What Is Real World Evidence in Healthcare

By Kevin Fernando · Sep 8, 2026 · 11 min read

Real-world evidence is clinical information derived from observational data collected during routine patient care, from electronic health records, insurance claims, registries, and digital devices, rather than from controlled experiments. It answers questions about how treatments work in diverse populations and real-world settings, whereas randomized trials answer questions about whether treatments work under controlled conditions.

Real-world evidence is clinical information derived from observational data collected during routine patient care, from electronic health records, insurance claims, registries, and digital devices, rather than from controlled experiments. It answers questions about how treatments work in diverse populations and real-world settings, whereas randomized trials answer questions about whether treatments work under controlled conditions.

Real-world evidence is observational data collected from actual patient care outside controlled research settings. It comes from electronic health records, insurance claims, disease registries, and digital health devices, sources that capture what happens when treatments are used by real patients in real practice, not in the controlled environment of a clinical trial.

The distinction matters because randomized controlled trials and real-world evidence answer different questions. A trial shows whether a drug works under ideal conditions with carefully selected patients. Real-world evidence shows whether it works, and how, in the messy reality of diverse patients, competing medications, varying adherence, and long-term use. Neither replaces the other. Understanding what each reveals, where each comes from, and what each cannot tell you is essential for interpreting healthcare decisions: regulatory approvals, insurance coverage, and clinical practice guidelines all now draw on real-world evidence as a formal evidence type.

What real-world evidence actually is

Real-world evidence is clinical evidence derived from analysis of real-world data. Real-world data are the records generated during routine healthcare delivery: patient demographics, diagnoses, lab results, medications, procedures, and outcomes. The FDA's formal definition includes data from electronic health records, medical claims, disease registries, and digital health technologies. The European Medicines Agency uses similar language, defining real-world data as observational data stored in repositories such as electronic health records and disease registries.

The key distinction is observational, not experimental. Real-world evidence documents what actually happened to patients; it does not impose a research protocol on them. A clinician prescribes a medication, a patient takes it (or does not), outcomes occur, and the healthcare system records the result. Real-world evidence is the analysis of those records.

How real-world evidence differs from randomized clinical trials

Randomized controlled trials and real-world evidence studies operate under fundamentally different designs. Randomized trials use randomization and control groups with defined inclusion and exclusion criteria. Real-world evidence studies use observational designs without randomization and include the actual populations that received care. This difference is not a flaw in one or the other; it is a structural choice that makes each suited to different questions.

The RCT-DUPLICATE initiative, which compared results between randomized controlled trials and their non-randomized real-world emulations, demonstrated that heterogeneity between the two can be explained by differences in study design and population differences. When researchers attempted to replicate the methods of published randomized trials using real-world data, the results often diverged, not because one was right and the other wrong, but because the trial excluded patients the real-world data included.

Why randomized trials control what real-world evidence does not

Randomized trials achieve internal validity through randomization. By randomly assigning patients to treatment or control, a trial ensures that known and unknown confounding factors are distributed equally between groups. If sicker patients are more likely to receive a treatment, that bias could make the treatment look harmful even if it works. Randomization prevents that. A trial also enforces inclusion and exclusion criteria: it may enroll only patients aged 18-65 with no other major illnesses, ensuring a homogeneous population in which the treatment's effect is clear.

The trade-off is external validity. A trial that enrolls only ideal-case patients cannot tell you how the treatment performs in elderly patients, patients with multiple conditions, patients who forget doses, or patients in under-resourced settings. Real-world evidence includes all of these because it does not select them out. But without randomization, real-world evidence cannot isolate the treatment's effect from the countless other factors that influence outcomes. A patient who receives a medication may also change diet, start exercise, reduce stress, or receive other treatments. Real-world data records all of this together, making it impossible to say definitively which factor caused which outcome.

Advantages of real-world evidence for healthcare decisions

Real-world evidence reveals outcomes that randomized trials cannot measure or measure poorly. Long-term effectiveness is one: a trial may follow patients for two years; real-world data can follow them for a decade. Rare adverse events are another: a trial enrolls thousands; real-world data may include millions of patients, making it possible to detect side effects that occur in one per hundred thousand users. Effectiveness in diverse populations is a third: trials enroll selected patients; real-world data includes elderly patients, patients with comorbidities, patients of different races and ethnicities, and patients in different healthcare systems.

A systematic literature review examining the efficacy-effectiveness gap by comparing randomized controlled trials and observational studies of glucose-lowering drugs found that this gap, the difference between how well a drug works in trials versus in practice, is a real and important phenomenon. Real-world evidence quantifies it. It also measures adherence and persistence: how many patients actually take the medication as prescribed, and for how long. Trials often show perfect adherence because they monitor patients closely; real-world data shows the messy reality.

Where real-world evidence comes from

Real-world evidence is generated from five main sources, each with different coverage, completeness, and limitations.

Electronic Health Records (EHRs) are digital versions of a patient's paper chart containing health information collected and maintained by healthcare providers over time, including diagnoses, lab results, medications, physician notes, and more. EHRs are broader than electronic medical records (EMRs); EMRs are usually limited to a single provider or practice, while EHRs cover patient records across different care settings and providers. EHRs have rich clinical detail, physician notes, lab values, imaging reports, but coverage is limited to patients who use that healthcare system, and data quality depends on how thoroughly clinicians document.

Medical claims data, also known as administrative data, are electronic records that collect information on millions of doctors' appointments, bills, insurance information, and other patient-provider communications. Claims data typically include diagnosis codes (ICD-10), procedure codes (CPT), National Provider Identifier (NPI), and service dates. Claims data cover large populations because they are generated by insurance companies processing millions of transactions. But they contain only what was billed; they lack clinical detail like lab values or medication dosages. The Agency for Healthcare Research and Quality (AHRQ) was developing an all-payer claims database in 2025 to compile medical claims, pharmacy claims, dental claims, and eligibility and provider files from private and public payers.

Patient registries are organized systems that use observational study methods to collect uniform data (clinical and other) to evaluate specified outcomes for a population defined by a particular disease, condition, or exposure, and that serve a predetermined scientific, clinical, or policy purpose. Patient registries have traditionally been researcher-generated by research institutions, academic clinical institutions, or individual research teams. Terms such as clinical registries, clinical data registries, disease registries, and outcomes registries are also used to describe patient registry data collection methods. Registries are often highly specialized, for example, a registry for patients with a rare cancer, and contain detailed clinical information, but they are smaller than claims databases and may be subject to selection bias if participation is voluntary.

Digital health tools such as mobile apps, wearable devices, and software are transforming the healthcare landscape by providing robust data sets from user activity for real-world evidence generation. These devices capture continuous data: heart rate, activity, sleep, blood glucose, weight. A collaboration announced in March 2026 between Verily Life Sciences and Samsung Electronics integrated Samsung's Galaxy Watch 8 with Verily's precision health platform, Pre, to provide an integrated solution for generating evidence and monitoring real-world populations. Digital tools offer real-time, granular data but reach only populations with access to and willingness to use the technology.

How real-world evidence is used in healthcare decisions

Real-world data and real-world evidence are used throughout the lifecycle of medicinal products to provide evidence about their effectiveness and safety. Three domains dominate: regulatory approval, reimbursement, and clinical practice.

Regulatory and approval decisions

The FDA created a framework in 2018 for evaluating the potential use of real-world evidence to help support the approval of a new indication for a drug already approved or to help support or satisfy drug post-approval study requirements. The 21st Century Cures Act of 2016 mandated that FDA establish a program for evaluating real-world data in regulatory decision-making. In September 2025, the FDA established the RWE Accelerate program as a unified approach to guide real-world evidence efforts across the agency.

Among 218 labeling expansions granted from January 2022 to May 2024, real-world evidence was found in FDA documents for 3 approvals and elsewhere for 52 approvals, representing 23.3%, 27.7%, and 23.7% of approvals in 2022, 2023, and 2024 respectively. Real-world evidence was most commonly found in FDA labeling expansion submissions for oncology (43.6%), infection (9.1%), and dermatology (7.3%). Most real-world evidence studies used in labeling expansion were retrospective (65.9%), employed a cohort study design (87.5%), and used electronic health records (EHR) data (75.0%).

Reimbursement and pricing

Insurance companies and government payers use real-world evidence to assess whether a treatment delivers value in actual practice. A drug may show benefit in a trial but fail to deliver that benefit when used by diverse patients in diverse settings. Real-world evidence can demonstrate that the trial benefit translates to real-world benefit, or reveal that it does not. Payers use this information to decide whether to cover a treatment, at what price, and for which patients. Real-world evidence of long-term outcomes, rare side effects, and performance in specific subgroups informs these decisions. If real-world data shows that a drug works well in elderly patients but not in younger ones, a payer may cover it only for the elderly population.

Clinical practice and treatment guidelines

Clinicians and health systems use real-world evidence to inform treatment decisions and protocols. A hospital system may analyze its own EHR data to determine which antibiotic regimen produces the best outcomes for a specific infection in its patient population. Professional societies incorporate real-world evidence into clinical practice guidelines. For example, a guideline might recommend a first-line treatment based on randomized trial data but note that real-world evidence shows lower adherence with that treatment in certain populations, and therefore recommend a second-line option that patients are more likely to take consistently. Real-world evidence also identifies subgroups: a treatment may work well overall but poorly in patients with a specific genetic marker or comorbidity, information that real-world data can reveal.

Limitations and common misreadings of real-world evidence

Real-world evidence has structural limitations that researchers and decision-makers must account for.

Confounding bias occurs when important confounding factors are lacking in large healthcare databases, and the value of these databases depends on completeness and validity of information on confounding factors. Confounding by indication is a specific type: the indication for treatment is itself a risk factor for the outcome being studied. A patient prescribed a medication for high blood pressure is at higher risk of heart attack than the general population, not because the medication causes it, but because high blood pressure causes it. If real-world data does not account for baseline blood pressure, the medication will appear harmful. Methodological biases in observational studies include competing risks, immortal-time bias, and confounding bias, which can affect estimates of treatment effectiveness.

Selection bias occurs when investigations are conducted in settings that may not be representative of the underlying population. In a case study of metastatic cancer, referred patient samples showed relative overestimation of systemic treatment uptake and median overall survival compared to population-based samples, demonstrating selection bias. Selection bias in electronic health records and administrative data does not diminish with increasing sample size and requires specific attention from researchers. A larger dataset does not cure selection bias; it only makes the bias larger.

Data quality varies. EHR data depend on how thoroughly clinicians document. Claims data lack clinical detail. Registries may have missing data or incomplete follow-up. Digital health data depend on device accuracy and user compliance.

The correlation-causation trap is the most common misreading. Real-world evidence can show that two things are associated, patients on treatment A have better outcomes than patients on treatment B, but cannot prove that A caused the better outcomes. Unmeasured confounding, selection bias, or reverse causation (sicker patients receive treatment B) may explain the association. Treating correlation as causation is the error that leads to false confidence in real-world findings.

How real-world evidence fits with other evidence types

Real-world evidence is one piece of an evidence hierarchy. At the top are randomized controlled trials, which establish causation. Below are observational studies and real-world evidence, which establish association. Below those are case reports and expert opinion.

But hierarchy does not mean ranking by importance. A randomized trial showing that a drug works in ideal patients and real-world evidence showing that it does not work in elderly patients are both true and both important. The trial answers one question; the real-world evidence answers another. The two can conflict: the trial may show benefit while real-world data shows harm, or vice versa. When they conflict, the conflict itself is information. It may indicate that the trial population was too narrow, that real-world use differs from trial protocol, or that unmeasured confounding in the real-world data is producing a false result.

The most rigorous approach combines both. Regulatory agencies increasingly require randomized trial data to establish efficacy and real-world evidence to establish effectiveness in diverse populations. Payers use both to make coverage decisions. Clinicians use both to inform practice. Real-world evidence is not a replacement for randomized trials. It is a complement that answers the questions trials cannot.

Do this automatically

Collecting and analyzing real-world evidence manually, extracting data from electronic health records across hundreds of patients, cross-referencing insurance claims, integrating registry information, and synthesizing findings into actionable insights, requires months of labor-intensive work by clinical teams. Software can automate the aggregation, standardization, and analysis of these diverse data sources to generate real-world evidence at scale and speed. Healthcare research runs this kind of analysis end to end and cites every figure back to its original source, across as many subjects as you need, with Mimir attaching a citation to every figure. It will not predict a trial's outcome or advise on anyone's treatment. Those decisions stay with clinicians and regulators, and deliberately so. Every plan includes it, including the free one.

Start your first report and run it on something you already know well.