In a column for NPR, Dr. Mara Gordon argues that while artificial intelligence scribes offer clinicians significant time savings, the technology introduces complex concerns regarding patient privacy and the potential for medical record errors.
Gordon, a family physician in Camden, N.J., notes that she is an eager adopter of the software, which allows her to maintain eye contact with patients rather than focusing on a computer screen.
The software utilizes automatic speech recognition to transcribe office visits, typically storing the data in a HIPAA-compliant cloud before generating a clinical note. Despite these safeguards, Gordon highlights that the technology is not infallible.
She recalls an instance where a scribe incorrectly summarized a patient’s diagnosis, claiming he had been misdiagnosed when he had actually been diagnosed with metastatic cancer.
“It looks really good because it’s really good at sounding like us, and it’s really good at pretending to be us,” said Dr. Eileen Kim, an assistant professor of internal medicine at the University of California San Diego. “But it’s still just pretending in a way. There’s no real brain.”
AI Notes May Cause Errors and Privacy Concerns
Kim warns that overburdened physicians may not have the time to adequately review these AI-generated notes, which can lead to the propagation of errors throughout a patient’s medical chart.
While most AI scribe services are designed to be HIPAA-compliant and encrypt data to protect against breaches, the software remains a point of tension for some patients. Gordon acknowledges that some individuals fear the technology could be used to identify them or that sensitive conversations are being recorded by third-party corporations.
Physicians Retain Legal Responsibility for Medical Records
Physicians remain legally and professionally responsible for the contents of the medical record. Gordon explains that she edits every AI-generated note before signing it, ensuring the clinical decision-making is accurate and transparent. Most software agreements reinforce this liability, placing the burden of fact-checking on the doctor regardless of the tool’s assistance. Beyond accuracy, health equity advocates have raised concerns that AI scribes primarily support English-speaking patients.
Research and the Future of AI
Some researchers view the vast collection of office-visit transcripts as “digital gold” that could eventually be used to analyze best practices in physician-patient communication. For now, however, Gordon clarifies that these recordings are not used for research purposes unless a patient is specifically enrolled in a study.
Patients who feel uncomfortable with AI in the exam room generally have the option to opt out, with doctors reverting to traditional manual note-taking.
Kim notes that while a patient might avoid an AI scribe, it is nearly impossible to opt out of AI in medicine entirely, as machine learning is increasingly integrated into billing, risk calculators, and early warning systems for conditions like sepsis.
“It’s just everywhere,” Kim told NPR. “It would be impossible to separate. We use it throughout the healthcare system.”
Common Questions About AI Medical Scribes
Can I refuse to have an AI scribe record my appointment?
Yes. If a patient does not want the software used, doctors can continue to take notes manually on a computer, which remains the standard practice for many providers.
Are these recordings stored on the doctor’s personal phone?
No. While some doctors use apps on their smartphones to capture the audio, the recordings are uploaded to a HIPAA-compliant cloud and are typically deleted by the service provider after a set period.
Does the AI replace the doctor’s judgment?
No. According to Gordon, doctors are required to read and sign the notes generated by the AI, and they remain responsible for all clinical decisions and the accuracy of the medical record.