The European Health Data Space (EHDS) promises to transform how health data is shared, accessed, and used across the EU. But turning this ambitious vision into operational reality won’t be easy. Henrik Matthies, Co-CEO and Co-Founder of Honic, shares why EHDS could become the world’s largest integrated health data space and what it will take for Europe to stay ahead in the global race for data-driven medicine.
EHDS entered into force in March 2025 and has since been awaited with great hope. Does the final legislation enable the harnessing of the full potential of data in science?
First, the EHDS has a bold vision: enabling a harmonized EU data space where interoperable institutions manage the compliant flow of health data for better care and research. The EHDS has the potential to establish the world’s largest integrated health data space and could become a true booster for data-driven medicine, fueled by near-time real-world data (RWD) research.
As always, the devil is in the details, and it is now up to each of the 27 EU member governments to translate this vision into operational practice. Since all 27 healthcare systems work completely differently, it is a true challenge to establish common standards, harmonize data flows, and reduce bureaucracy to a minimum.
Moreover, the EHDS legislation has not yet been fully harmonized with the EU AI Act, GDPR, and other regulations governing the generation and use of health data.
However, more digitally advanced EU nations like Portugal, Finland, and Estonia have shown for years that a harmonized health data exchange — in this case, ePrescription — is already possible.
Do you believe the EHDS has the potential to give European medical research a real competitive edge globally? If yes, what will it take to achieve that?
Yes, the potential is there! The EHDS vision is the right move for a more harmonized EU health system and market. The EHDS was designed in the face of the pandemic, when it became obvious that the status quo was inappropriate for managing such a crisis. Yet, in 2021, the rapid advances of artificial intelligence could not have been foreseen. For me, the make-or-break point for the EHDS will be whether technological advances like AI will outpace the implementation speed of the EHDS, which is planned to be in full force as late as 2031.
Regulators now need to “square the circle,” as we would put it in Germany: harmonize AI regulation with the EHDS and GDPR, ensure that bureaucracy does not stall the engine, uphold high EU data privacy and security standards, and uncover and harshly sanction misuse. Then the EHDS can harness its full potential for research.
The EHDS promises broad access to retrospective healthcare data. From your perspective, what new kinds of research questions will become possible once these large datasets are available? Would you call EHDS a “game-changer” in medical research or just a small step forward?
I do not have insights into all 27 healthcare systems and research communities, so I will mainly speak from a German perspective: to date, we lack the means (data access, data standards, infrastructure) to analyze the medical care status quo without major time lags and significant biases. All relevant data is there, but it is stored in silos, often generated without common semantic standards and syntax, and gated by various data holders who have little incentive to share their data. An interoperable infrastructure is lacking.
With the EHDS in place, the silos would vanish, data standards would be commonly used, interoperable infrastructures implemented, and we would be able to set up near-time dashboards for all kinds of diseases and indication areas. We would understand the quality of care as it is, as we could follow full patient journeys in inpatient and outpatient care, as well as rehab, elderly care, and many other current data silos.
Researchers could, for example, build representative control arms for clinical studies at scale, reducing necessary budgets while increasing the speed to market of new treatments.
I am sure we will identify many additional use cases that will boost medical research across the EU once health data is accessible and harmonized for research across several member states.
Can you share some examples or early use cases you’ve seen where access to large-scale health data has already changed how research is done or has had a big impact?
Several medical aspects of the COVID-19 pandemic remain uncovered to this day. But RWD may contain answers. Here is one: I helped build a digital-sovereign, compliant, and secure RWD research platform in Germany called Honic. In close cooperation with the German data privacy authority and patient organizations, large-scale RWD, especially from outpatient care, was and is being made accessible for research. Today, we have longitudinal RWD from more than 11 million Germans on the platform. By 2026, this number will rise to 40 million.
Our research team joined forces with several professors from LMU, one of Germany’s leading medical universities, to analyze German lab data on the Honic platform — specifically vitamin D levels — from a pre-pandemic cohort and a comparable pandemic cohort. Vitamin D levels dropped significantly during the pandemic, especially among elderly women. Usually, this type of research would have been conducted with a few hundred patients, raising many concerns about bias and representativeness. It would also have involved a lot of manual work aggregating data from various study centers, resulting in a 2–3 year project.
In our case, we analyzed data from 292,187 patients, and Nature Communications published the results less than 1.5 years after the project began, in early October 2025. The dataset's size and the speed with which it was published in a major global journal make this study a valuable reference and a strong motivator for further research on RWD.
Some countries, like Finland, are further ahead in the secondary use of data for research. Why do some countries outperform, while others struggle when it comes to building scalable research environments?
Research is most advanced in progressive, digitalized healthcare systems that focus on access to and quality of health data.
Every year, high-level delegations from Germany travel to the Nordics and other more digitally advanced healthcare systems to understand why the German system lags behind. They often return puzzled, summarizing that their European colleagues also “cook with water.” But somehow, other nations focus on implementing and iterating in an agile, learning system instead of endlessly discussing and problematizing the myriad of opportunities, which prevents fundamental changes altogether.
The most advanced digitalized healthcare systems today usually started with small budgets, like Estonia, established clear responsibilities within the system, and had rather simple healthcare systems to begin with, especially compared to Germany’s. The more legacy systems and fenced business models exist, the more responsibilities become diffused, and the harder it is to digitize. Digitalization changes roles and, as a result, business models, and these are the two most fundamental shifts a system can experience. In Germany, change is often perceived as a threat rather than as a necessary path to a better solution.
That is one of the reasons we decided to start Honic as a startup rather than a public institution, even though my co-founders and I had previously worked for the German government. In a startup, you can act fast, iterate, learn, and operate with a high degree of freedom. And the solution has to be lasting and continuously evolving. Otherwise, you risk your company’s existence.
Where do you see the biggest technical, legal, or cultural roadblocks when it comes to making health data available for research? Are scientists really avoiding data-driven research because they are afraid of GDPR, while data are stuck in silos or not interoperable?
Scientists usually apply what they have learned in their academic careers. And for the vast majority, only RCTs with clinical data or research with registry or claims data are pre-approved, which is desirable. Research with RWD at this depth and breadth is new. It requires pioneers to establish new best practices, experience both limitations and benefits, and understand where RWD is complementary or supplementary to established data pools such as clinical, registry, or claims data.
And, of course, it was extremely difficult to access RWD until recently. The vast majority of RWD research is consent-based, which is almost impossible to collect at scale in outpatient care settings. Thus, most RWD research has been based on inpatient data, limiting its scope. With Honic, we established a multi-layered data protection and security architecture that enables the use of health data for research without patient consent. Among many other technical and operational measures, all medical data is pseudonymized at the data holder level by an external trust center, and no data leaves our secure platform environment.
Many challenges remain, as described above. RWD seldom follows semantic standards, syntax is often broken and poorly documented, and data holders have few incentives to share their data with third parties.
Yet, since the EHDS has been passed, public discussions have shifted away from problems and challenges aimed at protecting the status quo and towards realizing that the new normal will be data sharing across silos and EU nations. Now, there is much more emphasis on smart implementation and integrating existing data pools and structures.
I believe that if scientists can access health data easily and scalably, and pioneers create first-best practices, we will see a wave of new RWD research projects with novel insights that established research data sources have not yet generated, mainly when used in isolation. One of the game-changers will be how well we can link, for example, RWD with claims data or inpatient with outpatient data.
The official timeline for the EHDS is ambitious. By 2027, National Health Data Access Bodies should be established, and by 2029, Patient Summaries should be available across all EU member states. Do you think it’s realistic?
It is ambitious, given the status quo and the significant variation among the 27 EU healthcare systems. Yet, 2029 seems rather unambitious when you confront it with technological advancements, especially in AI, in the USA and China.
I do not believe that all EU nations will succeed within the given timeline. Still, I would argue that the EHDS is already a success if the most advanced EU nations can exchange their data in a compliant and secure fashion and build a first common data space, even if it does not yet comprise all of Europe. Other nations will then have even more motivation to catch up, so as not to lose more research capacity to the more advanced countries.
Is the EHDS focusing on the right types of data?
I would rephrase that question more broadly: Does our current medical research — and, by extension, our medical system — focus on the right data points?
My estimation: a clear no. Most of our research focuses on “medical” data, i.e., data generated by medical professionals during care. That is an essential snapshot of a patient’s health, but we know that many other factors influence health, and that health parameters vary over days, months, cycles, and seasons. A patient’s zip code and socio-economic status, social environment (e.g., whether they are lonely), wearable data that monitor 24/7 rather than once per quarter, patient-reported outcome measures (PROMs), weather, and season all play a role. A lot of data can be interpreted medically, even if generated in non-medical contexts, such as social media interactions affecting mental health or search behavior patterns.
In many countries, these various “non-medical” data points cannot be linked to medical data at scale. I believe it would significantly improve research and care to get a holistic view of patient journeys, including relevant “non-medical” data, and to focus on the biggest levers, which may sometimes mean finding a companion for a patient before increasing a medication's dosage.
What does AI change in making health data accessible and usable for research? And does generative AI have a role to play in this context?
AI is evolving at such a pace that any answer can only be temporary. Now, AI can support research in finding patterns, extracting structured data from unstructured documents in some use cases, translating (e.g., language to code), and analyzing large amounts of data when results can be cross-checked. A game-changer would be to feed all types of health data into an AI to generate harmonized data that adheres to international standards. But while this may work well for syntax, it does not work reliably for semantics. Harmonization can only be as good as the generation and documentation of the original data.
I would question any AI result that I cannot reproduce or understand, or verify whether it represents the correct answer to my question.
That said, I believe AI will become a great toolbox for future research, among many other use cases, and we should always be open to exploring its best applications while upholding EU data privacy and security principles. I therefore strongly recommend restricting access to any health data to digital-sovereign EU infrastructures, ensuring that no data point leaves EU jurisdiction. Otherwise, we risk losing control over our most sensitive data once and for all. In these times, we should avoid making ourselves more dependent and vulnerable to blackmail. Our company, Honic, is built entirely on EU technology. Digital sovereignty is a major trust builder, enabling us to have health data from 40 million Germans on our research platform by 2026.
Dr. Henrik Matthies is the Co-CEO and Co-Founder of Honic, the leading German RWD research platform. Until 2021, he served as managing director of the health innovation hub, the digital health think tank of the Federal Ministry of Health in Germany. He is a serial entrepreneur, co-founded, among others, Mimi Hearing Technology, a digital health pioneer in Europe.
