For most of their existence, electronic health records have been treated as administrative infrastructure, a place to document a visit, justify a claim and satisfy a compliance requirement. That framing no longer reflects their potential. The same records that clinicians fill out during a ten-minute visit are quietly becoming one of the richest, most continuously updated sources of evidence about how diseases actually behave and how treatments actually perform once they leave the controlled environment of a clinical trial. The question facing health IT and life sciences leaders alike is no longer whether EHR data can support research, but whether the systems, standards and incentives on both sides are mature enough to use it responsibly.
A regulatory push, not just a technology trend
This shift has been building for nearly a decade and it is now backed by formal regulatory frameworks. The 21st Century Cures Act, passed in 2016, directed the FDA to formally evaluate how real-world evidence could support drug approvals and post-market surveillance, which led to the agency’s 2018 RWE framework outlining how sponsors should submit real-world data in regulatory applications. [1]
Europe has moved in a similar direction. The European Medicines Agency’s DARWIN EU network is now able to access data from around 250 million patients through 40 data partners across 18 European countries, and the agency’s current strategy explicitly treats real-world evidence as a core input rather than a supplementary one. [2] Health authorities are not asking life sciences organizations to choose between randomized trials and real-world data; they are asking them to use both and EHRs sit at the center of that second category.
Where the bridge is still under construction
The gap between health IT and life sciences R&D is not primarily about ambition; it is about infrastructure. Clinical documentation is built for point-of-care decision-making, not for longitudinal research. A single condition can be recorded under different codes, in different fields or as free text buried in a physician’s note, depending on the system, the specialty and even the individual clinician. Research teams need standardized, harmonized and de-identified data at scale, and getting there requires real infrastructure. Federated data networks that map disparate EHR systems onto common data models are one answer. A 2025 descriptive study in JMIR of one such European network, built specifically to harmonize real-world data across dozens of hospitals and countries, describes an infrastructure designed explicitly to address the heterogeneity of coding standards and governance structures that has historically made cross-institutional research so difficult. [3] The underlying principle behind these efforts is the FAIR standard: as a 2023 editorial in Frontiers in Public Health sets out, real-world data has to be findable, accessible, interoperable and reusable before it can generate evidence anyone can trust. [4]
This is where health IT leadership matters as much as clinical or regulatory expertise. Interoperability is usually discussed as a patient-safety or care-coordination issue and it is, but it is equally a research-readiness issue. A hospital system with clean, standardized, well-governed data is not just easier to operate; it is also a viable research partner. Organizations that treat data architecture, terminology mapping and governance as one-time compliance projects rather than ongoing capabilities will find themselves structurally unable to participate in the evidence economy that regulators are now building.
AI is changing what counts as usable EHR data
The most significant recent development is not a new data source but a new ability to use the messy data that already exists. Much of the clinically meaningful information in an EHR, including disease progression, symptom severity and treatment response, lives in unstructured physician notes rather than structured fields. Natural language processing and large language models are increasingly able to extract this information at scale, turning years of narrative documentation into research-ready variables. Industry sentiment suggests this shift is already well underway. A 2025 survey of 150 senior pharma and biotech executives, conducted by studioID and TriNetX, found that more than half had already paired AI with real-world data, and 93 percent believed AI could make that data more accessible and impactful. [5] This matters because it changes the calculus for smaller research programs and mid-sized life sciences organizations that previously lacked the resources to manually curate large observational datasets. The bottleneck is shifting from data extraction to data quality assurance, which means governance, validation, and auditability of AI-derived data now deserve as much attention as the extraction technology itself.
What this means in practice
For health IT and life sciences organizations trying to act on this convergence, a few priorities stand out. First, invest in EHR data standardization before investing in analytics; a sophisticated model built on inconsistent source data will simply produce inconsistent evidence faster. Second, build governance frameworks that make provenance and quality auditable, particularly as AI-assisted extraction becomes more common, because regulators will expect to trace a research conclusion back to a verifiable source. Third, design for interoperability from the start rather than retrofitting it, since the value of any single EHR system multiplies significantly when it can be federated with others under a common data model. And fourth, build genuinely cross-functional teams: clinical informaticists, biostatisticians and regulatory specialists need to be working from the same data model and the same assumptions, not translating between separate silos after the fact.
None of this replaces the randomized controlled trial and it should not try to. What it does is close the loop between how a treatment performs in a controlled study and how it actually behaves across the diversity of real patients, comorbidities and care settings that a trial can never fully capture. The organizations that get ahead here will not be the ones with the most data, but the ones that treat their EHR infrastructure as a research asset from day one, built with the interoperability, governance and data quality that real-world evidence demands. The technical and regulatory groundwork for that shift is already in place. What remains is the discipline to build for it.
Author Bio
