Yvette Cooper, Secretary of State for Health and Social Care of the United Kingdom

A hospital trust in the Midlands tells a freedom of information request that the Federated Data Platform is not live and that it has no intention of rolling it out. On the national register, the same trust appears among the organisations realising benefits from the system. Nobody need have acted in bad faith. The entry reflects a definition, and the definition is the story.

The platform, supplied by Palantir under a £330m contract that began in 2024, is meant to join up operational data held in separate systems. Trusts can then validate waiting lists, schedule theatres, plan discharges and track cancer pathways. More than 100 freedom of information requests found that 29 trusts recorded as realising benefits said they were not using the system or its applications. NHS England says the listing includes trusts that use the platform for data submission and cost calculation, and that it has said so before. Under that reading the register is accurate. It can also mislead anyone who takes "benefits" to mean better care for patients.

The timing gives the problem weight. MPs, including the chairs of two Commons committees, are pressing the health secretary to activate the break clause in 2027. The vendor is a contentious supplier for reasons that have little to do with hospital performance, among them its defence and security work and its founder's public views on democracy. It has hired a former Labour deputy leader as a lobbyist and is writing to critics individually. In that climate the operational evidence has to carry the whole decision. At present it cannot.

The published uptake figures show how narrow the operational footprint is. NHS England reports 142 trusts live and 137 reporting benefits. Yet the patient-facing products are credited to far fewer organisations: 40 trusts for theatre scheduling, 37 for inpatient waiting list management, 31 for referral to treatment validation, 30 for discharge planning, 16 for outpatient validation and five for cancer pathways. Whatever the platform has achieved, it has achieved in a minority of the estate. The headline count describes connection. It says little about use.

The reported results deserve a fair reading. Waiting list validation is unglamorous and valuable. Removing patients who have died, who appear twice, or who were treated elsewhere gives clinicians an honest list and shows them who has waited longest. The referral to treatment tool has reviewed more than 5.7 million records and recorded over 1.1 million removals, and the outpatient tool a further 218,756. The discharge tool is credited with a 14.3 per cent fall in delay days for long-stay patients. Delayed discharge is one of the persistent causes of blocked beds, so if that figure holds it matters to a service whose pressure is felt in corridors and ambulance queues.

The difficulty is that these are before-and-after comparisons. NHS England concedes it cannot draw conclusions about cause and effect, and it accepted this summer that the 117,539 additional theatre patients could not be attributed to the platform. The Office for Statistics Regulation has examined that claim and is now taking the benefits listing forward through its casework function. There is a second consideration. A validated waiting list is a more accurate list, and it is also a shorter one. A government judged on waiting time targets has an obvious interest in tools that reduce the reported number, which is a good reason for scrutiny that does not depend on the department's own analysis.

The experience of trusts adds another layer. Kent and Medway told the requesters it had been focused on other priorities. For a trust managing financial constraint, industrial relations and winter planning, that is a rational answer. Adopting a data platform takes analyst time, integration work and clinical engagement, all of which are scarce. A national programme that treats sign-up and technical connection as success misreads how change happens in provider organisations. Some 170 trusts have signed a memorandum of understanding and 142 are live, but the products that touch patients are used in a smaller number. Trust leaders should expect to be asked precisely which tools they use, and they should answer precisely, because inflated returns tend to resurface through freedom of information requests.

Policymakers have a short list of tasks if the break clause decision is to rest on evidence. They need an independent evaluation with comparison trusts, so that improvements can be separated from the underlying recovery in performance. They need product-level adoption published by trust. They need a definition of benefit that separates analytical back-office use from clinical operations, and the register should follow it. The vendor says concerns about lock-in are myths. That is a testable claim, and publishing the exit terms and migration costs would test it. Critics who want a UK-owned supplier need to show that one could deliver comparable functionality at comparable scale within a realistic timetable.

Health-tech and life sciences companies are watching closely. The evidence standard set in this dispute will apply to every supplier that follows, and firms with robust outcome data have reason to welcome a stricter regime. Patients have the least information of anyone. They are asked to accept that their records will be joined up on a platform whose results are asserted more often than demonstrated.

The platform may turn out to be useful, and parts of the evidence suggest it could be. A 2027 break clause and a regulator already examining the figures leave ministers little time to commission the evaluation that the benefits register failed to provide. If the system works, a clean assessment will say so in terms critics find hard to dismiss. If it does not, the state needs to know before the window closes.