Here’s the situation most pharma and MedTech data teams are actually in: they know they need real-world data. They’ve approved the budget. They’ve started the vendor conversations. And then — three months, two NDAs, and one very expensive data access agreement later — they discover the dataset doesn’t cover their indication at the regional granularity they need, or the coding methodology introduces systematic bias they can’t correct for, or the data is nine months stale and useless for competitive intelligence. The strategic window has narrowed. The launch timeline has slipped. And someone in the executive suite is asking where the insights are.
This page exists because that sequence of events is entirely preventable — but only if you understand the RWD landscape clearly before you enter it. Not after you’ve committed to a vendor. Before.
What follows is an honest navigation guide to real-world and synthetic healthcare data: what the different source types actually are, what they can and cannot do, how to evaluate them on the dimensions that genuinely matter, and where the compliance and privacy landscape requires careful navigation — particularly across Continental European markets. We won’t name vendors. We won’t rank providers. We will tell you what you need to know to avoid the mistakes we’ve seen derail programmes across the industry.
If you’re a Head of Data Science trying to brief your leadership on RWD sourcing options, an RWE Director scoping a Continental European evidence programme, or a HEOR lead building the analytical foundation for a payer submission — this is the reference you’ve been looking for. Bookmark it. You’ll come back to it.
What RWD Medical Data Actually Is — and Why the Taxonomy Matters
Real-world data is a broad term, and that breadth is part of the problem. Vendors use it to describe sources as different as statutory health insurance claims, electronic health records, patient registries, wearable device outputs, and social determinants of health datasets. Understanding the differences between these sources — what each captures, what each systematically misses, and what each is genuinely suited for — is the foundation of any credible real-world evidence strategy.
According to Sherman et al. (2016), real-world data encompasses health-related data collected outside the context of conventional randomised controlled trials, from a variety of sources including electronic health records, claims and billing activities, product and disease registries, and patient-generated data. That foundational definition from NEJM is still the right starting point — but the landscape has expanded considerably since then.
Here’s how the major source types actually differ in practice:
Health Insurance Claims Data and Aggregated Patient Data
Claims data — drawn from statutory health insurance systems across Continental Europe and from payer databases in other markets — is the backbone of most real-world evidence programmes. It captures what was billed: diagnoses (ICD-10 coded), procedures, prescription dispensing (ATC classification), specialist referrals, hospitalisations, and healthcare utilisation patterns over time. The population coverage is typically very broad, the longitudinal depth is strong, and the data is structured in a way that supports systematic analysis.
What it doesn’t capture is clinical detail. The severity of a diagnosis, the rationale for a prescribing decision, the patient’s functional status, the clinical outcome at a granular level — these are not in claims data. This isn’t a deficiency so much as a design characteristic. Claims data was built for billing, not research. Working with it effectively means understanding what questions it can credibly answer and what questions require a different data source. Our dedicated post on unlocking pharma insights with real-world and synthetic data goes deeper on this. For teams working with claims-based patient-level data, rigorous phenotype validation in claims data is the methodological step that determines whether your disease cohort accurately represents the population you intend to study.
Electronic Health Records and Linked EHR-Claims Datasets
EHR data is clinically richer than claims — it can capture lab values, physician notes, vital signs, and treatment response in ways that claims data simply cannot. The limitation is that EHR data is typically fragmented across care settings, provider systems, and national infrastructure frameworks. A patient’s oncology records, their primary care history, and their pharmacy dispensing data often sit in separate systems with no automatic linkage.
Linked EHR-claims datasets attempt to solve this — and where they work well, they’re powerful for outcomes research, HEOR analysis, and regulatory submissions. But linkage quality varies dramatically, and the methodology used to match records across sources matters enormously for the validity of any analysis built on top of it. Understanding this is covered in detail in our post on analysis-ready real-world evidence and healthcare data.
Synthetic Patient Cohorts and Privacy-Preserving Datasets
Synthetic healthcare data is generated algorithmically from real-world source data — most commonly using Generative Adversarial Networks (GANs) with differential privacy guarantees — to produce datasets that preserve the statistical properties of the underlying population without containing any actual patient records. We cover this in much more depth in the section below, but the short version is: synthetic data is not made-up data. When generated correctly and validated rigorously, it mirrors real patient populations with high fidelity. Its primary advantages are in GDPR Article 9 compliance, rare disease research where patient volume is limited, and scenario modelling where you need to explore analytical pathways before committing to real-data access agreements. As referenced by Dang (2023) and the FDA’s Real-World Evidence Framework, the use of such privacy-preserving approaches is growing rapidly.
Disease Registries and Specialty Pharmacy Data
Registries provide deep clinical detail on specific patient populations — often with longitudinal follow-up, patient-reported outcomes, and treatment history that claims data can’t match. Their limitation is coverage: a registry focused on a rare autoimmune disease might have three hundred patients across five centres. That’s deeply informative for natural history research and trial design, but insufficient for population-level epidemiology or competitive intelligence analysis.
Specialty pharmacy data captures dispensing events for high-cost, complex therapies — often the exact assets that pharma commercial teams most need to track. It’s a valuable complement to claims data for brands where the dispensing channel is narrow and identifiable, though coverage and access models vary considerably by market.
DRG and Hospital Discharge Data
Diagnosis-Related Group (DRG) data and hospital discharge records provide insight into inpatient care patterns, resource utilisation, length of stay, and readmission rates. For MedTech market access programmes, hospital economics research, and post-market surveillance of devices and procedures, this source type is often underutilised. It’s structurally different from claims data — more episodic, less longitudinal — but powerful for specific analytical questions.
Patient-Reported Outcomes and Behavioural Health Signal Data
PRO data — collected through digital health platforms, disease management programmes, and clinical systems — captures the patient experience in ways that administrative data simply cannot. Quality of life, symptom burden, treatment adherence, and patient-reported functional status are increasingly important inputs to HEOR models and regulatory submissions. The challenge is standardisation: PRO data is collected through dozens of different instruments with varying levels of regulatory validation, and integrating it meaningfully with administrative data sources requires careful methodological work.
The Difference Between RWD and RWE — Why It Matters More Than You Think
Let me be direct about something that creates genuine confusion in procurement conversations and internal strategy discussions: real-world data and real-world evidence are not the same thing, and treating them as interchangeable is one of the more reliable ways to invest in the wrong thing.
Real-world data is the raw input — the claims records, EHR entries, registry entries, and dispensing records that exist in the world because patients received healthcare. Real-world evidence is the analytical output — the insights, conclusions, and evidentiary claims that emerge from applying appropriate study designs and analytical methods to that data.
The distinction matters because data access does not equal evidence generation. Licensing a dataset from a claims data provider does not give you real-world evidence. It gives you the raw material from which evidence can be generated, provided you apply the right study design, control for the right confounders, address the missing data patterns, and document your methodology to the standard required by EMA, HTA bodies, or payers.
As Blonde et al. (2018) and Zhu et al. (2023) both emphasise, the translation from raw observational data to credible real-world evidence requires methodological rigour that is separate from and additional to data access. This is where many programmes underinvest — and where the gap between “we have the data” and “we have the evidence” becomes painfully visible during an HTA submission or payer negotiation.
When you’re evaluating RWD sources, the right question isn’t just “does this dataset cover my population?” It’s “does this dataset, combined with the right study design and analytical methods, support the evidence I need to generate, for the specific decision-making context I’m operating in?” Those are different questions. The second one is harder. And it’s the one that actually matters.
Key Use Cases: Matching Data Types to Business Questions
The most common mistake in real-world data strategy is approaching sourcing as a generic capability decision rather than a use-case-specific one. Different analytical questions require different data sources, different coverage characteristics, and different levels of longitudinal depth. Getting this mapping right at the outset saves months of engineering work and prevents the discovery of critical gaps after contracts are signed.
According to the ISPOR Real-World Data Task Force, aligning data source selection with specific decision-making contexts is foundational to generating credible real-world evidence. Here’s how that alignment plays out across the most common use cases:
Market Sizing and Diagnosed Prevalence
Claims data and aggregated insurance datasets are the primary tool here. Broad population coverage — across key Continental European markets and their statutory health insurance systems — allows you to estimate diagnosed prevalence with the statistical power needed for credible market sizing. The key variables are ICD-10 coding completeness, the breadth of the population capture relative to the total insured population, and data recency. A dataset that lags reality by six months is a serious problem for sizing an indication where diagnostic rates are changing quickly.
For launch planning specifically, our post on data-driven product launch planning with aggregated patient data covers the practical implications in depth.
HEOR and Cost-Effectiveness Analysis
Linked EHR-claims datasets with outcomes tracking are typically required here. Budget impact models and cost-effectiveness analyses need healthcare utilisation data — hospitalisations, specialist visits, procedures — alongside treatment costs and, ideally, clinical outcomes proxies. The methodological standard is high: EMA and national HTA bodies expect evidence of appropriate confounder adjustment, handling of informative censoring, and sensitivity analyses that test the robustness of your findings. Claims data alone is often insufficient; linked datasets with clinical enrichment are increasingly the expected standard.
Patient Journey Mapping and Treatment Sequencing
Longitudinal claims data with treatment sequence analysis is the core tool. Understanding how patients move through lines of therapy — from diagnosis to first treatment, through switching events, to second-line and beyond — requires continuous observation over time. The critical variable is longitudinal depth: some data sources can follow patients for five or more years, while others refresh frequently but don’t maintain extended patient histories. For loss-of-exclusivity planning and competitive intelligence on treatment dynamics, that depth distinction is decisive.
Synthetic Control Arms and Trial Design
This is a rapidly evolving area. As Thorlund et al. (2020) describe, synthetic and external control arms use real-world data to construct a comparator population that substitutes for or supplements a randomised control arm — particularly valuable in rare diseases, paediatric indications, and settings where randomisation is ethically or practically difficult. The data requirements are stringent: you need sufficient patient volume in the target population, high-quality covariate capture for propensity matching, and documentation of methodological decisions that will withstand regulatory scrutiny.
Post-Market Surveillance and Safety Monitoring
Broad population claims coverage is essential. Post-market commitments — particularly for conditional marketing authorisations and accelerated approvals — require surveillance across large, diverse populations to detect safety signals with statistical power. The geographic coverage of the dataset, its update frequency, and its completeness across patient subgroups all matter considerably here.
Competitive Intelligence and Launch Tracking
Near-real-time prescription and treatment pattern data — ideally with monthly or quarterly updates — is the tool of choice for commercial analytics teams tracking brand performance against competitors. The challenge is that the freshest data often sacrifices some depth, while the deepest longitudinal datasets are often the slowest to update. Understanding this trade-off, and selecting the right source for your specific intelligence need, is not always obvious from vendor pitch materials. Penberthy et al. (2022) provide useful context on data source selection for commercial analytics in oncology — but the principles extend broadly across therapeutic areas.
Regulatory Submissions and Label Expansions
Validated, analysis-ready datasets with documented quality are non-negotiable for regulatory-grade evidence. The FDA’s 2024 guidance on assessing EHR and medical claims data for regulatory decision-making is explicit about the quality standards expected. EMA’s DARWIN EU network is establishing similar expectations for European submissions. The implication: not all datasets that appear structurally similar are equally appropriate for regulatory use, and the documentation of validation methodology matters as much as the data itself.
The Synthetic Data Opportunity — What It Is, Where It Works, and Where It Doesn’t
Synthetic patient data is genuinely exciting — and genuinely misunderstood. We’ve sat in enough meetings where someone’s dismissed it as “made-up data” to know that the terminology creates a credibility problem before the conversation has started. So let’s be precise about what it is and what it isn’t.
High-quality synthetic healthcare data is generated algorithmically from real-world source data — typically using Generative Adversarial Networks (GANs) with differential privacy guarantees. The generation process learns the statistical structure of the source population — the distributions, correlations, and conditional relationships between variables — and produces a synthetic dataset that mirrors those properties without containing any actual patient records. As Jordon et al. (2021) demonstrate, when generated and validated correctly, synthetic data can preserve the utility of real patient-level data for a wide range of analytical applications.
The GDPR compliance advantage is significant. Anonymised synthetic data is not subject to GDPR Article 9 restrictions on special category health data in the same way that pseudonymised patient-level data is — provided the anonymisation methodology is robust, documented, and reviewable by data protection authorities. For Continental European analyses where access to patient-level claims data triggers complex regulatory requirements, synthetic data can unlock analytical pathways that would otherwise require months of DPA negotiation and legal review.
But here’s what I’d caution against: treating synthetic data as a universal substitute for real-world data. It isn’t. Chen et al. (2023) and Hernandez et al. (2023) both identify the boundaries clearly. Synthetic data is powerful for scenario modelling, algorithm training, and early-phase feasibility analysis. It’s a genuine enabler for rare disease research where patient volume in any single real-world source is insufficient. And it’s an effective tool for teams that need to develop and validate analytical approaches before committing to expensive real-data access agreements.
What it is not — at least not yet, in most regulatory contexts — is a substitute for real patient-level data in primary regulatory submissions, HTA dossiers, or high-stakes causal inference studies. Regulatory acceptance is growing, as Pasculli et al. (2025) document, but the evidentiary bar for synthetic-data-based submissions remains higher than for real-world-data-based submissions. Know where the boundary lies before you design your evidence programme around a synthetic data strategy. And understand that protecting patient privacy in synthetic health data remains an active area of regulatory development — the standards are still evolving.
The right framing is: synthetic data as a complement to real-world data, not a replacement. Used strategically — for the analytical questions it’s genuinely suited to — it adds significant value and capability. Used as a shortcut around the complexity of real-data access, it creates analytical fragility that surfaces at the worst possible moments. Our dedicated post on aggregated and synthetic insurance data explores this in practical detail.
What to Look For When Evaluating RWD Sources
The challenge many pharmaceutical companies face isn’t finding real-world data providers — it’s determining which combination of capabilities actually aligns with their strategic priorities. The vendor landscape is crowded, the pitch materials are sophisticated, and the claims made about coverage, quality, and compliance often sound identical across providers who are genuinely quite different in their actual capabilities.
The RECORD statement — the reporting guidelines for studies using routinely collected health data — provides a useful framework for thinking about what rigorous data use looks like. The ISPOR-ISPE Good Practices guidelines are the methodological standard for comparative effectiveness research. And the FDA’s 2024 guidance on assessing EHR and medical claims data sets out what regulatory-grade data quality looks like in practice. These are the frameworks you should be applying when evaluating data sources — not the vendor’s own characterisation of their data quality.
Here’s what actually separates adequate from excellent data partnerships across the dimensions that matter:
Coverage Breadth and Geographic Granularity
Coverage claims require scrutiny. A dataset described as “broad Pan-European coverage” might mean ten million lives or two hundred million — and that distinction is decisive for rare disease research, sub-national analysis, and therapeutic areas with low diagnosed prevalence. Geographic granularity matters too: national-level data is often insufficient for analyses that need to reflect regional healthcare delivery variation or sub-national HTA submission requirements. Ask for coverage statistics by therapeutic area and market, not aggregate headline figures.
Data Recency and Update Frequency
This varies dramatically across providers and is rarely disclosed prominently in pitch materials. Some datasets are updated monthly; others have a twelve-month or longer lag between real-world events and availability for analysis. For competitive intelligence, brand tracking, and launch analytics, data recency is not a secondary consideration — it’s a primary one. A dataset that reflects reality from nine months ago is telling a story that may no longer be true.
Linkage Quality Across Sources
Linked datasets — where claims are combined with EHR records, patient registries, or specialty pharmacy data — are only as valuable as the quality of the linkage. Linkage methodology matters enormously: deterministic matching on shared identifiers, probabilistic matching algorithms, and privacy-preserving record linkage approaches all produce different results and carry different biases. The linkage quality documentation provided by a data partner is one of the most revealing indicators of their overall methodological rigour.
Validation Methodology and Quality Documentation
Ask for it. Specifically. Validated datasets come with documented evidence of systematic quality checks — coding consistency, completeness assessments, comparison against reference population statistics. Datasets without this documentation are making an implicit claim that you should verify externally, which means additional time and cost before you can trust the outputs. The GRACE principles provide a useful benchmark for what high-quality observational data looks like.
GDPR and Compliance Architecture
Particularly for Continental European data, the compliance architecture of a data source matters legally, not just operationally. Is the data anonymised or pseudonymised? That distinction is legally significant under GDPR — pseudonymised data remains personal data under Article 9, while appropriately anonymised data may not. What is the documented anonymisation methodology? Has it been reviewed by a data protection authority? What contractual protections exist for data access, storage, and processing? For most organisations, legal and privacy counsel will have questions that go well beyond what sales materials address. Expect that review to take time, and budget for it.
Analytical Sophistication and Delivery Model
Analysis-ready datasets — pre-processed, validated, and structured for immediate querying — dramatically reduce the time-to-insight compared to raw data delivery that requires extensive engineering before analysis can begin. Analytical sophistication also varies: some providers offer query-based access to pre-structured datasets, while others provide AI-powered platforms for real-time pattern recognition. Neither model is inherently superior — the right choice depends on your internal capabilities, your analytical objectives, and your timeline. For more on this dimension, see our post on analysis-ready real-world evidence data.
Cost Structures and Contractual Flexibility
Cost models across the RWD landscape range from subscription-based platform access to per-query pricing to project-based engagement models. None is universally better — the right model depends on your usage pattern, your internal analytical capacity, and the frequency with which you’ll need access to the data. Contractual flexibility matters too: long-term exclusivity arrangements that lock you into a single provider before you’ve validated their capabilities for your use cases are a significant risk. Negotiate for phased access, pilot projects, or defined proof-of-concept engagements before committing to multi-year relationships.
Understanding which data sources match specific use cases — across all these dimensions simultaneously — requires deep market knowledge that most companies lack internally. This is where evaluating pharma data providers without the sales pitch becomes genuinely difficult, and where expert advisory support adds its most obvious value.
The Privacy and Compliance Landscape
Privacy compliance in real-world data is not a box-ticking exercise. It’s a substantive analytical constraint that shapes what data you can access, how you can use it, where it can be stored and processed, and what you can publish based on it. Getting it wrong doesn’t just create legal risk — it slows programmes, triggers regulatory scrutiny, and can invalidate analyses that were built on a foundation that data protection authorities later challenge.
The GDPR framework governs health data access across all EU member states and EEA markets. Health data is defined as special category personal data under Article 9 — meaning its processing requires explicit legal basis, robust technical safeguards, and, in most cases, some form of anonymisation, aggregation, or pseudonymisation before it can be used for research purposes. The distinction between anonymisation and pseudonymisation is legally significant: pseudonymised data (where a key exists that could re-identify individuals) remains personal data under GDPR and is subject to the full Article 9 restrictions. Truly anonymised data, where re-identification is not reasonably possible, is treated differently — but the threshold for demonstrating genuine anonymisation is high and is subject to national regulatory interpretation.
National implementations of GDPR vary by member state in ways that matter practically. The legal basis available for health data processing differs across national frameworks; what’s permissible in one market may not be permissible under the same approach in another. Cross-border data transfers for pan-Continental European analyses trigger additional requirements — particularly where data is processed outside the EEA. For most pharma teams, this complexity requires legal counsel with specific expertise in European health data regulation, not generic GDPR compliance expertise.
The EMA’s DARWIN EU network — documented at the EMA’s DARWIN EU page — is establishing a framework for federated data access across European health data networks that addresses some of these cross-border challenges. Under a federated model, analyses are run locally within each national data environment rather than pooling patient-level data centrally. This approach has significant compliance advantages but also methodological constraints — not all analytical questions can be answered through federated queries, and the results require careful harmonisation across national coding and classification systems.
For US-focused programmes or global companies operating across both markets, HIPAA adds a parallel compliance framework. The HHS guidance on de-identification of protected health information establishes two approaches — Expert Determination and Safe Harbor — with different requirements and evidentiary standards. Understanding which approach a US data provider has used for de-identification, and whether it meets the standard appropriate for your use case, requires technical review that goes beyond what most vendor sales materials address. And as Walonoski et al. (2025) discuss, protecting patient privacy in synthetic health data adds an additional layer of regulatory consideration that is actively evolving.
The practical implication: build compliance review into your data sourcing timeline, not as an afterthought after vendor selection. Legal and privacy counsel who haven’t reviewed a data source before licensing negotiations begin are a consistent source of late-stage delays in real-world data programmes. Engage them early, give them the technical documentation they need to make an informed assessment, and don’t assume that a vendor’s claim of GDPR compliance is equivalent to a legal confirmation that their data is appropriate for your specific use case.
The RWD Vendor Landscape: Why Navigation Matters
The real-world data vendor market is genuinely complex. There are dozens of providers across the Continental European and global markets, with overlapping but distinct capabilities, varying geographic strengths, and different specialisations by therapeutic area, data type, and analytical model. Some maintain broad population coverage across multiple markets. Others offer deep clinical enrichment within specific disease areas. Some have invested heavily in analysis-ready delivery models; others provide access to raw data that requires significant engineering before it can support analysis. Some have established track records with EMA and national HTA bodies; others are newer entrants whose regulatory acceptability is still being established in practice.
Most organisations discover this complexity only after they’ve entered the vendor market. The vendor’s incentive — quite understandably — is to present their capabilities as comprehensively as possible. The buyer’s challenge is to evaluate those claims against their specific requirements, on dimensions that the vendor may not have volunteered.
As Makady et al. (2017) documented in their analysis of real-world data policies across national HTA bodies, the methodological requirements for acceptable real-world evidence vary significantly by country and by decision context. A dataset that works well for commercial analytics may be inadequate for a formal HTA submission. A source that’s powerful for oncology may have serious coverage gaps in rare metabolic diseases. Understanding these distinctions before you commit to a vendor — rather than after — is the difference between a data investment that delivers value and one that gets written off.
The gap between licensing claims data and extracting actionable intelligence is substantial. That gap is where expert guidance becomes genuinely valuable — not because the landscape is impenetrable, but because navigating it well requires familiarity with what each source type actually delivers in practice, how different providers perform on the dimensions that matter for specific use cases, and where the mismatches between vendor capability claims and real-world performance are most common. Expert advisors who understand both your business objectives and the vendor landscape can accelerate your path to value while avoiding costly mismatches.
The GRACE principles for recognising high-quality observational studies are a useful independent benchmark for what good data-driven evidence looks like — and applying them to vendor evaluation conversations is a more reliable approach than relying on vendor-provided quality documentation alone.
Internal Build vs. External Partnership: An Honest Assessment
Every organisation with a serious real-world data need eventually asks some version of this question: should we build internal RWD infrastructure, or should we access external analysis-ready datasets? It’s a reasonable question, and the honest answer is: it depends — on your stage, your therapeutic focus, your analytical ambition, and your time horizon.
Building internal RWD infrastructure — data engineering pipelines, data lakes, in-house analytical platforms — gives you control, customisability, and the ability to integrate proprietary data sources that external providers don’t have access to. It also takes years and tens of millions of dollars to build properly. The organisations that have done it successfully are large-cap pharma companies with extensive data science functions, dedicated RWE teams, and the organisational patience to absorb multi-year build timelines. For smaller biotech or specialty pharma companies with one or two late-stage assets, the build option is rarely credible.
As Kesselheim and Doherty (2021) note in their assessment of real-world evidence in drug regulation, the organisational infrastructure required to generate regulatory-grade RWE consistently is non-trivial. Schneeweiss et al. (2019) add methodological context on study design that’s relevant to any team trying to build this capability internally.
The more interesting question for most organisations isn’t “build or buy” in the abstract — it’s “what capabilities do we genuinely need to own internally, and where is external partnership more efficient?” Internal ownership makes sense for: proprietary data assets that you generate through patient engagement programmes or clinical operations; analytical capabilities that are genuinely differentiating and that you’ll use at scale; and methodological expertise in causal inference and study design that defines your evidentiary strategy. External partnership makes sense for: population-level claims data that no single company can generate internally; geographic coverage across markets where you don’t operate; and analysis-ready delivery models that bypass the data engineering bottleneck entirely.
Most organisations that approach this honestly end up with a hybrid model. The question is how to structure the hybrid so that internal investment goes into capabilities that are genuinely differentiating, and external partnerships are selected for fit with specific use cases rather than breadth of generic capability. Defining that architecture clearly — before entering vendor conversations — is where expert advisory support typically adds the most leverage.
Getting Started: Defining Your RWD Requirements
Here’s what I’ve noticed consistently: companies that enter the vendor market without clearly defined requirements spend significantly longer in vendor evaluation, make more costly mistakes in vendor selection, and generate the most internal friction around data investments. The clarity of your requirements — before you talk to any provider — is the single most controllable variable in the success of a real-world data programme.
The FDA’s 2021 guidance on considerations for RWD and RWE and ISPOR’s real-world evidence transparency initiative both frame requirements definition as foundational to any credible evidence programme. The frameworks are designed for regulatory submissions but the underlying logic applies equally to commercial analytics and market access use cases.
Before entering any vendor conversation, define these dimensions explicitly:
Therapeutic area and indication specificity. Not “oncology” — the specific tumour type, line of therapy, and patient population you need to analyse. Coverage varies dramatically at this level of specificity, and providers who claim comprehensive oncology data may have significant gaps in the exact patient population you need.
Geographic scope and granularity. Which Continental European markets matter for your evidence programme? Do you need sub-national granularity, or national-level data sufficient? Are you submitting to HTA bodies in multiple markets simultaneously? The answers shape the coverage requirements decisively.
Analytical objectives and study design implications. Is this a descriptive epidemiology study, a comparative effectiveness analysis, a safety surveillance programme, or a synthetic control arm for a regulatory submission? Each analytical objective implies different methodological requirements, which in turn imply different data specifications. Define the study design before you define the data requirements — not the other way around.
Compliance constraints. What legal basis does your organisation have for processing European health data? Has legal counsel been engaged? What anonymisation standard is required for your specific use case? Are there cross-border data transfer implications that need to be addressed before data access agreements can be signed?
Timeline and time-to-value expectations. When do you need insights? Working backwards from the strategic deadline — a launch date, an HTA submission window, a payer negotiation — gives you the constraint that governs everything else. Analysis-ready datasets are worth a premium if the alternative is six months of data engineering before your first query.
Budget parameters and cost model preferences. What’s the total cost of ownership you’re working with — including not just licensing fees but data engineering, analytical services, and the internal resource required to manage the relationship? And what cost model suits your usage pattern: subscription access, per-project pricing, or a hybrid?
If you’re working through these questions and finding that the answers are less clear than you’d like — or that different internal stakeholders have different views on what the requirements actually are — that’s a signal that expert advisory support before entering the vendor market would save significant time and cost. Navigating this landscape requires expertise in both pharmaceutical commercialisation and the evolving real-world evidence ecosystem. Understanding which data sources match specific use cases, across the full set of requirements, requires deep market knowledge that most companies lack internally.
That’s exactly the conversation we have with every company we work with at MEDDDICAL. Not a vendor pitch — an honest assessment of what you need, what the landscape can deliver, and where the gaps and risks are. If you’d like to start that conversation, contact us to discuss your specific evidence requirements and explore how expert RWE advisory can accelerate your path to value.
Sources & References
- Lohr, S. (2014). For Big-Data Scientists, ‘Janitor Work’ Is Key Hurdle to Insights. The New York Times. https://www.nytimes.com
- Berger, M. L., et al. (2017). A framework for regulatory use of real-world evidence. Journal of Comparative Effectiveness Research, 6(1), 5–8. https://www.tandfonline.com
- Sherman, R. E., et al. (2016). Real-world evidence — what is it and what can it tell us? New England Journal of Medicine, 375(23), 2293–2297. https://www.nejm.org
- Dang, A. (2023). Real-world evidence: a primer. Pharmaceutical Medicine, 37(1), 25–36. https://pubmed.ncbi.nlm.nih.gov/36604368/
- U.S. Food and Drug Administration (2018). Framework for FDA’s Real-World Evidence Program. https://www.fda.gov/media/120060/download
- Blonde, L., et al. (2018). Interpretation and impact of real-world clinical data for the practicing clinician. Advances in Therapy, 35(11), 1763–1774. https://link.springer.com
- Zhu, R., et al. (2023). Clinical pharmacology applications of real-world data and real-world evidence in drug development and approval. Clinical Pharmacology & Therapeutics, 114(3), 516–527. https://pubmed.ncbi.nlm.nih.gov/37393555/
- Garrison, L. P., et al. (2007). Using real-world data for coverage and payment decisions: the ISPOR Real-World Data Task Force report. Value in Health, 10(5), 326–335. https://www.valueinhealthjournal.com
- Penberthy, L., et al. (2022). An overview of real-world data sources for oncology and considerations for research. CA: A Cancer Journal for Clinicians, 72(3), 228–245. https://acsjournals.onlinelibrary.wiley.com
- Thorlund, K., et al. (2020). Synthetic and external controls in clinical trials — a primer for researchers. Clinical Epidemiology. https://pubmed.ncbi.nlm.nih.gov/32440224/
- Jordon, J., et al. (2021). Synthetic data in machine learning for medicine and healthcare. Nature Biomedical Engineering, 5, 493–497. https://www.nature.com/articles/s41551-021-00751-8
- Chen, R. J., et al. (2023). Harnessing the power of synthetic data in healthcare: innovation, application, and privacy. npj Digital Medicine. https://www.nature.com/articles/s41746-023-00927-3
- Pasculli, et al. (2025). Synthetic data in healthcare and drug development: definitions, regulatory frameworks, issues. CPT: Pharmacometrics & Systems Pharmacology. https://ascpt.onlinelibrary.wiley.com/doi/full/10.1002/psp4.70021
- Hernandez, M., et al. (2023). Synthetic data in health care: a narrative review. PLOS ONE. https://pmc.ncbi.nlm.nih.gov/articles/PMC9931305/
- Benchimol, E. I., et al. (2015). The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. PLOS Medicine, 12(10). https://journals.plos.org/plosmedicine
- ISPOR-ISPE Special Task Force. Good practices for real-world data studies of treatment and/or comparative effectiveness. https://www.ispor.org
- U.S. Food and Drug Administration (2024). Real-world data: assessing electronic health records and medical claims data to support regulatory decision-making for drug and biological products. https://www.fda.gov
- U.S. Department of Health & Human Services. Guidance regarding methods for de-identification of protected health information in accordance with HIPAA Privacy Rule. https://www.hhs.gov/hipaa
- Walonoski, J. A., et al. (2025). Protecting patient privacy in tabular synthetic health data: a regulatory perspective. npj Digital Medicine. https://www.nature.com/articles/s41746-025-02112-0
- European Medicines Agency (2023). Data analysis and real-world interrogation network (DARWIN EU). https://www.ema.europa.eu
- Makady, A., et al. (2017). Policies for use of real-world data in health technology assessment. Value in Health, 20(3), 520–532. https://www.valueinhealthjournal.com
- Dreyer, N. A., et al. (2018). GRACE principles: recognizing high-quality observational studies of comparative effectiveness. American Journal of Managed Care, 16(6), 467–471. https://www.ajmc.com
- Kesselheim, A. S., & Doherty, M. (2021). Real-world evidence in drug regulation — challenges and opportunities. Health Affairs, 40(2). https://www.healthaffairs.org
- Schneeweiss, S., et al. (2019). Graphical depiction of longitudinal study designs in health care databases. Annals of Internal Medicine, 170(6), 398–406. https://www.acpjournals.org
- U.S. Food and Drug Administration (2021). Considerations for the use of real-world data and real-world evidence to support regulatory decision-making for drug and biological products. https://www.fda.gov
- ISPOR (2022). Real-world evidence transparency initiative: guiding principles. https://www.ispor.org