Why Science? | FDA Electronic Health Records in Clinical Investigations

FDA describes EHR systems as electronic platforms containing individual health records, generally maintained by healthcare providers, organisations or institutions and used to deliver care. Research teams often do not control these systems, so they must understand ownership, access and change processes before relying on the data.

For an ehr is primarily a care platform, imagine two maps laid over each other: the care workflow and the research protocol. FDA describes EHR systems as electronic platforms containing individual health records, generally maintained by healthcare providers, organisations or institutions and used to deliver care. Research teams often do not control these systems, so they must understand ownership, access and change processes before relying on the data. Keep this boundary visible: A clinically useful EHR is not automatically configured or documented for a research purpose.

A strong science note separates observation from interpretation. Identify who owns the EHR, who administers access, which records are authoritative and how system changes are communicated. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Next chapter

FDA identifies the document as final guidance dated July 2018 under docket FDA-2016-D-1224, issued by CDER, CBER and CDRH. It covers EHR data in prospective investigations of drugs, biologics, devices and combination products, including certain foreign studies submitted to FDA.

The promise of the july 2018 guidance is final is useful reuse. The scientific test is whether context survives. FDA identifies the document as final guidance dated July 2018 under docket FDA-2016-D-1224, issued by CDER, CBER and CDRH. It covers EHR data in prospective investigations of drugs, biologics, devices and combination products, including certain foreign studies submitted to FDA. The result stops here: The document has defined exclusions and should not be stretched to every observational or registry use.

A strong science note separates observation from interpretation. Create a source receipt with title, final status, date, docket, issuing centres, scope and exclusions. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

The guidance applies to EHRs and EHR clinical data warehouses that store and integrate EHR data. It says the document does not apply to other forms of health IT such as mobile and telehealth technologies, medical devices, remote monitoring devices, assistive technologies and sensors.

A good interoperability story for ehr scope is narrower than all health technology names people, systems and meanings. The guidance applies to EHRs and EHR clinical data warehouses that store and integrate EHR data. It says the document does not apply to other forms of health IT such as mobile and telehealth technologies, medical devices, remote monitoring devices, assistive technologies and sensors. Preserve this guardrail: A digital health source should not be labelled an EHR merely because it contains health information.

A strong science note separates observation from interpretation. Classify five data sources as EHR, EHR warehouse, EDC, DHT or other health IT and justify each boundary. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

An EHR may contain clinical notes, laboratory results, medication records, images and orders, but availability does not make each element relevant or reliable for a protocol endpoint. The sponsor should define the research concept and assess whether the EHR captures it with suitable timing, completeness and meaning.

For the trial question comes before the available fields, imagine two maps laid over each other: the care workflow and the research protocol. An EHR may contain clinical notes, laboratory results, medication records, images and orders, but availability does not make each element relevant or reliable for a protocol endpoint. The sponsor should define the research concept and assess whether the EHR captures it with suitable timing, completeness and meaning. Keep this boundary visible: A large dataset cannot compensate for a variable that does not represent the intended clinical concept.

A strong science note separates observation from interpretation. For one endpoint, map the concept to the EHR field, workflow, timing, units and likely missingness. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

FDA defines interoperability as the ability of two or more products, technologies or systems to exchange information and use the exchanged information without special effort by the user. Successful transport is only half the task; codes, units, context and semantics must survive the exchange.

The promise of interoperability means exchange plus usable meaning is useful reuse. The scientific test is whether context survives. FDA defines interoperability as the ability of two or more products, technologies or systems to exchange information and use the exchanged information without special effort by the user. Successful transport is only half the task; codes, units, context and semantics must survive the exchange. The result stops here: A file transfer can succeed technically while changing the scientific meaning of the data.

A strong science note separates observation from interpretation. Test one exchanged laboratory value for patient match, code, unit, reference range, time and interpretation. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

When systems cannot exchange data electronically, staff may transcribe EHR information into an eCRF or paper form. This resembles traditional source transcription and can introduce entry error. Quality controls should reflect the importance and complexity of the data rather than assume manual entry is harmless.

A good interoperability story for non-interoperable systems depend on transcription names people, systems and meanings. When systems cannot exchange data electronically, staff may transcribe EHR information into an eCRF or paper form. This resembles traditional source transcription and can introduce entry error. Quality controls should reflect the importance and complexity of the data rather than assume manual entry is harmless. Preserve this guardrail: Source-data verification can detect some errors but cannot justify an unclear or inefficient design.

A strong science note separates observation from interpretation. Measure transcription error in a sample and trace each error to workflow, field design or training. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Electronic transmission can let demographics, vital signs, laboratory results and medication data populate eCRFs. This may reduce duplicate work and transcription error, yet mappings, transformations and acceptance rules need validation and version control. Site staff should know which fields arrived automatically.

For interoperable systems can prepopulate ecrfs, imagine two maps laid over each other: the care workflow and the research protocol. Electronic transmission can let demographics, vital signs, laboratory results and medication data populate eCRFs. This may reduce duplicate work and transcription error, yet mappings, transformations and acceptance rules need validation and version control. Site staff should know which fields arrived automatically. Keep this boundary visible: Prepopulation should not make incorrect or irrelevant data appear investigator-confirmed.

A strong science note separates observation from interpretation. Mark automatically populated fields, validation rules, exceptions and the evidence of investigator review. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

A fully integrated arrangement may use research modules, tabs or custom fields inside the EHR so research data are entered at the point of care. Integration can simplify workflow but raises questions about separating care documentation from research-only data, access roles and retention.

The promise of fully integrated systems bring research into care records is useful reuse. The scientific test is whether context survives. A fully integrated arrangement may use research modules, tabs or custom fields inside the EHR so research data are entered at the point of care. Integration can simplify workflow but raises questions about separating care documentation from research-only data, access roles and retention. The result stops here: Integration should not blur whether a value was collected for care, research or both.

A strong science note separates observation from interpretation. Label every integrated field by purpose, originator, visibility, retention rule and downstream destination. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

FDA encourages existing open data standards when possible while preserving integrity and security. Standards can make codes, units and message structures more consistent across sites and vendors. Implementation profiles and local mappings still need testing because the same standard may be configured differently.

A good interoperability story for open standards help systems understand one another names people, systems and meanings. FDA encourages existing open data standards when possible while preserving integrity and security. Standards can make codes, units and message structures more consistent across sites and vendors. Implementation profiles and local mappings still need testing because the same standard may be configured differently. Preserve this guardrail: Using a named standard does not prove that two implementations exchange equivalent meaning.

A strong science note separates observation from interpretation. Create a conformance test for one code set, required field, unit and error response. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Structured fields can support automated exchange and analysis, while clinical narratives may contain context not captured in codes. Natural-language processing or manual abstraction can extract information from notes, but methods, versions and review rules should be documented and validated for the intended use.

For structured and unstructured data need different controls, imagine two maps laid over each other: the care workflow and the research protocol. Structured fields can support automated exchange and analysis, while clinical narratives may contain context not captured in codes. Natural-language processing or manual abstraction can extract information from notes, but methods, versions and review rules should be documented and validated for the intended use. Keep this boundary visible: An algorithmic extraction is a derived research measure, not the original clinician narrative.

A strong science note separates observation from interpretation. Trace a concept from free text through annotation, algorithm, confidence threshold and final variable. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

FDA encourages use of certified EHR technology where appropriate because certification can provide information about capabilities and standards. Research teams still need to assess the actual system, configuration, data flow and intended trial use, including controls outside the certified functions.

The promise of certified health it can support—but not settle—assessment is useful reuse. The scientific test is whether context survives. FDA encourages use of certified EHR technology where appropriate because certification can provide information about capabilities and standards. Research teams still need to assess the actual system, configuration, data flow and intended trial use, including controls outside the certified functions. The result stops here: Certification does not establish fitness for every protocol, interface or endpoint.

A strong science note separates observation from interpretation. Compare the certified capability list with the trial’s required functions and record every gap. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

EHR values arise from routine-care processes: who orders a test, when documentation occurs, how corrections are made and what incentives shape coding. Research teams should understand these workflows because completeness and timing can differ across sites, departments and patient groups.

A good interoperability story for data quality should reflect clinical workflows names people, systems and meanings. EHR values arise from routine-care processes: who orders a test, when documentation occurs, how corrections are made and what incentives shape coding. Research teams should understand these workflows because completeness and timing can differ across sites, departments and patient groups. Preserve this guardrail: A missing EHR field may mean not measured, measured elsewhere, not documented or inaccessible.

A strong science note separates observation from interpretation. Observe the workflow for one critical datum and list every point where it can be delayed, changed or omitted. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

The record should identify the EHR source, originator when available, date and time, transfer route, mappings and transformations. Audit trails and data-lineage records help distinguish the original clinical entry from research selections, conversions and corrections.

For provenance should survive ehr-to-edc transfer, imagine two maps laid over each other: the care workflow and the research protocol. The record should identify the EHR source, originator when available, date and time, transfer route, mappings and transformations. Audit trails and data-lineage records help distinguish the original clinical entry from research selections, conversions and corrections. Keep this boundary visible: A copied value without provenance cannot be confidently reconciled with the care record.

A strong science note separates observation from interpretation. Build a lineage diagram from care event through EHR field, interface, EDC field and analysis dataset. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

EHRs contain sensitive information beyond what a protocol may need. Consent, authorisation, IRB review, role-based access, minimum necessary extraction, secure transfer and incident plans help protect participants. Research access should not expose unrelated records merely because the system can display them.

The promise of privacy protections must fit dual-purpose data is useful reuse. The scientific test is whether context survives. EHRs contain sensitive information beyond what a protocol may need. Consent, authorisation, IRB review, role-based access, minimum necessary extraction, secure transfer and incident plans help protect participants. Research access should not expose unrelated records merely because the system can display them. The result stops here: De-identification or limited access reduces risk but does not remove every re-identification or misuse possibility.

A strong science note separates observation from interpretation. Create a field-level access plan showing why each requested element is necessary for the research. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

FDA’s acceptance of EHR-derived data depends on the agency’s ability to verify quality and integrity during inspection. Sponsors and sites should arrange access to relevant records, metadata and audit history while respecting privacy and institutional security controls.

A good interoperability story for fda inspection requires verifiable access names people, systems and meanings. FDA’s acceptance of EHR-derived data depends on the agency’s ability to verify quality and integrity during inspection. Sponsors and sites should arrange access to relevant records, metadata and audit history while respecting privacy and institutional security controls. Preserve this guardrail: A summary extract may not be enough when FDA must verify the source and its history.

A strong science note separates observation from interpretation. Run an inspection rehearsal for one transferred value and record the time and evidence needed to verify it. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Clinical investigators and sponsors must meet applicable recordkeeping and retention requirements even when EHR vendors, interfaces or healthcare organisations change. Plans should address readable exports, certified copies, metadata, code dictionaries, mapping versions and continued authorised access.

For retention plans should outlast interfaces, imagine two maps laid over each other: the care workflow and the research protocol. Clinical investigators and sponsors must meet applicable recordkeeping and retention requirements even when EHR vendors, interfaces or healthcare organisations change. Plans should address readable exports, certified copies, metadata, code dictionaries, mapping versions and continued authorised access. Keep this boundary visible: Keeping the EDC copy may be insufficient when interpretation depends on EHR context that was not transferred.

A strong science note separates observation from interpretation. Test a long-term archive package without using the live EHR or current interface. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

The receipt records the protocol question, EHR owner, system version, care workflow, fields, standards, mappings, transfer mode, validation, provenance, review, privacy, inspection and retention. It lets a later reader see exactly how a care record became trial evidence.

The promise of a complete ehr evidence receipt keeps care and research aligned is useful reuse. The scientific test is whether context survives. The receipt records the protocol question, EHR owner, system version, care workflow, fields, standards, mappings, transfer mode, validation, provenance, review, privacy, inspection and retention. It lets a later reader see exactly how a care record became trial evidence. The result stops here: A transparent data route cannot remove bias created by who receives care or what clinicians choose to document.

A strong science note separates observation from interpretation. Ask an independent reviewer to reproduce the variable and list the populations or workflows it may underrepresent. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter