AI in the NHS must support doctors

Artificial intelligence is moving rapidly into everyday clinical practice. From drafting discharge summaries to summarising patient records, AI tools have the potential to save time in a health service under relentless pressure.

But as the Government pushes the NHS to accelerate its shift from analogue to digital care, a big question remains largely unanswered: when AI-assisted systems get something wrong, who ultimately carries responsibility?

This question is being pulled into sharper focus as the debate grows around the plan to roll out the use of AI-assisted discharge summaries following a trial last year at Chelsea and Westminster NHS Foundation Trust.

There appears to be many potential benefits: the manual processes for discharge documentation are time-intensive, often causing delays in sending summaries to GPs. In addition, time-pressured doctors can sometimes omit key information leading to failures in continuity of care between hospitals and primary care.

Yet the safety and regulatory questions surrounding the introduction of alternative AI systems remain largely unresolved.

A report earlier this year from the Royal College of Physicians (RCP) warns that while doctors broadly support the introduction of AI tools in healthcare, the NHS lacks the digital infrastructure needed to deploy them safely and effectively.

According to the RCP, 70% of physicians are supportive of AI being widely implemented in the NHS, but more than two thirds believe the health service currently lacks the digital foundations required.

The view from frontline clinicians is similarly nuanced.

In a recent survey conducted by the Medical and Dental Defence Union of Scotland (MDDUS) involving 1,175 doctors across the UK, just over half said they were not confident that the government’s proposed shift from analogue to digital healthcare could realistically be achieved.

Doctors are not opposed to digital innovation – nearly half of respondents believed digitally delivered care could have a positive impact on their day-to-day practice, particularly through more efficient ways of working, reduced administrative burden and improved access to care for patients.

But many reported being concerned that poorly implemented digital tools could create new pressures rather than relieve existing ones.

Among doctors who believed digital care could negatively affect their practice, the most frequently cited concerns were unrealistic expectations about the speed of care and increased workload from follow-up queries, both highlighted by 77% of respondents. A further 74% cited concerns about inaccuracies in AI-generated advice.

In other words, the concern is not simply that AI might get things wrong. It is that digital systems could create additional layers of work while leaving clinicians responsible for checking and correcting their outputs.

This is important because, regardless of how sophisticated the technology becomes, the professional responsibility for clinical decisions still rests with the doctor.

In our medico-legal practice we see the consequences when documentation is incomplete, inaccurate or misunderstood. Discharge summaries are a good example. They influence what happens next for a patient and often become the definitive record of a hospital admission.

If an AI tool helps produce that documentation more quickly, it may bring welcome efficiencies. But if key clinical details are wrong or omitted, the algorithm will not be held accountable. Instead, the accountability will rest with the clinician whose name is attached to the record.

This is precisely why the current pace of technological adoption must be matched by equal attention to governance and regulation.

The RCP is right to call for clearer national guidance on safe and approved AI tools and for robust regulatory frameworks to support their introduction. Clinicians also need access to proper training and clarity about how these systems should be used in practice.

Without that framework, there is a risk that adoption will outpace accountability.

Artificial intelligence has real potential to support clinicians and improve patient care. Used well, it could help address some of the administrative burden that currently consumes so much clinical time.

But digital transformation cannot simply be measured by how quickly new technologies are introduced. It needs to be judged on whether safety, accountability and professional responsibility keep pace with that change.

AI should support clinical judgement, not transfer new risks onto the doctors expected to use it.

---

This page was correct at the time of publication. Any guidance is intended as general guidance for members only. If you are a member and need specific advice relating to your own circumstances, please contact one of our advisers.