EHDS as "Guinea Pig": Europe's Ambitious Plan for Health Data

Contents

On the occasion of the upcoming "Berlin Forum," the Association of Scientific Medical Societies (AWMF) highlighted the opportunities and challenges of two central European health projects at a press conference: the European Health Data Space (EHDS) and the new European Health Technology Assessment for medicines (EU-HTA). On the one hand, the data space is intended to improve direct patient care across borders (primary use) and, on the other hand, to enable the use of health data for research (secondary use).

Data Exchange for Holidaymakers and Co.

Prof. Rolf-Detlef Treede, President of the AWMF, outlined the relevance of the initiatives, which date back to the "Cross-border Health-Care-Act“ of 2011 (PDF). The EHDS is intended to improve cross-border care. "This is important in various scenarios that I outlined, not just on holiday, but for many who work across borders or simply for optimal care by centers that are perhaps close to the border," said Treede. However, he warned against poor implementation and inadequate data quality: "What is relevant for revenue today is based on data from the past. But if you only encode what is relevant for revenue now, you cannot tackle any new developments at all."

Furthermore, he fears that Germany is lagging behind in modernizing diagnostic data. "My fear is that we are currently asleep at the wheel regarding something that we ourselves helped to drive until 2019, namely the switch from diagnosis coding version 10 to version 11." In contrast to the old version, ICD-11 is a database that "is updated annually and always reflects the latest state of medical science."

"What's in it for the patient?"

According to Jana Hassel, spokesperson for digital policy at the Federal Working Group for Self-Help (BAG Selbsthilfe), the EHDS has the potential to significantly improve care, but the central question must always be: "What do patients actually get out of it?" This question is "not always put first. From our perspective, this has been a problem so far," said Hassel.

She sees concrete advantages, for example, that "through mass evaluation of data, we could also more quickly determine when risks, side effects, and interactions with medications occur." The data space is also promising for research into rare diseases. A prerequisite is high data quality, because "distorted or incomplete data significantly reduce the benefit. They can even be a cause of harm to patients." Furthermore, comprehensive protection against disadvantages is essential: "No one is willing to share their data voluntarily if they have to fear that sharing will lead to disadvantages for them."

Dr. Stephanie Weber from the Federal Institute for Drugs and Medical Devices (BfArM) explained the operational implementation. The EHDS is "the first data space of many that are being set up in Europe. In this respect, it is also a bit of a guinea pig, trying out how a large data space can function for Europe." A central challenge is the standardization of data across language and system boundaries.

Corona Pandemic as a Catalyst

The Corona pandemic also served as a catalyst for the project, during which the digital vaccination certificate showed that a Europe-wide technical infrastructure can function. Specifically, it concerns primary data categories such as e-prescriptions, patient summary records, imaging, laboratory findings, and discharge reports. The data for this is to come from the electronic patient record (ePA) and be exchanged between countries via a "National Contact Point eHealth." When asked when the more modern diagnosis classification ICD-11 would be introduced in Germany, she explained that this is a complex process because the old version ICD-10 is deeply embedded in billing and quality assurance systems. Therefore, a specific date cannot be given.

"We are practicing Europe"

According to AWMF Chairman Prof. Bernhard Wörmann, the launch of the European Health Technology Assessment (EU-HTA) procedure is also important. It is intended to address the unequal access to new medicines in the EU. "We are practicing Europe," said Wörmann. While the approval of medicines has long been centralized by the EMA, actual access in the member states is extremely heterogeneous. Specifically, this means that within the 27 EU states, there are countries that only have one-fifth of the approved preparations, according to Wörmann.

The new process is intended to centrally assess the benefit of a medicine compared to the current standard of therapy. The difficulty here is: "What is currently proving to be a significant hurdle is that the standard in the different European states is not the same." Nevertheless, Wörmann sees a great opportunity: If the assessment is available centrally, doctors and patients in countries with previously poor access can put pressure on their payers. "They have an instrument in their hands with which they can say, this is how it was assessed. And now you please take the next step. I consider that to be progress."

Europe's AI Answer?

The EHDS and other EU data spaces were also recently discussed during the "Dauphine Digital Days" in Paris. Experts from research, industry, and start-ups spoke about their visions, in which trustworthy data spaces in the EHDS and specialized AI systems are intended to provide the decisive competitive advantage. "The European game is about putting mind over matter, mathematics over hardware," explained Prof. Jakob Rehof, Director of the Lamarr Institute, right at the beginning. For the computer scientist, who himself worked at Microsoft for almost ten years, the future lies in "vertical AI systems" – highly specialized models for industries such as physics, medicine, or logistics. These are to be based on high-quality data shared in European data spaces like the EHDS.

Emmanuel Bacry, scientific director of the French Health Data Hub, described the EHDS as a "unique initiative." "All health data, regardless of whether it comes from private companies or public institutions, must be made available for research in the public interest," said Bacry. This is a crucial step to reduce dependence on non-European AI solutions.

"Building Data Spaces is Not a Given"

However, building such data spaces is not a given. Hubert Tardieu, board member of the cloud initiative Gaia-X, called for a change in strategy. The era of pure pilot projects is over. "We can no longer subsidize data spaces if we have no idea about their economic viability." For him, creating viable business models is the central challenge of the "second season" of European data spaces.

According to Selma Souihel, deputy director of the AI program at the French research institute Inria, sovereignty also means the ability to choose allies. For Anastasia Stasenko, co-founder of the AI startup Pleias, sovereignty must above all mean competitiveness. Her company consciously develops small, efficient language models trained on data that fully complies with copyright. She also referred to projects like the Spanish Salamandra model family, developed among others by the Barcelona Supercomputing Center. Salamandra, available on Hugging Face, was trained with European data and covers 35 languages, including Spanish, Catalan, Basque, Galician, and others.

Stasenko advocated for focusing on specialized models that excel at specific tasks, rather than copying huge all-rounders, "as others also do." She referred to the concept of the "cognitive core," also described by Andrej Karpathy, one of the founders of OpenAI: small, task-specific models that are massively expanded through synthetic data. "We must understand sovereign AI as competitive AI," she demanded.

"Traceability is the core of AI sovereignty," said Laurent Lafaye, co-founder of the data exchange platform Dawex. Only if it can be proven at any time which data and which software components are in an AI system can one speak of real control. This requires complete documentation of the entire supply chain. Europe's path to AI sovereignty is not possible by copying American strategies, but through its own approach based on quality, transparency, and cooperation. The goal, according to Souihel, is "to make Europe a place where AI is invented and not imported."

(mack)