Your doctor’s AI notetaker may be making things up, Ontario audit finds

In a significant revelation for the healthcare sector, an audit conducted by Ontario's auditor general has raised serious concerns about the reliability of AI medical scribes. These tools have been increasingly adopted by overwhelmed physicians to help document patient interactions, diagnoses, and treatment plans. However, the findings suggest that these AI systems are frequently producing inaccurate or incomplete records, which could potentially jeopardize patient safety. The audit, detailed in a report on the Use of Artificial Intelligence within Ontario's Government, evaluated transcription tests from two simulated doctor-patient interactions across 20 AI scribe vendors approved by the provincial government. Alarmingly, all vendors exhibited some form of inaccuracy or incompleteness during these basic assessments. The results highlighted that nine of the vendors generated fabricated patient information, twelve recorded details incorrectly, and seventeen failed to capture critical aspects of mental health discussions. Specific examples from the audit illustrate the possible ramifications of these errors. Instances were noted where AI scribes fabricated referrals for blood tests or therapy that never occurred, misidentified prescription medications, and neglected crucial details regarding mental health issues. Such inaccuracies could lead to inadequate or harmful treatment plans, ultimately affecting patient outcomes and care quality.