The NHS has a thirty-year habit that no strategy document has yet broken. When a data problem becomes visible enough to demand a response, the response is a platform. A new platform for pathology. A new platform for population health. A new platform for analytics. The platforms arrive, the problems persist, and eventually someone commissions another strategy that recommends another platform.
This is not a technology failure. It is a governance failure dressed as a technology decision. And the evidence that it will continue to fail is not theoretical. The Nordic countries ran the experiment at scale, spent billions, and produced the same result.
The wrong question, at great expense
Norway, Sweden, Finland, and Denmark each concluded, at different points over the past decade, that the route to connected health data was a single electronic health record system rolled out across the entire healthcare system. One system, one data model, one version of what a patient record looks like. The logic was sound on paper. The results were not. Costs escalated beyond projections. Clinicians were dissatisfied. Patient safety did not improve. Several countries are now locked into systems they cannot operate effectively and cannot afford to leave.
The failure was not in the ambition. Connecting health data across a national system is the right ambition. The failure was in the question. Interoperability and uniformity are not the same thing. A psychiatrist, an emergency physician, and a social worker will all interact with the same patient record, but they work in entirely different ways, under entirely different time pressures, with entirely different definitions of what a good outcome looks like. Forcing them onto a single system does not connect their data. It forces a compromise that serves none of them adequately, and produces resistance that ultimately undermines the entire programme.
The question England keeps asking is: which platform? The question worth asking is: what does the foundation need to be, so that any platform sitting on top of it produces data the rest of the system can use?
The foundation that was never built
The NHS does not have a data architecture. It has a collection of systems, each procured to solve a specific problem, each with its own data model, its own clinical terminology standards, its own interpretation of what a patient record contains and how it should be structured. None of them were designed to share a common semantic foundation with each other.
Pathology, the discipline responsible for over a billion laboratory tests a year in England, still routes its results using a messaging standard originally designed in 1987 for United Nations trade documentation. The clinical coding system that standard depends on was deprecated in 2019, when the rest of the NHS moved to a modern international terminology. Pathology had not moved, because the messaging layer beneath it was physically incapable of carrying the new codes. When COVID required new diagnostic codes at speed, there was no mechanism to create them within the existing infrastructure. The workaround was to route descriptions through a free-text field in a format never designed to carry clinical meaning. A billion-test-a-year clinical discipline routed around its own plumbing in a national emergency.
That is one discipline. The pattern repeats across imaging, prescribing, referrals, and community care. Each has accumulated its own standards, its own supplier relationships, its own reasons why migration is complicated. NHS England's own flagship analytics infrastructure, the Common Data Model that defines how data is structured and the Federated Data Platform that processes it, are reported not to be semantically aligned with each other. If accurate, this means the organisation responsible for building a national data architecture does not yet share a common data language within its own estate.
That is the baseline from which a national programme is supposed to begin.
What infrastructure actually requires
NHS England has articulated something that is architecturally correct: data should be treated as national infrastructure, existing independently of any specific application or platform. The data is not the platform. The platform is a consumer of the data. If the data is consistent, portable, and semantically coherent, any platform sitting on top of it becomes replaceable. Suppliers lose the ability to create lock-in through proprietary data models. The NHS gains the ability to procure on merit rather than on switching costs.
This is the right framing. It is also a decade of hard work from being a reality.
What makes it hard is not the technology. The modern international standards for health data exchange and clinical terminology are mature, widely adopted, and internationally maintained. The technical building blocks are available. What is missing is the governance to impose them, and the political durability to maintain that imposition across supplier contracts, procurement cycles, and ministerial reshuffles.
Infrastructure requires someone with authority to define the standard, enforce adoption across every supplier in the market, resolve the data controller and data processor questions that arise when patient data moves between organisations, and hold that line regardless of which political priority arrives next. It requires that authority to persist not for one Parliament but for several. You are not implementing software. You are destabilising an ecosystem built over decades by multiple competing interests, and the stabilisation work is measured in years, not sprint cycles.
Ten Health Secretaries in sixteen years have not provided that continuity. Each arrived with a platform. None stayed long enough to build the foundation the platform was supposed to sit on.
The question worth asking
There is a version of this that gets fixed. It requires treating the National Data Architecture as infrastructure in the same sense that roads and power grids are infrastructure: built once, maintained permanently, not reinvented by each incoming administration. It requires common data standards enforced by contract rather than encouraged by guidance. It requires the data models underpinning NHS England's own analytical estate to be aligned before anything further is built on top of them.
The political risk sitting above all of this is real. Reform UK's leadership has said explicitly that they are open to replacing NHS funding through general taxation with an insurance-based model. Whatever one thinks of that as a healthcare proposition, its implications for data governance are severe. Insurance-based systems fragment data ownership across payers, each of whom becomes a data controller with commercial incentives to retain rather than share patient information. The United States has spent decades and considerable sums on federal interoperability mandates and still does not have what a properly governed national data architecture requires. Moving toward that model mid-programme does not pause the infrastructure work. It makes it structurally impossible.
Infrastructure thinking is incompatible with a four-year political cycle. That is the lesson the Nordics learned at great expense. We have not yet decided whether to learn it differently, or to spend the same money discovering the same thing.
You cannot buy your way out of a foundational problem. You also cannot govern your way out of one if the people governing change every eighteen months.