From Data Access to Public Trust: What the French Health Data Hub Brought to Wales

HDF 2025 Wales Blog | Lessons from the French Health Data Hub

Lead
At the Health Data Forum Global Hybrid Summit Wales 2025, Emmanuel Bacry's presentation on the French Health Data Hub offered one of the clearest examples of how a national health data platform can move from policy ambition to operational practice. The follow-up input from the Health Data Hub adds an important layer: the future of secondary use depends not only on access to data, but on workflows, transparency, interoperability, linkage capacity, de-identification and trust.

One of the strongest messages emerging from the Health Data Forum Wales 2025 was that health data systems are no longer being judged only by what they promise. They are being judged by what they can actually make possible.

• Can they support public-interest research?
• Can they help researchers, innovators and health systems access the right data safely?
• Can they link claims, registries, cohorts and clinical datasets without weakening privacy?
• Can they create transparency for citizens?
• Can they help Europe prepare for the real-world implementation of the European Health Data Space?

These questions were brought vividly into the conversation in Wales through the contributions of Emmanuel Bacry and the French Health Data Hub.

A National Platform for Public-Interest Research

The French Health Data Hub was created as a public-interest group to facilitate access to health data for public-interest research. This includes academic research, public-sector projects and, where appropriate, private-sector innovation serving a public-interest purpose.

That distinction matters.

The Health Data Hub is not simply a database. It is part of a national infrastructure for making health data usable, governable and secure. Its role spans governance, data access support, a secure technological platform and practical assistance to projects across their user journey.

In Wales, Emmanuel explained that the Health Data Hub is already supporting more than 200 innovative projects. This makes it a particularly valuable case study for countries and regions now designing their own data lakes, data spaces or health data access bodies.

The lesson is clear: secondary use is not only about opening access. It is about creating the conditions that allow access to happen responsibly.

The SNDS: A Footprint of Health Pathways

At the centre of the French model is the National Health Data System, known as the SNDS.

Emmanuel described it as a reimbursement-based database rather than a clinical database. Because France has broad healthcare reimbursement coverage, the SNDS offers a powerful footprint of healthcare pathways across the population.

This gives the system particular value for pharmacoepidemiology, pharmacovigilance and, increasingly, for linkage with clinical databases. A clinical database may contain rich information about a patient within a specific care episode or registry, but may not show what happened before or after that episode. The SNDS can help fill this longitudinal gap.

This is especially important for AI and real-world evidence.

AI in healthcare often fails not because algorithms are weak, but because the data context is incomplete. Claims data, registries and clinical datasets each tell part of the story. The power comes when they can be responsibly connected.

Making Data Ready Before Access Is Requested

One of the most interesting features of the French approach is the notion of a catalogue of databases identified as nationally important.

These datasets are not merely listed. They are intended to be prepared, documented and, where relevant, made ready for use on the Health Data Hub platform. Some are updated regularly so that, once an authorisation is granted, a secure project space can be opened more rapidly.

This speaks to a practical challenge that many health data systems face: authorisation may be granted, but the data may still be difficult to find, prepare, link, document or use.

The Health Data Hub's follow-up input also highlights work on a granular metadata catalogue aligned with the European Health Data Space and the HealthDCAT-AP standard. This is particularly important because EHDS will require not only data access bodies but also discoverability. Researchers and innovators must be able to understand what data exists, under what conditions it can be accessed, and what it can realistically support.

In other words, metadata is not administrative decoration. It is part of the infrastructure of trust and usability.

Secure Project Environments and Responsible Use

For projects using the Health Data Hub platform, data is made available in secure project environments. These environments are designed to enable users to work with pseudonymised data while preventing unauthorised data extraction.

This is an important operational principle.

If we want to support secondary use, we need to move beyond the simplistic question of whether data is shared or not shared. The more relevant question is: under what governance, security and technical conditions can data be used?

Secure environments allow authorised users to work with approved datasets while maintaining strong controls over data export, software access and project-specific identifiers. The Health Data Hub model also includes human validation before data leaves the secure space.

This is where trust becomes practical.

Trust is not only a communication message. It is built into workflows, permissions, encryption, documentation, audit trails, review processes and the everyday design of the platform.

Data Linkage as a National Capability

One of the most valuable parts of Emmanuel's presentation was his explanation of data linkage.

Linking datasets is often where the promise of health data becomes real. It is also where complexity, risk and operational burden increase sharply.

The Health Data Hub follow-up document describes two main approaches when the French patient identifier is not initially available: direct matching by reconstructing the identifier when civil-status information is available, and indirect matching using shared characteristics such as date of care, institution code, procedure code, or diagnosis.

The document also describes the "My linkage space" tool, developed to industrialise routines that would otherwise be too manual and resource-intensive.

This is a significant lesson for other countries.

Data linkage cannot be treated as an occasional technical workaround. It needs to become a capability. If health systems want to connect claims, registries, hospital data warehouses, cohorts and clinical datasets, they need safe, repeatable and well-governed linkage processes.

Without this, secondary use will remain slow, fragile and uneven.

De-identification: More Than Removing Names

The Health Data Hub's follow-up input also gives useful detail on de-identification and anonymisation.

All ingested data is already pseudonymised, with direct identifiers hashed. The Health Data Hub applies multiple layers of encryption and uses project-specific identifiers within secure environments, so the same patient will not share the same identifier across projects.

Before data can be exported, anonymisation is required. The follow-up document refers to established privacy models such as k-anonymity, l-diversity and other thresholds designed to reduce re-identification risks while preserving data utility. It also describes a combination of automated checks, manual review and scenario-based risk assessment.

This matters because the language of anonymisation is often used too casually.

In practice, de-identification is not a single switch. It is a set of technical, legal and organisational measures. It must be assessed in context, especially when working with rich health data, rare diseases, genomics, imaging or linked longitudinal datasets.

For the Health Data Forum, this is directly aligned with the Wales conversation on trust, transparency and the responsible use of health data.

The Human Side of Data Sharing

Perhaps the most practical advice Emmanuel offered in Wales came in response to a question about countries now building their own data lakes.

His answer was not primarily technical. It was relational.

The hardest challenge, he suggested, is onboarding data producers into a common project of health data sharing. Data producers need to understand why participation matters, what they gain from it, and why it is good for their databases, their institutions and their mission.

A legal obligation may force compliance, but it does not create commitment.

This is an essential point for Europe.

The success of EHDS will depend not only on regulation, but on whether data producers, hospitals, registries, public agencies, researchers and citizens feel part of a shared purpose. Interoperability also has to begin early. Standards should not be imposed after the architecture is already designed. They should be discussed and agreed with data producers from the start.

This may be one of the most important lessons from the French experience: infrastructure is technical, but adoption is social.

Europe, EHDS and the Next Stage

The French Health Data Hub has also played a significant role in European work around the European Health Data Space, including pilot activity for secondary use.

In Wales, Emmanuel described work on connecting national platforms, creating a central portal to query metadata, developing a common access form and running use cases to identify bottlenecks. One of those use cases, involving genetic and clinical data linkage, exposed how difficult the real-world implementation of EHDS can become.

That is precisely why pilots matter.

They are not valuable because everything works smoothly. They are valuable because they reveal where governance, data standards, linkage, consent, technical architecture and institutional readiness still need to mature.

For the Health Data Forum, the French experience is especially relevant to the wider Data First, AI Later movement.

AI will not become trustworthy in healthcare simply because better models are developed. It will depend on the quality, structure, governance, and legitimacy of the underlying data foundations.

What Wales Takes Forward

The Health Data Hub contribution to Wales helped make one thing clear: the future of secondary use will not be built by slogans.

It will be built through access workflows, project directories, metadata catalogues, data standards, secure environments, linkage capabilities, de-identification methods, help desks, documentation, and citizen-facing transparency.

It will require national platforms, but also networks of interoperable hospital data warehouses.

It will require technical architecture, but also trust.

It will require regulation as well as incentives.

It will require a public-interest purpose as well as operational excellence.

As Europe moves from EHDS policy to implementation, the French Health Data Hub offers a valuable reference point — not as a perfect model to copy, but as a living example of the work required to turn health data into usable public infrastructure.

That is one of the enduring lessons from Wales.

Before we scale AI, we must learn how to make health data work responsibly.

And before we ask citizens and institutions to trust the future, we must build the systems that deserve that trust.

Source note
This article is part of the Health Data Forum Wales 2025 Blog and is based on Emmanuel Bacry's presentation on the French Health Data Hub during the Health Data Forum Global Hybrid Summit Wales 2025, together with the follow-up input shared by the Health Data Hub after the Summit.