IMTS Expert Perspective
Enhancing Postmarket Surveillance Through Electronic Health Records
Electronic health records can extend medical device surveillance beyond spontaneous reports by providing longitudinal clinical context, broader patient representation, and opportunities for proactive evidence generation.
The direct answer: EHR data can make postmarket surveillance more proactive and clinically informative, but only when the device exposure, patient population, endpoints, follow up, and analytical methods are sufficiently reliable for the intended question.
Premarket testing cannot identify every issue that may emerge after a medical device enters routine clinical use. Postmarket surveillance therefore monitors whether safety, performance, and benefit risk conclusions remain supported across broader populations, longer follow up, changing practice patterns, and real world use conditions.
Traditional surveillance often depends heavily on complaints and spontaneous adverse event reports. Those sources remain important, but underreporting, incomplete information, uncertain denominators, and reporting bias can limit interpretation. EHRs provide another lens because they contain information recorded during routine patient care.
What EHRs can contribute
Routine clinical data can support earlier and more contextual surveillance.
Continuous observation
EHR data are collected during routine care, allowing surveillance to extend beyond voluntary adverse event reports and scheduled study visits.
Broader clinical context
Demographics, diagnoses, procedures, medications, laboratory results, imaging, and clinical notes can provide context for an observed device outcome.
Linked data sources
EHR information can be combined with registries, claims, device identifiers, mortality data, and other sources when governance and linkage methods permit.
Embedded evidence generation
Health systems may support prospective surveillance, pragmatic studies, or registry activities within routine clinical workflows.
EHRs can support passive monitoring of prespecified outcomes and active evidence generation through prospective protocols embedded in clinical practice. They can also reveal patterns that are difficult to see in isolated data sources, such as changes in laboratory values, repeat procedures, readmissions, medication use, or documented symptoms after device exposure.
The strongest use case begins with a defined surveillance question. A large dataset is not inherently useful if it cannot identify the device, measure the relevant outcome, establish follow up, or distinguish the signal from differences in patient risk and clinical practice.
The important limitations
Availability is not the same as fitness for purpose.
| Challenge | Why it matters |
|---|---|
| Device identification | The dataset must reliably identify the specific device, model, version, and exposure. Procedure codes alone may not provide sufficient granularity. |
| Data quality and completeness | Missing values, inconsistent documentation, changes in coding, fragmented care, and loss to follow up can distort the apparent outcome profile. |
| Endpoint validity | A field that is convenient to extract is not automatically a clinically meaningful or validated safety or performance endpoint. |
| Bias and confounding | Treatment selection, center practices, disease severity, and incomplete covariates must be addressed through an appropriate design and analysis plan. |
| Interoperability and governance | Different EHR systems, privacy protections, permissions, security requirements, and data use agreements can limit standardization and access. |
Clinical notes may contain valuable detail but require careful abstraction or validated natural language processing. Care delivered outside the participating health system may be missing. A negative result may mean that an event did not occur, that it occurred elsewhere, or that it was not documented in a searchable form.
Privacy, security, patient consent, data minimization, and institutional governance must be considered from the beginning. These controls are part of evidence quality because they determine what data can be accessed, linked, verified, and reproduced.
A practical development framework
Start with the regulatory question, not with the database.
- 01
Define the decision
Specify the safety signal, performance question, evidence gap, population, device, comparator, endpoint, and action the analysis is expected to inform.
- 02
Assess data fitness
Confirm device identification, variable definitions, data provenance, completeness, follow up, representativeness, and the ability to validate important outcomes.
- 03
Prespecify the design
Define inclusion criteria, index date, exposure, comparator, endpoints, covariates, missing data methods, bias controls, sensitivity analyses, and success criteria before examining results.
- 04
Integrate the findings
Determine whether the results affect the CER, risk management, labeling, PMS plan or report, PMCF, PSUR, CAPA, or another controlled document.
EHR surveillance is most credible when the analysis can trace each conclusion from a defined clinical question through a fit for purpose dataset to a prespecified regulatory action.
The conclusion
EHRs can strengthen surveillance, but they do not eliminate the need for rigorous study design.
Electronic health records offer a valuable opportunity to observe device use and outcomes in routine practice, detect potential signals, and generate evidence with less burden than a stand alone study in selected settings. Their limitations must be made visible rather than treated as minor technical details.
Manufacturers should combine EHR evidence with complaints, vigilance, literature, registries, PMCF, and other relevant sources. The goal is not to replace one surveillance method with another. It is to build a more complete and responsive evidence system.
This article provides general information and does not replace device specific regulatory, privacy, legal, or statistical assessment.