A database of officer misconduct records has gone dark in Washington. A survey measuring violence against children has been quietly withdrawn. Real-time pollution readings from American embassies abroad have stopped updating. None of these losses arrived with an announcement. They accumulated, one deletion at a time, until a group of former federal data scientists built a tracker simply to establish how much had disappeared. Their count now stands at 28 datasets deleted outright and 338 more altered, with health information making up roughly two out of every five entries. Surveillance systems tracking HIV infections, overdose deaths and violence against women have been stripped of questions on gender identity or stopped updating altogether, casualties of an administration determined to erase findings that sit awkwardly with its priorities.

It is tempting for British readers to file this under American exceptionalism, another instance of a political culture willing to bend institutions to its will. That would be a mistake. The lesson from Washington is not about America's particular pathologies. It is about what happens to a health system's institutional memory once the infrastructure holding its data sits close enough to political and commercial pressure that priorities, rather than clinical need, can determine what gets measured, retained or quietly dropped. That question is live in England right now, and it has a name: the Federated Data Platform.

The FDP, delivered by a consortium led by Palantir under a contract that could reach £330 million, is fast becoming the central nervous system through which NHS trusts manage beds, waiting lists, staffing and discharge. NHS England insists the arrangement is watertight, that Palantir is merely a processor with no rights over the data itself and no ability to repurpose it. Ministers have repeated those assurances at the despatch box. Yet the platform has drawn sustained parliamentary criticism precisely because a single vendor, chaired by a figure openly sceptical of the NHS's existence, now sits inside the plumbing of a public health system with 57 million patients. The Health and Social Care Committee has called for the contract to be scrapped when its break clause arrives in 2027. The National Data Guardian has warned about vendor lock-in. None of this proves that data will be manipulated for political convenience. It does show how thin the line can be between operational efficiency and dependency on an entity whose interests are not simply the NHS's own.

The American episode illustrates why that distinction matters more than it might appear. Datasets do not need to be deleted outright to lose their value. They can be narrowed, definitions can shift, and questions that produce inconvenient findings can be dropped from the next survey cycle. Britain has its own quieter version of this problem already. The Cass Review found that six of England's seven adult gender clinics refused for years to share data that would have allowed clinicians to track outcomes for thousands of young patients, a gap serious enough that ministers have had to legislate this year to compel the information into a formal linkage study. The cause was different from Washington's, rooted in institutional reluctance rather than executive order, but the consequence was the same: a health system unable to learn from its own history because the record was incomplete.

What protects the NHS from drifting toward the American pattern is not technology but structure. Bodies such as the ONS and UKHSA operate under statutory independence that insulates their data functions from ministerial or commercial direction in a way that a subscription contract with a private American firm simply does not replicate. That distinction ought to be central to any decision on renewing or ending the FDP arrangement, rather than a footnote to conversations about cost and functionality. Preserving public trust in health data is not a matter of reassurance from ministers or suppliers. It rests on who controls the infrastructure long after the political weather has changed, and whether that control is answerable to the public rather than to a contract renewal date.

The datasets vanishing from federal websites this year were not lost to negligence. They were the predictable output of a system where political will can reach directly into the machinery that decides what a nation is allowed to know about its own health. The NHS has time to decide whether its own machinery is built to resist the same pressure, but not indefinitely.