Since May 1997, I've been producing most of my patient notes by dictating not into a tape recorder but into a computer. I've used voice recognition software to produce legible, reasonably complete encounter notes for the 25 to 35 patients I see each day — and I generally finish within a few minutes of my last visit. Clearly, this technology will not be the answer for everyone; it's only one solution among many. But I believe it can be a great help for many family physicians.

Your first question might well be, “If I have a tape recorder and a secretary or a transcription service, why bother with a computer?” That's how I used to do it, too, and it has its advantages. You can dictate into a tape recorder very quickly, and you can use conversational speech while relying on your secretary to correct the irregularities. Current voice recognition software also allows you to dictate continuously and rapidly, although the dictation must be precise to be effective (the programs can't fill in words and context that a human listener would reconstruct). But voice recognition software offers two important advantages over traditional dictation: Your notes are available as soon as you dictate them, and you don't have to pay anyone for transcription.

Better systems

I became interested in computerized voice recognition after coming to two realizations: Eventually, patient records will be kept on computers, and I'm not a typist. I first tried to use voice recognition software a few years ago. But until recently, the technology required users to pause between words to make the software work — so-called “discrete” voice recognition. I tried a discrete voice recognition system, but I couldn't make it work for me because learning the program and using it required more effort than I was willing to put in.

It wasn't until the spring of 1997 that practical, affordable, continuous speech recognition — which no longer forces you to pause between words and gives you more freedom in your verbal commands — became available. At that point, I purchased Dragon Systems' Dragon NaturallySpeaking (IBM and Lernout & Hauspie soon came out with similar products — ViaVoice and Voice Xpress, respectively), and I've been using subsequent versions of the Dragon Systems product ever since.

After a brief training session, I immediately realized that continuous voice recognition software is much more effective than the discrete variety. But I encountered a fairly steep learning curve as I tried to find ways to minimize the corrections I had to make manually and as I added medical terminology to the system word by word. (Vendors now offer medical vocabulary programs for their voice recognition software.) My persistence was rewarded as the program gradually improved in accuracy and learned my specific vocabulary.

Over the past year and a half, medical voice recognition systems have become easier to use and much more accurate. I currently use Professional Edition 3.0 of Dragon NaturallySpeaking, which sells for about $700, coupled with a medical vocabulary designed for family practice by Voice Automated, which costs about $300. I've also improved my own use of the software by designing several voice macros that make dictation much easier in recurring situations. (With a single voice command, the macros insert blocks of text that I tend to repeat.) In addition, a number of programs are available that automatically format notes once you've created them.

Making the choice

Is a medical voice recognition program right for you? If you write your notes by hand and would like to have typewritten notes but don't want to pay for transcription (which might cost $6,000 to $10,000 per year for a family physician), then this represents a reasonable solution. After a modest learning curve, producing neat, thorough notes should take no longer than writing notes by hand — and your notes might also be more complete.

If you currently use a tape recorder and a transcriptionist, your decision is more difficult. Using voice recognition software will take longer than dictating into a tape recorder (and that assumes you have a basic knowledge of using computers and word processing programs). Using my system means that I work about 30 minutes longer each day, although I don't miss that time. Since physician time is worth considerable money, this “expense” may offset the cost savings you gain from not paying a transcriptionist. But keep in mind that using voice recognition software offers other advantages. You don't have to review your notes after they're transcribed. You can write referral letters on the spot (sometimes I give them to patients before they even leave my office). And I believe that seeing your thoughts in writing as you enter your notes leads to a more precise and complete description of the patient encounter.

I'm enthusiastic and optimistic about the ways family physicians might use voice recognition technology. In the near future, voice recognition probably will be an integral part of computerized patient record systems. Perhaps in the more distant future, we'll be able simply to ask our computers to retrieve pertinent patient information and have it appear on our desktops. That may sound like science fiction, but voice recognition was science fiction just a few years ago. As we head into the 21st century, we have a historic opportunity to improve our handling of patient information and by doing so help improve our patients' health. Medical voice recognition is one exciting way we can do that.

Editor's note: See also a comparison of the three voice recognition systems mentioned in this article.