Speech recognition errors occurred in 71 percent of emergency department notes, and 21.1 percent of notes with errors were judged as critical with potential implications for patient care says a recent study in the International Journal of Medical Informatics.

Investigators looked at a random sample of 100 dictated notes and found 128 errors or 1.3 errors per note.

More than half of the errors were ascribed to speaker mispronunciation. Although when I use speech recognition software, it sometimes does not accurately discern what I am clearly saying.

Other errors involved deleted and added words, nonsense, and homonyms.

An example of a nonsense error was “patient up been admitted for stable gait.”

Some of the critical errors (with possible interpretations) were as follows:

Cardiac exam is regular regular (irregular irregular)
Temperature 12.9 (102.9)
Exposure was a pap (bat) was found in room the family house
Cranial nerves II through XII intact, he is out of 5 motor strength (5 of 5 motor strength)
Pulling (pooling) of secretions

The authors concluded, “As this was a pilot study, we did not evaluate whether the errors were associated with actual adverse events.”

I have used speech recognition software for several years with occasional amusing results. I have seen numerous gaffes in hospital charts. No one proofreads dictated notes before electronically signing them.

Of the many such errors I have seen, I can’t recall a single one that led to an adverse patient outcome.

Luck? Not enough subjects in my experience (Type II error)? Or as I speculated in a blog post about electronic medical records five years ago, could it be that no one is reading most of these notes anyway?

Here’s what I said back then:

The ability to copy and paste coupled with the ease of dictation results in voluminous notes. As you may know, coding (directly linked to reimbursement) for visits is based on the extent of the care given. “If it’s not documented, it didn’t happen,” goes the saying. But now we have the inverse. It is so easy to document that notes are easily puffed up to “document” extensive encounters with every patient.

Because they are so long, they are difficult to read, and the black-on-white appearance of the words on so many screens causes the reader to skim over most of the note and go straight to the plan or recommendations.

I think no one is reading the notes.

“Skeptical Scalpel” is a surgeon who blogs at his self-titled site, Skeptical Scalpel.  

Image credit: Shutterstock.com