Published on

Medical scribes document and transcribe physician interaction with patients during clinical appointments.  Does it matter if the scribe is human?

In recent years, the number of human medical scribes has grown throughout the healthcare industry.  Scribes accompany providers as they note patient history, features of a physical examination, lab results, create orders or prescriptions, and other interaction during an appointment.  The use of an effective scribe can increase the amount of time and attention that a patient receives from a provider as well as reduce the time a provider later spends completing documentation.

In the past, acting as a scribe offered experience to students considering medical school. The role evolved into a stand-alone job for those interested in clinical care and documentation.  With the advent of artificial intelligence (AI), and the rapid development of AI-driven applications, human medical scribes are likely to be replaced by algorithms in the not-too-distant future.

Human jobs that can be performed equally well by an algorithm will probably be outmoded by AI.  Subscribing to an AI service reduces the need for human hires, payment of benefits, and HR. There is no need for vacation time, and no issues with overwork.

Recent research published in the Journal of Informatics in Health and Biomedicine (JAMIA) compared automatic speech recognition (ASR) with natural language processing (NLP).  The study evaluated two AI products, Google ASR and Amazon ASR.  Both applications are trained on clinical language encounters. For the study, researchers provided data from 36 primary care visits to the bots. The authors then reviewed the results for word errors.

For these products at present, study authors found that non-lexical conversational sounds (NLCS) were a trip-up for ASR bots.  NLCS are the common sounds that all of us use to communicate, such as “uh-huh,” or any of the “Mm-hm” sounds that can be used to mean yes or no.  To a human medical scribe, the context is obvious; to an AI scribe, not so much.

Healthcare is an industry where detail is critical.  As noted in this research, “incorrect recognition of them [NLCS] could result in inaccuracies in clinical documentation and introduce new patient safety risks.”  When screening for symptoms and history, clarity is important, just as it is vital that a physician document what information was provided to a patient and to ensure that information is understood.

Late to the game, the US is now talking about regulation of AI.  The genie is long out of the bottle. It is understandable that AI stumbled out of the gate in this research.  But later versions of this powerful, generative technology will improve its ability to understand context, content, and non-lexical conversational sounds.  The only question now is how quickly that will happen and the type and extent of adverse events that will occur because of it.

Experienced legal service to Maryland patients injured by medical malpractice

Schochor, Staton, Goldberg, and Cardea, P.A. has more than 38 years of successful experience representing patients and their families injured through medical mistake.  If you suffer serious injury due to a serious mistake made by your provider, contact us or call 410-234-1000 to schedule a free consultation today.