"Mmm hmm." "Uh-uh." When your physician answers your questions with those phrases, you understand. But the artificial intelligence listening to your conversation might not.
That's according to a new study published in the Journal of the American Medical Informatics Association.
The researchers analyzed 36 primary care encounters that used Google Speech-to-Text Clinical Conversation and Amazon Transcribe Medical. Nonlexical conversational sounds were recognized only 40.8 percent of the time by the Google software and 57.2 percent by the Amazon program, the Jan. 23 study found.
While nonlexical conversational sounds only accounted for 2.4 percent of the total words used, "incorrect recognition of them could result in inaccuracies in clinical documentation and introduce new patient safety risks," the authors wrote.
The researchers are affiliated with UC Irvine (Calif.), California University of Science and Medicine in Colton, University of Maryland in Baltimore, University of North Carolina at Chapel Hill, and UC San Diego in La Jolla, Calif.