Twenty-five years inside European digital health, and the one problem underneath all of it.

For most of my working life I have been trying to get health systems to talk to each other. That sentence would have described the job in 1999, and it would have described it in 2015. The job stayed the same, and the layer it sits on changed. Making systems talk to each other, interoperability, was the first twenty years. Deciding who controls what they say is the work now. “Here,” for me, is that shift. It never felt like a career. It felt like one problem slowly revealing how deep it actually went.

I came up through standards. I sat on the board of MedCom, Denmark’s national health-data network, from 1997 to 2009, the years when a small country quietly became the place where digital health messages actually flowed, at national scale, between GPs, hospitals, pharmacies and laboratories, every day, without anyone thinking it remarkable. Later, from 2011 to 2018, I was Vice Chair of the Continua Health Alliance in Europe, working on the interoperability stack for connected personal health devices: the unglamorous business of agreeing how a blood-pressure cuff hands its reading to a record without either side needing to know the other in advance.

Two decades of that teaches you something the field is still slow to admit. You can make any two systems exchange a message and still leave the person whose body produced the data with no say in where it goes. Interoperability solves the plumbing. It does not touch the question of who the plumbing serves. We built schema after schema, profile after profile, and every one moved data more smoothly between institutions while doing nothing to move control toward the citizen. MedCom, Continua and the standards a generation of us spent careers on gave Europe a connected system that the patient still stands outside of.

That is the lesson I carry out of the standards era, and it is the spine of everything since.

Denmark went further on adoption than almost anywhere. Its standards were no cleverer than anyone else's. It went further because it built the trust and the institutions around the standards: the connective tissue that turns a schema into a service people actually use. That distinction, between the standard and the trust the standard needs, became portable, and I spent years carrying it.

I took it to the Gulf, where health systems are being built greenfield and you get the rare luxury of designing without forty years of legacy to unwind. In Saudi Arabia and Qatar the question is how to architect well from the start, with nothing to retrofit. And I took it to Southeast Asia, where I am now based, in Bangkok, where deployment rather than design is the real examination. A model that works on a slide has to survive contact with a different language, a different workforce, a different idea of what a clinic even is. European rails, Asian deployment, Gulf scale. Each context taught the same thing in a new accent: the schema travels easily; the trust does not travel at all unless you build it deliberately, locally, every single time.

Which brings me to now, and to two pieces of European law arriving at once. The European Health Data Space is moving into implementation, and the eIDAS 2.0 digital identity wallet must be in every citizen’s hands by the end of 2026. Together they are meant to give Europeans real control over their health data: to see it, move it across borders, and decide whether it is used for research and for training algorithms.

The trap national systems are walking into is one of sequencing. They are building the consent model for primary care first, because that is the urgent clinical need. And that consent topology, poured early, quietly decides what secondary use is ever possible. The hard problem is met first, at the moment you design primary-care consent, by people who believe they are only solving the clinic. Build the topology coarse, and you cannot later give a citizen a granular, revocable, cross-border say over research and AI without tearing up the layer you just laid.

I did not arrive at that by reading the regulation. I arrived at it from the work. I authored the consent and governance thinking out of CRANE (GA 965277), a pre-commercial procurement that put citizen-controlled data at the centre across three rural regions in Sweden, Spain and Norway, and out of the CronIQUE PPI that carries it forward. CRANE’s blunt finding was that the citizen-side layer, the trusted place where consent is held and control is exercised, can be built and validated, and that nobody currently funds or buys it. And ROSIA (GA 101017606) has just delivered: a digital rehabilitation model proven to run across three different national health systems at once, which is the proof that none of this is theoretical. Across all of it sits the same gap. The layer the citizen actually touches is the one no one has been asked to own.

So here is where the line lands. For twenty years the load-bearing problem was interoperability, making systems talk to each other. That problem is, broadly, solved. The standards exist, the rails are laid, the messages flow. The load-bearing problem for the next decade is sovereignty: deciding who controls what those systems say about a person, and building the layer through which they decide it. Patient-data sovereignty cannot be bolted on when the research requests start arriving. It is the layer the whole next phase rests on, and it is being poured, well or badly, right now, mostly by people who have not noticed they are pouring it.

That is what I mean by “here.” I stayed in the same field while it grew deeper. The work was always the same problem. It just took twenty-five years to reach the part that was underneath it the whole time.