In recent years, speech recognition technology has been hailed as a way to eliminate transcription costs. The marketing hype claims that it is 95% accurate. Which means that, if you are using speech recognition for your dictation, there is a good chance that five out of every hundred medications and their dosages could appear incorrectly in your patient's medical records. Is that a statistic you can be comfortable with?

While speech recognition enthusiasts think they are improving the situation by taking the human factor out of the dictation/transcription process, oftentimes they are only making matters worse. It's true that some doctors can keep their minds focused on doing one thing at a time (which helps when you’re using speech recognition). Most of them, however, are trying to do three things at the same time while fielding interruptions from support staff. Concentration is at a premium -- as evidenced by these verbatim quotes from dictating physicians:

“Impression colon common migraine headache plan colon patient to common migraine headache patient to plan colon patient to try Midrin and Fioricet on a p.r.n. basis.”

“Past history the usual up to the usual.”

Is that how you want your reports to read?

The following two examples are taken from dictation by a physician who used to be the chairman of his hospital’s Medical Records Committee:

“The patient is a 30 year old right handed married mother of one child age four who is a former office manager in a family business in steel fabrication and who was placed on permanent disability in February of 1997.”

“The patient is a 35 year old right handed woman who lives with a partner who has been pregnant on two occasions and who has had two abortions and who did take birth control pills which she discontinued about 10 years ago because of profuse bleeding.”

Misplaced modifiers, dangling participles, blazing contradictions and run-on sentences! These are the stock in trade of most dictating physicians. Add in foreign accents, incorrect lab values, rotated right/left designations, wrong genders and you start to encounter problems that speech recognition cannot and will not solve.

When doctors with impaired language skills cannot properly communicate their thoughts in English, they end up providing substandard data which can endanger a patient's life. Most people who examine medical records are looking at printed text and not listening to sound. They cannot imagine how much editing is necessary to clean up dictated text and correct a physician’s mistakes before the data gets into the chart!

Speech recognition doesn't do that.  The software does absolutely nothing to improve the language skills of the person using it.