Real-World Data (RWD)

Key facts

FDA framework

Framework for FDA Real-World Evidence Program, December 2018

Statutory basis

FD&C Act Section 505F, added by the 21st Century Cures Act (2016)

Common sources

EHRs, claims and billing data, registries, digital health technologies

What is real-world data?

FDA defines real-world data as data relating to patient health status and/or the delivery of health care routinely collected from a variety of sources. The defining feature is how the data are generated: they come from routine care, billing, monitoring or daily life, rather than from a protocol written to answer a research question.

Real-world data are the raw material, and real-world evidence (RWE) is what analysis of those data produces. A claims file is RWD; a well-designed study using that file to measure a drug's effectiveness in routine practice produces RWE.

Types of real-world data

Common sources include:

• Medical and pharmacy claims: diagnoses, procedures and dispensed drugs coded with ICD-10-CM, CPT, HCPCS and NDC codes, including Medicare and Medicaid claims.
• Electronic health records (EHRs): clinical notes, laboratory results, vital signs and medication orders.
• Product and disease registries: structured follow-up of patients with a condition or treatment.
• Patient-generated data: surveys, apps, and readings from wearables and other digital health technologies.

Why real-world data matter

RWD show how drugs and devices are used, by whom and with what results, across far larger and more varied populations than trials enroll. Commercial teams use claims data to size markets, map treatment patterns and find prescribers; market access teams model budget impact; and clinical operations teams check trial feasibility and patient counts near sites.

FDA uses RWD for postmarketing safety surveillance, including through its Sentinel System, which analyzes claims and electronic health record data.

Limitations of real-world data

RWD were not collected for research. Claims record what was billed, not always what happened clinically; EHR data are fragmented across health systems; and missing data, coding changes and gaps in who is captured can bias results. Linking sources on patient identity raises HIPAA privacy requirements, so research datasets are usually de-identified or accessed under specific agreements. Whether a dataset is fit for purpose depends on the question being asked.