Palantir’s NHS efficiency claim draws fresh clinical scrutiny
Prof Caroline Mitchell says Palantir tools built for fragmented US reimbursement may not fit the NHS, after NHS England deleted a claim about 137 trusts using them.
By Amara Osei
4 min read

What's News
- A November 12 report said NHS England deleted a “misleading” claim involving 137 trusts and Palantir AI tools.
- A local NHS consultant used seven bolt-on software systems for one consultation, including a billing-focused system.
- A senior US hospital administrator said most US health data systems primarily record care for reimbursement.
- Mitchell cited Emis as a UK primary-care system developed with GP leadership.
- Mitchell warned that algorithms developed for the US healthcare system may be immature in the NHS context.
A November 12 report cited by Prof Caroline Mitchell said NHS England had deleted a “misleading” claim that 137 trusts were benefiting from Palantir’s AI tools. In the letter, Mitchell said the episode suggests NHS England may have overstated Palantir’s ability to improve NHS efficiency. She said few frontline clinicians would be surprised.
Prof Caroline Mitchell, a GP in Sheffield and professor of general practice research at Keele University, wrote the letter from a frontline clinical perspective. Her criticism centers on system fit, algorithmic maturity and the aims built into health records.
What does the efficiency dispute turn on?
Mitchell said global digital policy often assumes software designed for the fragmented, insurance-based US healthcare system can be adapted to a universal public service elsewhere. She called the concern about algorithms legitimate because systems developed for the US market may still be immature in the NHS context.
Her objection concerned the design assumptions behind imported tools, not simply the wording of the deleted benefit claim. Her letter did not cite a Palantir contract value, NHS savings target or rollout timetable. She joined system compatibility with algorithmic maturity, two distinct tests for any imported AI tool.
The concern also changes what an efficiency claim must demonstrate. In Mitchell’s account, a US product may achieve results in a reimbursement-led insurance system without delivering the same gains in a universal service built around lifelong records and primary care. That was the comparison Mitchell asked NHS policymakers to make.
Why does clinical workflow matter?
Mitchell cited a consultant in a local NHS trust who had to work across seven bolt-on software systems for a single consultation. One of those systems focused on billing functions, according to her account. She used the example to place the efficiency debate inside the clinician’s working day rather than around a broad promise about better technology.
Seven systems for one consultation showed why imported software cannot be judged only by its own capabilities. It must also work with the existing patchwork of clinical records and billing functions, according to Mitchell’s argument. She did not quantify time spent on the seven systems or calculate savings from replacing any of them.
How do US and NHS records differ?
A senior US hospital administrator told Mitchell that the primary purpose of most US health data systems is to record care for reimbursement. By contrast, he said, “NHS general practice records are designed to support lifelong patient care and safe prescribing.”
He was also struck by the NHS app’s focus on coded healthcare rather than how much of an insurance allowance had been used up that year. The administrator’s comparison identified two recordkeeping priorities: payment processing in the US and continuing care in the NHS.
For an AI efficiency claim, that difference is material because algorithms must operate on records and workflows shaped by each system’s underlying purpose. Mitchell’s point was that the efficiency case cannot be separated from the records and workflows the tool is meant to improve.
Why did Mitchell cite Emis?
Mitchell asked why the NHS needs a US firm to manage its data when the UK has a strong record of healthcare-software innovation. She cited Emis, developed with GP leadership, which she said transformed primary care IT decades ago.
Her conclusion was blunt: “Software built for an insurance market may simply not be the best fit for universal healthcare systems.” The Emis example made her argument about comparable operating environments rather than national origins. In her comparison, Emis was built for primary care, while Palantir’s tools come from a market organized around insurance reimbursement and fragmented provider systems.
What would settle the efficiency claim?
Mitchell’s criticism leaves a practical test for NHS England and Palantir: show how imported AI tools and data systems work within NHS records built for lifelong care and safe prescribing. Any efficiency case must also address whether algorithms developed for the US market are mature enough for decisions in an NHS context, according to her letter.
The 137-trust claim was presented as evidence of broad benefits, but the deletion and Mitchell’s critique leave those benefits unproved in the terms she identifies. Unless NHS England can show that workflows, reimbursement assumptions and algorithmic maturity do not defeat the claimed efficiency gains, the Palantir case remains narrower than the deleted benefit claim implied.
Source: theguardian.com
More from Amara Osei
Show full bio
Senior reporter covering companies, deals and strategy in entrepreneurship, startups and venture funding at Business Bearings.
2,336 articles