A Practical Path Forward

By Adam Skali

Europe’s health systems generate a wealth of valuable data—clinical records, diagnostics, treatment outcomes, and more. Yet much of this information remains underused, locked within national systems or isolated institutions. This fragmentation limits the EU’s ability to conduct large-scale research, monitor public health trends, or develop robust AI tools. At a time when healthcare is increasingly shaped by data, this disconnect is a missed opportunity.
Recognising this, the EU has introduced the European Health Data Space (EHDS) to enable more effective data use across borders. But strong privacy protections—particularly the General Data Protection Regulation (GDPR)—mean that simply pooling sensitive health data into a central platform isn’t viable. Legal requirements vary from country to country, and concerns about trust and security remain.

A more pragmatic solution is needed—one that allows collaboration without requiring countries or institutions to give up control over their data. That’s where federated data networks come in.
The core idea is simple: instead of moving data to a central location, you move the analysis to where the data resides. Hospitals and research centres keep full control over their records. When a study or AI training needs to be run, a model or query is sent to each site. Each site runs the analysis locally and returns only the results—aggregated statistics or model updates. No raw patient data is transferred.

This approach aligns well with Europe’s legal and ethical expectations. It supports GDPR principles like data minimisation and privacy by design. It allows institutions to respect local laws while contributing to European-scale insights. And it builds trust—because the data stays under the control of those who collected it.

Federated networks also support one of the EU’s most important ambitions: enabling innovation, especially in the field of AI. Building effective AI tools in healthcare requires access to large, diverse, and high-quality datasets. But doing that in a centralised way often runs into legal and technical barriers. Federated learning—where AI models are trained across multiple sites without centralising the data—addresses this challenge directly.

Models trained in this way tend to be more robust. They draw on more varied data sources, which helps reduce bias and makes them more generalisable across different patient populations.

This is particularly valuable in fields like cancer care, imaging, and rare disease research, where no single dataset is large or diverse enough on its own. Several major projects, including efforts by the NHS in the UK and research networks in the UK and EU, have shown that federated learning can deliver meaningful results—safely and efficiently.

Some examples include:

• EHDEN: A pan-European network standardising hospital data via the OMOP Common Data Model to enable privacy-preserving federated analyses for research.
• CINECA: A federated infrastructure linking national genomic databases across Europe, allowing secure analysis of highly sensitive genomic data without moving it.

That said, successful federation isn’t automatic. It requires shared data standards, strong technical infrastructure, and a clear governance model. Each participating site needs the ability to process data securely and consistently. And collaboration requires trust—not just in technology, but in how decisions are made, how consent is managed, and how results are shared.

For Europe, one possible path forward would be to have EHDS implemented with federated data as a central design principle. Regulatory guidance should help harmonise how GDPR is interpreted in these contexts, so organisations have the confidence to participate. Investment is needed—particularly to support less-resourced hospitals and regions. And high-value, high-visibility use cases should be prioritised to build momentum and demonstrate real-world impact.

Federated data offers Europe a way to make better use of its collective strengths—linking knowledge without compromising privacy. It’s not just a technical solution. It’s a practical model for collaboration that respects national sovereignty, builds public trust, and enables meaningful innovation. If approached thoughtfully, it can help the EU turn its fragmented health data landscape into a strategic advantage