The NHS statistics watchdog has been asked to investigate Government data used to justify a nationwide AI rollout for GP appointments.
Scientists say NHS England has used an internal, unpublished evaluation to greenlight the flagship artificial intelligence programme without releasing its evidence for independent scrutiny.
They have now asked the Office for Statistics Regulation (OSR) to investigate whether ministers have breached official standards of transparency and trustworthiness.
The new AI system is to be used to prioritise and triage people seeking an urgent same-day GP appointment.
Instead of speaking to a receptionist, patients are instead asked to complete an AI-powered symptom questionnaire through the NHS App or a web link.
The software asks follow-up questions based on their answers before producing a summary for the duty GP, who reviews the case before the patient is contacted about the next step.
The use of this system is likely to reignite concerns over the growing use of AI in the NHS. Scientists have warned that this technology can produce dangerously convincing false information, overlook serious illnesses and amplify hidden biases in patient data.
Critics also fear older patients who are forgetful, live with dementia or are not confident using technology could be left behind as the NHS becomes dependent on AI and smartphone apps.
The current row centres on NHS England's widely reported July 3 announcement that it would roll out an AI triage tool through the NHS App to millions of GP patients.
As evidence the technology works, NHS England claimed an AI pilot at a Sussex GP practice had cut the number of patients queuing on the phone by 29 per cent, helping end the daily 8am scramble for appointments.
The figure was widely press-released and used to back a nationwide rollout across England.
Oxford University academics Prof Carl Heneghan and Dr Tom Jefferson say the scientific study cited by NHS England does not contain the evidence for that claim and that any technology affecting patient care must be backed by rigorous, publicly available evidence before it is rolled out nationwide.
LATEST DEVELOPMENTS:
- Cambridge professor RESIGNS after investigation launched into accusations of plagiarism
- Andy Burnham urged to defend Christian freedom in 'defining test' for new PM
- Martin O'Neill in hospital after Celtic boss undergoes surgery
Instead, they say the paper compared the AI's clinical decisions with those made by GPs and measured patient satisfaction.
It contains no analysis of phone queues, call waiting times or changes in telephone demand.
NHS England has said the 29 per cent wait reduction figure was based on an unpublished internal NHS England evaluation of a GP practice and was based on GP telephone data.
The Oxford academics say this means outside experts cannot check the accuracy of the headline claim.
In a formal complaint sent to Ed Humpherson, Director General for the NHS statistics watchdog, the Office for Statistics Regulation, Prof Heneghan and Dr Jefferson say this is unacceptable when a statistic is being used to justify a major national policy.
Writing on their Substack ‘Trust the Evidence’ they state: "If a headline statistic is sufficiently robust to justify a national policy announcement, shouldn’t the evaluation behind it be public? Without the underlying analysis, independent scrutiny is impossible."
And they add: "This is no longer simply about one AI pilot. It is about the standards of evidence that should accompany public claims made in support of national policy."
Their complaint asks the Office for Statistics Regulation to examine whether NHS England's use of an unpublished evaluation complies with the Code of Practice for Statistics, which requires public bodies to be transparent about the evidence behind numerical claims.
The academics have also asked NHS England to publish the full evaluation report.
The statistics regulator OSR and NHS England have been approached for comment.
Our Standards: The GB News Editorial Charter
