-minute visit allowed. The widely cited review of EHR data-quality assessment methods by Weiskopf and Weng frames this directly: EHR data constitute a secondary use of information captured for clinical and administrative purposes, and their quality must be assessed against the specific research question rather than assumed from their origin in a clinical system1. The same chart can be well suited to one question (whether the patient received the drug) and unsuitable for another (the dose and schedule) if the second detail was recorded only in a scanned PDF or a free-text note.

Data-quality dimensions assessed by reviewers

The review by Weiskopf and Weng, based on a systematic survey of how EHR data quality has been measured in the literature, consolidates the checks into five recurring dimensions: completeness (whether the required data are present at all), correctness (whether the data match the truth, usually checked against a reference standard or chart review), concordance (whether different parts of the record, or different sources, agree with each other), plausibility (whether the value is physiologically or logically sensible), and currency (whether the value remains accurate at the time it is used)1. In practice, assessing the credibility of an EHR-based study amounts to establishing which of these five dimensions were checked, and how.

Regulatory expectations for EHR-derived evidence

EHR data have moved from a research convenience to a recognised source of real-world evidence that can support regulatory submissions. The U.S. 21st Century Cures Act (2016), at section 3022, directed the FDA to establish a framework for evaluating real-world evidence, including EHR-derived data, for use in regulatory decisions2. The FDA’s subsequent guidance on real-world data (updated 2024) sets out what it expects a sponsor to demonstrate before relying on an EHR-derived dataset: that the relevance and reliability of the data for the specific regulatory question have been assessed and documented, not merely that the dataset is large3. A 2025 study by Pant and colleagues, assessing how consistently EHR-derived variables are documented and validated across real-world evidence studies submitted in regulatory contexts, found substantial variability in how well studies reported the provenance and validation of their EHR-derived variables, which is precisely the gap that the FDA guidance is intended to close4.

Capabilities and limitations of common data models

Because EHR systems from different vendors and sites store the same clinical concept in different tables, fields and codes, multi-site EHR research typically maps local data into a common data model before analysis. The Observational Medical Outcomes Partnership (OMOP) Common Data Model, maintained by the OHDSI collaborative, is the model most widely used for this purpose: it standardises the structure (the table in which each type of fact is stored) and the vocabulary (the code set used to represent it), so that the same analytic code can run across sites without being rewritten for each one5. A common data model standardises structure and vocabulary, not the underlying clinical accuracy of what was recorded; mapping an incorrect or missing value into OMOP format does not make it correct. The five quality dimensions described above must still be checked on the mapped data; a common data model makes such checking easier to perform at scale but does not remove the need for it.

Key points

Key points

EHR data are a by-product of care rather than a research instrument, so their research value depends on assessing completeness, correctness, concordance, plausibility and currency against the specific research question, and, where the data will support a regulatory claim, on documenting that assessment as current FDA real-world-data guidance expects.

Practical application

Practical application

  1. The secondary-use gap should be identified first. Before analysis, researchers should record the purpose for which each EHR field was captured (billing, order entry, clinical note) and whether that purpose matches the meaning the study requires.
  2. Quality checks should be selected according to the research question rather than a generic audit. A medication-exposure study requires correctness and currency of dosing fields; a comorbidity study may require concordance across diagnosis codes and problem lists. Checks should be chosen according to what the analysis depends on.
  3. Provenance should be documented if the result may support a regulatory claim. The EHR fields, sites, code sets and validation steps used should be recorded in the structured form that FDA real-world-data guidance expects reviewers to see.
  4. A common data model should be treated as a mapping step rather than a validation step. If data are mapped to OMOP or another common data model, this should be stated together with what was standardised, and the quality checks performed on the underlying values should still be reported.

Frequently asked questions

EHR data are one major source of real-world data, alongside claims data, registries and patient-reported data; the 21st Century Cures Act and FDA guidance treat real-world data as the broader category and EHR data as a common instance of it.

No. A larger sample narrows confidence intervals around whatever the data measure; it does not correct systematic missingness, miscoding or concordance problems, which bias the estimate regardless of sample size.

Which of the completeness, correctness, concordance, plausibility and currency dimensions were assessed for the variables on which the conclusions depend, and how, as opposed to merely citing the dataset’s overall size or source.

References

  1. Weiskopf NG, Weng C. Methods and dimensions of electronic health record data quality assessment: enabling reuse for clinical research. J Am Med Inform Assoc. 2013 Jan-Feb;20(1):144-51. read 10 October 2026. View on PubMed Central
  2. U.S. Congress. 21st Century Cures Act, Public Law 114-255, Section 3022 (Real World Evidence). 2016. read 10 October 2026. View on Congress.gov
  3. U.S. Food and Drug Administration. Real-World Data: Assessing Registries to Support Regulatory Decision-Making for Drug and Biological Products. Guidance for Industry. 2024. read 10 October 2026. View FDA guidance
  4. Pant A, et al. Documentation and validation of electronic health record-derived variables in real-world evidence studies: a systematic assessment. BMC Med Inform Decis Mak. 2025. read 10 October 2026. View on BMC Medical Informatics and Decision Making
  5. Observational Health Data Sciences and Informatics (OHDSI). OMOP Common Data Model. read 10 October 2026. View OMOP CDM documentation

Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.

How to cite this article

Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.

Pubrica Academy. (2026, October 10). The Impact of Electronic Health Records on Research: Data Quality, Regulation and Common Data Models. Pubrica. https://pubrica.com/academy/concepts-definitions/impact-of-electronic-health-records/

This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.

EHR secondary-use quality checklist

This checklist should be added to the methods working file before any EHR-derived dataset is analysed.

[ ] State what each key field was originally captured for (billing / order / note)
[ ] Completeness: % missing on each analysis variable, by site if multi-site
[ ] Correctness: validated against a reference standard or chart review? (yes/no, method)
[ ] Concordance: checked across record sources for key variables? (diagnosis codes vs. problem list, etc.)
[ ] Plausibility: implausible-value rule applied and % flagged
[ ] Currency: time between data capture and value's use in analysis
[ ] Common data model used? Which one, which version
[ ] Documented per current FDA real-world-data guidance, if findings may support a regulatory claim

Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.

How Pubrica can help

Pubrica’s data-analytics team designs and documents the quality checks that an EHR-based manuscript or regulatory submission requires (completeness, correctness, concordance, plausibility and currency) for the specific variables on which the conclusions depend, and prepares the methods section that reviewers expect.

Find something else

Search the Pubrica Academy

Access is free and requires no account. Enter a topic, or open the full library to filter by category, sub-topic or resource type.

Documentation of data-quality checks for EHR-derived data

A scoping assessment is free and takes two business days. It provides a written assessment of what the work requires and which specialist would undertake it.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.