Healthwatch England has raised concerns about the safety implications of artificial intelligence scribing technology being rapidly deployed across the NHS. The watchdog documented multiple instances where AI tools made significant errors in transcribing patient consultations, including misidentifying medications, incorrectly recording diagnoses, and omitting important medical instructions.
One prominent case involved a patient who received an incorrect diagnosis of demyelination when the AI scribe failed to properly capture the results of an MRI scan, ultimately recording “null demyelination” instead. The error caused considerable distress to the patient, who is an NHS health professional, and was only corrected after she questioned the record. In another instance, an AI system confused a prescribed medication with a similarly named drug, an error caught by the patient rather than the healthcare provider. A third case saw an AI-generated summary fail to record instructions for the patient to obtain a repeat prescription for migraine medication, potentially leaving them unable to access needed treatment.
According to Healthwatch, patients have detected numerous errors that healthcare professionals missed, raising concerns that inaccuracies could persist in medical records if patients do not catch and report them. The organization expressed particular concern that the Medicines and Healthcare products Regulatory Agency has decided not to classify AI scribes as medical devices, meaning there will be no nationwide regulatory oversight to ensure their safety and effectiveness. Currently, 27 different AI scribes are in use across GP practices and hospitals in England.
Experts have noted additional challenges with the technology. Researchers found that errors are more likely when consultations involve multiple people, patients with complex medical histories, or those whose first language is not English. Some have also raised concerns about “hallucinations,” where AI systems refer to medications or topics that were never mentioned during the consultation. Despite these issues, more than half of surveyed GPs believed their AI records were more accurate than those they produced themselves, suggesting varied perspectives on the technology’s reliability among healthcare providers.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI