What is a Health Data Space and how does it work?

 

What is a Health Data Space and how does it work?

A structured guide to understanding health data spaces: their architecture, governance, and what they actually unlock.

Healthcare doesn’t lack data. It lacks usable data.

Every hospital visit, prescription, lab test, and imaging scan generates data. At scale, health systems hold billions of records, a potential goldmine for research, clinical decision-making, and public health policy. And yet, most of that data sits locked in isolated systems, formatted inconsistently, governed by conflicting rules, and practically inaccessible to anyone who might put it to use.

The problem isn’t volume. It’s usability.

A radiologist in Madrid can’t easily share imaging data with a researcher in Amsterdam. A public health agency can’t query hospital records from multiple regions without a months-long approval process, if it’s even possible at all. Clinical trial sponsors struggle to find eligible patients scattered across dozens of institutions, each with its own data format and access policy.

Health data spaces are a direct response to this problem, and are becoming a central pillar of the European Health Data Space (EHDS). They are not just another data sharing initiative. They represent a structural shift in how health data is governed, connected, and made fit for use, across institutions, regions, and borders.

What is a Health Data Space? A working definition (EHDS context)

 

A health data space is a governed digital infrastructure that enables health data from multiple sources to be shared, accessed, and reused – securely and in compliance with applicable rules – without requiring that data to be physically centralised.

Four elements make this definition operational:

It is federated, not centralised. Data stays at the source; queries and approved access travel to the data, not the other way around.
It is governed. Rules – legal, technical, and ethical – determine who can access what, for what purpose, and under what conditions.
It is interoperable. Data from different systems can be understood and combined, because common standards are applied.
It is built for reuse. The same data that supports direct patient care can also, under appropriate controls, support research, innovation, and policy.

This last point is fundamental: a health data space is not just infrastructure for moving data. It is infrastructure for making data usable.

Why health data spaces need to exist

 

The underlying problem is structural. Health data is generated across a fragmented ecosystem – primary care, hospitals, pharmacies, diagnostic labs, wearables, registries – and each node operates with its own systems, standards, and rules. The result is a landscape of silos.

Health data interoperability standards like HL7 FHIR, SNOMED CT, and LOINC have made real progress in enabling data exchange. But technical connectivity, on its own, is insufficient. Connecting systems doesn’t automatically make data usable. You can have two hospitals exchange records in a shared format and still end up with data that is clinically ambiguous, inconsistently coded, or incomplete – because the underlying clinical documentation practices differ.

Beyond the technical layer, there are governance gaps. Who has the right to access which data, for what purpose, under what oversight? Without a clear, enforceable framework, organisations default to caution, locking data down rather than risking regulatory exposure. This is rational at the individual institutional level. It is catastrophic at the system level.

Health data spaces address both dimensions: they establish the technical architecture for sharing and the governance framework that makes sharing legitimate and safe.

Health Data Space Architecture: How it works in four layers

 

A health data space is best understood as a stack of four interdependent layers. Each one is necessary; none is sufficient on its own.

1 – Data Layer

The foundation. This is where health data actually lives, in hospital EHR systems, health data spaces and registries, diagnostic databases, pharmacy records, and increasingly, in patient-generated data from wearables and apps. The data layer is intentionally decentralised: sources keep ownership and control of their data. Nothing moves without authorisation from the upper layers.

2 – Interoperability Layer

This layer defines the common language. Standards like HL7 FHIR for data exchange, SNOMED CT for clinical terminology, DICOM for medical imaging, and ICD-10 for diagnostic coding allow data from disparate sources to be understood by other systems. Without this layer, connecting two hospitals would be like combining two databases with different column names and different units of measurement, technically possible, practically useless.

Importantly, the interoperability layer covers semantic as well as syntactic compatibility. It’s not enough that two systems can exchange a file; they need to agree on what the concepts in that file mean.

3 – Governance Layer

This is the layer that makes sharing legitimate. It defines data access policies, consent mechanisms, purpose limitations, audit trails, and accountability structures. It determines who can be a data holder, who can be a data user, what purposes are permitted, and what oversight bodies are responsible for enforcement.

In the EHDS framework, this role falls to Health Data Access Bodies (HDABs), national entities responsible for evaluating access applications, issuing data permits, and ensuring that processing occurs inside accredited secure environments. Governance is also where GDPR compliance, pseudonymisation obligations, and prohibitions on harmful uses are enforced.

4 – Access and Use Layer

The top layer is where authorised users – researchers, regulators, clinicians, innovators – actually work with data. Access typically occurs inside Trusted Research Environments (TREs) or Secure Processing Environments (SPEs): controlled, audited spaces where approved queries can be executed against pseudonymised data, without ever exposing the raw records to the external user. Outputs are vetted before release to minimise disclosure risks.

The architecture is designed so that data never needs to leave its source in identifiable form. Researchers bring their algorithms to the data; not the other way around.

From an operational perspective, a health data space is not only about interoperability, but about governed data flow: datasets become discoverable through metadata catalogues, access requests are submitted for a defined use case, permissions are assessed under the applicable legal and governance framework, and data is accessed or analysed within secure environments designed to preserve control, traceability and compliance. This is what enables secondary use at scale without reducing trust.

Primary use vs. secondary use of health data

 

This distinction is one of the most important – and most frequently confused – concepts in health data governance.

Primary use refers to the processing of health data to deliver care to the patient who generated it. A GP accessing your medical history to prescribe a treatment. A hospital sharing your discharge summary with a specialist. A pharmacist checking your current medications before dispensing. This is the original, core purpose for which health records exist.

Secondary use of health data refers to processing the same data for purposes beyond direct care: research, clinical trials, drug development, health technology assessment, epidemiological surveillance, AI model training, and policy analysis. The patient is not the direct beneficiary of any specific use; the beneficiary is the healthcare system, science, or society.

The regulatory and technical requirements for secondary use are substantially higher. Data must be pseudonymised or anonymised. Access must be purpose-limited and approved. Processing must occur inside secure environments. Users cannot attempt re-identification. And – critically – secondary use access can be declined: the EHDS explicitly prohibits secondary use for advertising, marketing, insurance discrimination, or any purpose that could be detrimental to the individuals whose data is used.

Health data spaces are designed to enable both use cases. The primary use infrastructure (in the EHDS, called MyHealth@EU) supports cross-border care continuity – for example, making patient summaries and e-prescriptions accessible to treating clinicians across EU member states. The secondary use infrastructure (HealthData@EU) enables approved researchers and innovators to query federated datasets under strict governance controls. For a deeper look at how this distinction works in practice – and why it is so frequently misunderstood – see Primary vs secondary use of health data: what everyone gets wrong.

Why interoperability alone is not enough

 

Connected data is not the same as usable data. This distinction is easy to underestimate, and it is where many health data initiatives fall short.

Consider what happens when hospital records from ten different institutions are made technically accessible through a shared API. The data might be formatted in FHIR. But the clinical coding may be inconsistent – one hospital uses SNOMED systematically; another uses free-text; a third codes the same condition under three different terms depending on the department. Diagnoses may be recorded differently depending on whether the encounter was inpatient or outpatient. Comorbidities may be documented in structured fields in some systems and buried in narrative discharge summaries in others.

At the syntactic level, the data is interoperable. At the clinical level, it is not ready for analysis.

This is the gap between connection and usability. Solving it requires more than standards compliance. It requires a process of clinical data curation, the harmonisation, validation, and structuring of raw clinical records into datasets that can actually answer research and clinical questions. In many health systems, a significant portion of clinically relevant information exists not in structured fields, but in unstructured clinical notes: physician observations, consultation summaries, radiology reports, discharge letters.

Making that information usable requires extracting it, normalising it, and mapping it to shared clinical concepts. Without this, interoperability creates the illusion of access while delivering noise.

The missing layer: making data clinically usable

 

This is perhaps the most underappreciated challenge in health data infrastructure. The regulatory and technical frameworks for data spaces are advancing rapidly. The EHDS is now law. National HDABs are being established. FHIR adoption is accelerating. Secure processing environments are being deployed.

And yet, the clinical quality of the underlying data – the thing that determines whether any of this infrastructure produces valid, actionable insights – receives comparatively little attention in the governance discussions.

Consider a research team granted access to a federated dataset of 500,000 patients with a specific diagnosis. They submit their query. They receive pseudonymised records. But the diagnosis was coded differently across institutions. The relevant comorbidities were documented in free-text notes, not structured fields. The medication history is incomplete because transitions between care settings weren’t captured. The dataset is technically accessible but clinically unreliable.

The infrastructure worked as designed. The science can’t proceed.

Closing this gap – between connected data and analytically usable data – is a distinct and largely unsolved problem. It requires investment in clinical data quality processes, terminology mapping, and tools capable of extracting structured information from unstructured clinical text. A health data space that stops at the governance and interoperability layers leaves this problem unaddressed. The value promised by the infrastructure remains locked inside the clinical record.

What health data spaces actually unlock

 

When all four layers work – data, interoperability, governance, and clinical usability – health data spaces become genuine enablers of evidence generation at scale.

The specific capabilities they unlock include:

Real-world evidence (RWE) at population scale. Researchers can query millions of patient records across institutions and countries to understand how treatments perform outside controlled trial conditions, capturing the heterogeneity of real patients, real prescribing patterns, and real outcomes.
Rare disease research. Patients with rare conditions are, by definition, scattered across many centres. Data spaces allow researchers to find and aggregate records that would never reach statistical significance within a single institution.
Post-market surveillance. Regulators and manufacturers can monitor the safety and effectiveness of medicines, devices, and diagnostics against large real-world populations – moving beyond the limitations of pre-approval trials.
AI model development and validation. Machine learning models for diagnosis, prognosis, and treatment recommendation require large, diverse, and clinically accurate datasets. Data spaces provide the infrastructure for building them, and for validating them across populations that weren’t in the training set.
Health policy and preparedness. Public health authorities can analyse disease patterns, treatment variation, and health inequalities at scale, and respond faster during emergencies.

None of this is speculative. Each of these use cases is already documented in pilot programmes across Europe and globally. The limiting factor is not the concept; it is the maturity of the infrastructure and the quality of the underlying data.

Real-world examples: EHDS and national implementations

 

The European Health Data Space (EHDS) – Regulation (EU) 2025/327, published in the EU Official Journal on 5 March 2025 – is the most comprehensive health data space framework currently in force. It creates a sector-specific legal and technical architecture for electronic health data across all 27 EU member states, covering both primary use (patient care) and secondary use (research and innovation).

Its implementation is phased. By March 2029, the first interoperability functionalities are expected to be operational: cross-border access to patient summaries and e-prescriptions, and the ability for Health Data Access Bodies to process secondary use requests. Full secondary use infrastructure – including genomic and imaging data – extends to 2031 and beyond.

The HealthData@EU Pilot (2022 – 2024) provided the first real-world test of the EHDS infrastructure, connecting national data platforms from 12 European countries. The pilot demonstrated cross-border data sharing is technically feasible, and identified the governance and semantic harmonisation challenges that remain to be solved.

At the national level, implementation varies considerably. Finland’s Kanta system – one of Europe’s most advanced national health data platforms – has been operational since 2010 and provides a model for EHDS-aligned infrastructure, including pseudonymised data access for researchers. Italy is restructuring its Fascicolo Sanitario Elettronico (FSE 2.0) to align with EHDS requirements, though regional fragmentation remains a significant challenge. Spain, Finland, Estonia, and the Netherlands are among the member states furthest along in preparing their national data access body structures.

Outside Europe, comparable frameworks exist. The UK’s NHS Secure Data Environment (SDE) programme provides a national federated infrastructure for research access to NHS data. Brazil’s RNDS (Rede Nacional de Dados em Saúde) is establishing national health data exchange infrastructure. These initiatives share the same underlying logic: federated access, governed secondary use, and secure processing environments.

The infrastructure is being built. The harder problem remains.

 

A health data space is not a database. It is not a data warehouse. It is not an API. It is a system of systems – technical, legal, and institutional – designed to make health data accessible to those who can use it responsibly, while keeping it under the control of those who generated it.

The frameworks are now in place. The EHDS is law. National bodies are being established. Standards are converging. The next ten years will see more health data made accessible for research and innovation than the previous fifty.

But infrastructure creates possibility, not outcomes. The question is what happens when researchers and clinicians actually sit down with the data. Whether the records they access reflect clinical reality – accurately, completely, consistently – will determine whether the infrastructure delivers on its promise.

Connected data is necessary. Usable data is the goal. Closing that gap is the work that still needs doing. And that is Savana’s mission. With its technology designed to extract clinical insights from unstructured text through analysing and harmonising procedures, it transforms medical information into governed, actionable data, compliant with current regulations. This is what enables real-world evidence to become scalable, reliable, and actionable across healthcare systems.

Sources & Further Reading

 

Learn more

Want to move from connected health data to governed, usable evidence?

Explore how Savana helps hospitals, healthcare authorities, pharma, and MedTech organisations transform fragmented clinical data into actionable insights.

ABOUT SAVANA

Savana is the only healthtech SaaS company delivering scientifically validated, multilingual Clinical NLP exclusively trained on EHRs, ensuring the deepest dRWE for hospitals, HTA bodies, pharma, and MedTech. It also powers the Smart Health Alliance, a global hospital network collaborating within Savana’s ecosystem to drive the most innovative and disruptive RWE use cases, including the market’s only AI-powered living registries.

Share it here:

Related Posts

The evidence behind the headlines.

Stay ahead of how real-world data is reshaping clinical practice, drug development, and healthcare policy.