Getting More From a Voice Recognition System
Using key words increases the accuracy of voice transcription technology and makes charting more cost-effective.
Fam Pract Manag. 2001 Mar;8(3):55-56.
Voice recognition computer technology has been a benefit for a few and a curse for many, but I’m living proof that physicians can successfully incorporate voice recognition systems into their practices. By using the key word, or “macro,” feature, I’ve made the best of what this imperfect technology has to offer and made dictation more accurate, more efficient and less expensive.
Key words are commands that, when spoken into your computer’s microphone, insert pre-defined phrases and paragraphs of text into your documents. Here are some examples:
Physical examination. I have defined a key word, “NormalMale,” that pulls up an entire normal physical exam for a male patient. To use it, I create a note and dictate word-for-word the patient’s history, medication list, etc. Then, when I reach the physical exam portion of the note, I merely say the key word “NormalMale,” and the entire physical exam description is automatically inserted into the note. This saves me considerable time, as I usually only have to make minor corrections. I use “NormalFemale” for a normal female exam, and I also have defined key words for the separate organ system exams.
Illnesses. Common acute illnesses also lend themselves well to key words. For acute bronchitis, I have a key word that produces a fairly complete progress note. I type or dictate the patient’s name and the current date, say my key word, and the computer displays the remainder of the note, which includes the usual symptom presentation for bronchitis, the usual physical findings (including many normal organ system exams), my usual treatment and follow-up plan and even the signature line. I intentionally make the note fairly general so that I can keep my revisions to a minimum; however, it still very accurately describes the patient encounter.
Medications. I also have a number of key words for commonly used medications. For example, I use “Amox5” to represent “amoxicillin in a dosage of 500 mg taken three times daily.” “Amox2” represents “amoxicillin in a dosage of 250 mg per 5 mL, 1 tsp. taken three times daily.”
Place the microphone correctly. The microphone should be located just forward of the corner of your mouth and about 1 to 2 inches away. If the microphone is placed too close to the mouth, it can pick up breath sounds, which may be interpreted as words. I prefer a headset-type microphone, which allows for some head movement while keeping the microphone at the correct distance from the mouth. Outside noise is usually not a problem with most voice recognition programs. If you are interrupted while using the program or have to respond to something going on around you, simply switch the microphone off using your computer mouse. This will prevent unwanted conversation from showing up in your dictation.
Speak distinctly, not slowly. Speed is much less critical than clarity when using voice recognition programs. Both IBM’s ViaVoice and Dragon NaturallySpeaking can keep up with a normal pace of speech. In fact, it is best to speak naturally rather than to intentionally pause between words. (Some of the older programs required that pause.) Dragon NaturallySpeaking uses context-sensitive algorithms in recognizing continuous speech, so speaking naturally actually improves its accuracy.
Complete the training session. All voice recognition programs require that you “train” the program to recognize your voice and pronunciation style. Follow the directions, and do not try to shortcut this training process, or accuracy will suffer. You can also retrain these programs if you discover ongoing difficulty with selected words. For example, the program may confuse the words “where” and “there,” which sound very similar. If the program guesses any word incorrectly, it is best to immediately retrain it to differentiate between both the correct and the incorrect word. Granted, this does take time, but it will increase your future accuracy.
Selecting key words
Key words must be sufficiently distinct from other spoken words that the system recognizes. For example, the key word “NormalMale” must be spoken as a slurred-together phrase. If I were to pause briefly between “Normal” and “Male,” the system wouldn’t recognize the command and would insert the two individual words into my document. To avoid this confusion, I often use Spanish words for key words; however, any language, even fabricated words, will work just as well. I use the Spanish word for child, niño, for my normal child physical exam. And I use the sound “sissss” to generate a fairly complete but negative review-of-systems paragraph.
You can designate key words for almost any common phrase you often find yourself repeating in your notes. I use about 30 key words. Select key words that are easy to remember or, if you don’t trust your memory, keep your list of key words near your computer. While it will initially take you extra time to define your key words, you’ll save much more time in the long run by not having to repeatedly enter this information in patient charts.
Choosing a system
If you don’t have a voice recognition system, I would encourage you to consider one. The software is more practical, faster and easier to use than the earlier versions, and when properly used, it’s 98 to 99 percent accurate. (For more tips on improving accuracy, see the box.)
Although I have used IBM’s ViaVoice (www.ibm.com/software/speech), I consider Dragon Systems’ Dragon NaturallySpeaking Medical Suite [recently renamed Dragon NaturallySpeaking Medical Solutions] to be the gold standard (www.dragonsys.com/products/naturallyspeaking/medical). Yes, the software is expensive (the medical version of Dragon NaturallySpeaking software costs approximately $1,000), but it can save you money. Before I began using a voice recognition system, my transcription services cost approximately $10,000 per year, not including the indirect costs associated with patient files having to be handled twice – once for the patient visit and once when the typed note was returned.
Now, with the help of key words, I often have the medical note completed, printed and ready to be filed before the patient has left our office.
Dr. Knight is a former medical director of High Sierra Medical Group, St. Mary’s Health Network in Sparks, Nev., and is currently on a medical mission, working in Peru and Bolivia. He used his voice recognition system to prepare this article.
Editor’s note: For more information, see “Voice Recognition Software: A Tool for Encounter Notes,” FPM, February 1999, page 55, and “A Comparison of Voice Recognition Programs,” FPM, April 1999, page 48.
Copyright © 2001 by the American Academy of Family Physicians.
This content is owned by the AAFP. A person viewing it online may make one printout of the material and may use that printout only for his or her personal, non-commercial reference. This material may not otherwise be downloaded, copied, printed, stored, transmitted or reproduced in any medium, whether now known or later invented, except as authorized in writing by the AAFP. Contact email@example.com for copyright questions and/or permission requests.
Want to use this article elsewhere? Get Permissions