NHS watchdog warns AI scribes misdiagnose and incorrectly record drug names

AI Scribes Making Critical Errors

Healthwatch England, an NHS watchdog, has cautioned that AI technology used to transcribe patient consultations is prone to errors, including misidentifying diagnoses and drug names. These inaccuracies could endanger patients by leading to incorrect medical records and subsequent care issues.

Examples of Patient Harm

In one instance, an AI scribe incorrectly summarized a patient's MRI scan as showing demyelination, a serious nerve condition, which was later corrected to "null demyelination" after the patient queried it. Another case involved an AI scribe confusing a prescribed drug with a different medication of a similar name. These errors were identified by patients, not medical professionals.

Impact on Patient Records and Care

The watchdog highlighted that such mistakes, if not detected, could become permanent parts of a patient's medical history, potentially affecting future treatment. Healthwatch has received multiple reports from patients who have identified these errors when healthcare professionals did not.

NHS Rollout and Government Expectations

The NHS is rapidly deploying AI scribes, with the government's 10-year health plan for England anticipating the technology will reduce administrative burdens on staff. This initiative is central to the NHS's transition from an analog to a digital-based service.

Call for Clarity and Reporting Mechanisms

Healthwatch acknowledges that healthcare is not error-free but emphasizes the urgent need for clear processes for patients to report and correct mistakes made by AI scribing tools or the professionals using them. The Patients Association also stressed that trust in this technology requires good communication and partnership with patients, which are currently lacking.