In a busy hospital setting, some providers may opt to streamline the process of documenting patient encounters by using speech recognition software to dictate notes that are saved to an electronic health record (EHR) system. However, if left unchecked, this process may introduce serious errors into providers’ notes.
A recent study in the International Journal of Medical Informatics demonstrates this problem.
In the study, researchers looked at an emergency department in an academic teaching hospital with an average of 42,000 visitors per year. They took a random sample of notes dictated by physicians in the ED and evaluated them for the presence of eight different types of errors:
- annunciation errors
- added words
- deletions
- homonyms
- nonsense errors
- spelling errors, and
- suffix or dictionary errors.
The result: Researchers found an average of 1.3 errors per note. And almost 15% of them were designated as “critical” errors that could potentially affect patient care – or even cause harm to patients.
Annunciation errors were the most common errors caused by speech recognition software – they were the cause of over half (53.9%) the errors discovered. Deletions were the second most common error, found in 18% of notes. Added words were in 11.7% of notes, and nonsense errors were in 10.9% of the notes. Problems with homonyms occurred with 4.7% of the notes.
Spelling errors were in fewer than 1% of the notes, and there were no issues with suffix or dictionary errors.
Researchers weren’t able to determine if these errors played a role in any adverse events, but the mistakes were significant enough to cause confusion when reading the notes.
Why double-checking is crucial
According to a post on Skeptical Scalpel, a blog written by a practicing surgeon, these errors are common with speech recognition software – partially because few people go back and read the notes after they’ve been recorded to check for mistakes.
Because it’s important to keep speech recognition errors from affecting patient treatment, it’s key for providers to be more diligent about spotting any issues they create in documentation. It’s a good idea for them to scan notes immediately after they’ve been dictated to see if any glaring errors stand out.
In addition, the Joint Commission has a series of standards hospitals should follow regarding dictated notes. Most of the recommended protocols involve improving the quality of providers’ documentation through regular review and reinforcement of best practices.
Education and training are essential to making sure errors caused by speech recognition software become less common. Using any transcription problems as teaching experiences, not excuses for punishment, is also important.
