Real-World Evidence Data Quality

Author: Morgane Ballon, Biostatistician
Published on: 04/08/2026

Estimated Reading Time: 4 min

Statistics & Programming Department

Introduction

Real-world evidence (RWE) plays a growing role in regulatory submissions, health technology assessment, and strategic decision-making in pharma and biotech. Yet, the strength of any RWE study fundamentally depends on the quality of the real-world data (RWD) underpinning it. Quality RWD doesn’t just mean clean and low missingness rates. It means the data are fit for the question being asked. That distinction matters enormously in practice, and everything that follows is about how to assess it before you commit.

Jump to section

1Research Questions

2Data Quality Framework

3Data Validation

4Why Excelya

5FAQ

Research Methodology

Always Start With the Research Question

Before evaluating any data source, the research question needs to be fully specified: target population, exposure definition, outcomes, covariates, and time horizon. Only then can you meaningfully ask whether the data can answer it.

The same dataset can be excellent for one purpose and entirely inadequate for another. For example, claims data often works well for estimating incidence or prevalence in epidemiological studies but falls short when granular clinical outcomes are needed; the kind of detail that electronic health records typically capture better.

Reference

The ISPOR SUITABILITY checklist, published in 2024, is a useful resource for assessing the suitability of electronic health records data for use in health technology assessments (HTAs) [1].

RWD

Framework

The Five Pillars of Real-World Data Quality

A pragmatic way to structure data quality assessment is around five core pillars, which broadly align with regulatory expectations and industry’s best practices.

Completeness

Completeness ensures that key variables are sufficiently populated to support robust inference. For example, high completeness is essential to capture rare adverse events in oncology, where even small gaps in event reporting can bias safety profiles.

Accuracy

Accuracy reflects how well recorded data represent true clinical events and is often assessed through validation exercises, such as comparing ICD-coded diagnoses against gold-standard clinical adjudication.

Consistency

Consistency refers to the stability of coding systems, variable definitions, and recording practices over time and across care settings, which is critical when analyzing longitudinal trends or multi-site datasets.

Timeliness

Timeliness measures how quickly data are captured and made available and is particularly important in postmarket surveillance, where delays in data refresh can distort early safety signal detection.

Granularity

Granularity determines whether you can reconstruct treatment pathways, dosing schedules, and exposure-outcome sequences with the precision a study requires.

Editorial Insight

The one thing that cuts across all five is relevance, whether the data actually capture endpoints that matter for the research objective. In immuno-oncology, for instance, the FDA has emphasized the importance of Patient-Reported Outcomes (PROs) [2]. A dataset without these outcomes may be technically complete, but it may be inadequate for the clinical context.

Completeness
Accuracy
Consistency
Timeliness
Granularity

Evaluation framework

Three Critical Validation Areas Before Using Real-World Data

Every real-world dataset should be evaluated through three complementary perspectives before it is considered fit for evidence generation.

Understand Data Origin, Governance, and Provenance

Understanding where data come from and how they are governed is essential for scientific credibility and regulatory confidence. Data provenance influences population coverage, variable definitions, and potential sources of bias. Robust governance frameworks, transparent quality controls, and compliance with data protection regulations such as GDPR should be clearly documented. Many RWE studies rely on linked datasets to enrich clinical, treatment, and outcome information. Assessing linkage methodology, match rates, and documentation is critical. Transparent dictionaries and audit trails are essential to support reproducibility and regulatory review.

Assess Bias and Representativeness

All real-world data inherently contain bias, including selection bias, information bias, and confounding, which must be systematically assessed and transparently acknowledged. Evaluating representativeness relative to the target population is therefore essential. For example, quantifying diagnoses and comparing them with clinical trial data can help characterize divergence and inform appropriate mitigation strategies. The ENCePP guide on Methodological Standards in Pharmacoepidemiology provides methods to address confounding such as disease risk scores and propensity scores [3].

Run a Feasibility Assessment Before Committing

Before finalizing study design, evaluate missingness, benchmark against external data, document limitations, and involve clinicians early to validate endpoints and assumptions.

Why Excelya

Build Reliable Real World Evidence with High Quality Real World Data

Assessing Real World Data (RWD) requires expertise across epidemiology, biostatistics, clinical science, and data engineering to ensure data are fit for the research question, study design, and regulatory purpose.

Excelya’s multidisciplinary team applies a structured, question driven approach to assess Real World Data quality, identify suitable databases, evaluate representativeness, mitigate potential bias, and strengthen regulatory confidence throughout the evidence generation process. From observational studies and HEOR to regulatory submissions, we help transform trusted Real World Data into reliable Real World Evidence (RWE).

Key Takeaway

Quality Real World Data is more than clean data. It is data that is scientifically robust, representative, and fit for purpose.

Excelya's Real World Evidence framework showing multidisciplinary expertise in epidemiology, biostatistics, clinical science, and data engineering to assess Real World Data quality and generate reliable Real World Evidence for regulatory submissions and HEOR.

References

Sources and Supporting Evidence

[1]

Fleurence, R.L., Kent, S., Adamson, B., et al. (2024) Assessing Real-World Data From Electronic Health Records for Health Technology Assessment: The SUITABILITY Checklist: A Good Practices Report of an ISPOR Task Force. Value in Health, 27(6), pp. 692–701. Available at: https://www.valueinhealthjournal.com/article/S1098-3015(24)00069-X/fulltext (Accessed: 24 July 2026).

[2]

U.S. Food and Drug Administration (FDA) (2024) Core Patient-Reported Outcomes in Cancer Clinical Trials: Guidance for Industry. Available at: https://www.fda.gov/regulatory- information/search-fda-guidance-documents/core-patient-reported-outcomes-cancer- clinical-trials (Accessed: 24 July 2026).

[3]

European Network of Centres for Pharmacoepidemiology and Pharmacovigilance (ENCePP) (2023) Guide on Methodological Standards in Pharmacoepidemiology, Revision 11. European Medicines Agency. Available at: https://encepp.europa.eu/encepp- toolkit/methodological-guide_en (Accessed: 24 July 2026).

Frequently Asked Questions

What makes Real-World Data suitable for a Real-World Evidence study?

Not all Real-World Data (RWD) are suitable for every Real-World Evidence (RWE) study. A dataset should be evaluated against the research question, study population, outcomes, exposure definitions, follow-up period, and regulatory objectives. Factors such as completeness, accuracy, consistency, timeliness, representativeness, and data provenance all influence whether the data are fit for purpose.

How is Real-World Data quality assessed?

Assessing Real-World Data quality involves more than checking for missing values. Researchers typically evaluate data completeness, accuracy, consistency, timeliness, and granularity while also reviewing governance, provenance, linkage methodology, representativeness, and potential sources of bias. Together, these factors determine whether a dataset can support reliable evidence generation.

Why is the research question important before selecting a Real-World Data source?

The research question determines which Real-World Data source is appropriate. The same dataset may be suitable for one study but not another. Defining the target population, outcomes, exposures, covariates, and study timeframe before evaluating data quality helps ensure the selected database can answer the intended scientific and regulatory question.

What challenges can affect the reliability of Real-World Evidence?

Several factors can affect the reliability of Real-World Evidence, including incomplete data, inconsistent coding practices, selection bias, information bias, confounding, poor representativeness, and insufficient clinical detail. Evaluating these limitations before study initiation improves study validity and supports more confident regulatory and clinical decision-making.

How does Excelya support Real-World Data quality assessment?

Excelya combines expertise in epidemiology, biostatistics, clinical science, and data engineering to evaluate Real-World Data for research and regulatory use. Its multidisciplinary teams help sponsors assess data quality, identify appropriate databases, evaluate representativeness, mitigate potential bias, and generate reliable Real-World Evidence (RWE) for observational studies, HEOR, and regulatory submissions.