Voice Recognition

March 1, 2003

Storage problems, transcription costs of more than $125,000 per year, and the imminent retirement of a workhorse transcriptionist led practice administrator John Bonini to one conclusion.

Storage problems, transcription costs of more than $125,000 per year, and the imminent retirement of a workhorse transcriptionist led practice administrator John Bonini to one conclusion.

"I wanted to go paperless," Bonini says.

He researched and tested multiple products before settling on an electronic medical record (EMR) he thought best fit the needs of the practice, N.Y.-based Nassau Orthopedic Surgeons, and its seven physicians. The technology overhaul went well - but Bonini thought the process could go even better.

"After using the EMR system for a while, the physicians found that it was very difficult to create templates for some things, such as history of present illness," Bonini says. "You can't create templates for all the things we might see in our practice. I mean, it's a little difficult to cover everything from 'back injury from roller coaster,' to, 'slipped on can of peaches in the supermarket.' "

So Bonini went one step further and added voice recognition software to the mix. He found that the software gave his physicians the ability to personalize reports and charts in a way that made them seem like they were actually created by a real human being, not a machine. Plus, he expected that, with time, voice recognition would boost efficiency, cut costs, and prove to be a valuable resource to physicians, staff, and patients alike.

Bonini was right. Together, the EMR and voice recognition provided the ideal combination of flexibility, convenience, and efficiency for this practice that sees more than 25,000 patients a year. In their second year of using voice recognition in conjunction with the EMR, Bonini estimates practice costs are down $100,000 while volume is up by about 3 percent. And all without sacrificing the personal touch Bonini felt was vital to the practice.

"With our EMR and voice recognition," says Bonini, "I can hit my Shangri-la."

Technology has improved

Not everyone is quite so enthusiastic about voice recognition software, but most experts agree that the technology has improved drastically since it was introduced to the healthcare industry in the 1980s. In fact, a 2001 study published in the American Journal of Emergency Medicine found that voice recognition technology had an accuracy rate of 98.5 percent, or 2.5 errors per chart, with a turnaround time of only 3.65 minutes.

While traditional transcription services were slightly more accurate with a rate of 99.7 percent, the turnaround time was substantially higher at 39.6 minutes. And with a price tag of approximately $2,200 per user, voice recognition isn't going to break the bank.
"The technology has really come a long way," says Sue Plummer Taylor, senior vice president of marketing for VantageMed, a company whose products, such as EMR, are designed with the ability to work with voice recognition.

To prove her point, Taylor cites a physician client whose use of voice recognition technology with an EMR allowed him to access critical patient information just hours after the patient had been seen in his office. When the patient showed up in an emergency room that night, the physician was able to view the notes from that day's appointment from his home, something that would have been impossible with traditional transcription methods.

"In this physician's mind, that was the most important issue - providing good patient care. It's why he went to medical school in the first place," Taylor explains. "In this case, having instant access to a medical record really made a substantial difference in this patient's care."

Savings are substantial

The cost savings with voice recognition technology can be substantial, too. The Skyline Family Practice in Front Royal, Va. found that the voice recognition software paid for itself within two months after they saved $600 per month on transcription costs - and boosted their efficiency in the process.

"Perhaps the biggest cost savings is in the speed with which progress notes and letters are generated," says Floyd "Tripp" Bradd, III, MD, a Skyline physician who helped implement voice recognition in the practice in 1998. "With voice recognition, the progress note is completed immediately. All the hands-on activity required in the transcriptionist system was removed."


While the ability to eliminate transcription charges - which can cost a practice up to several thousand dollars a month - is an immediate and tangible benefit, there are other cost savings to voice recognition that may not be apparent at first.

"Everything, and I mean everything, is done by voice now," Bonini says. "Eventually, we were able to get rid of all secretarial work. ... Even something as simple as a letter that needs to be written to a lawyer explaining the extent of a patient's injuries is done with voice recognition. We reduced our staff from 33 to 24 in six months, saving over $200,000 in gross payroll dollars."

But wouldn't such a drastic reduction in staff lead to strain and stress on those remaining? Not necessarily, according to Bonini. In fact, the efficiency of voice recognition in conjunction with the EMR created an environment that is quite the opposite.

"I have 100 percent employee and physician satisfaction," Bonini says. "And people who come in - patients, sales reps - say that it is the quietest and most relaxed practice they've ever visited."
In addition, Bonini says that the quality of care is up. "I think that voice recognition has increased the quality of our consultation notes by 30 to 40 percent," he says. "And patients are semi-dazzled when they see the doctor, right in front of them, speaking into a wireless laptop that takes down everything they say. Patients kind of just stare at the doctors when they explain that they just completed their chart and faxed it to their primary-care provider."
Patients aren't the only ones who have noticed - referrals are up, too. "Primary-care doctors are extremely impressed," Bonini says, noting that all consultation reports are completed and delivered within 24 hours of seeing the patient. "Sometimes the primary-care physician has our consultation notes before he even gets his own chart back from his transcriptionist."

Weigh the downside

Bradd feels there is no significant downside to voice recognition technology. However, he warns, the technology will be used by humans - and that's where the problems can arise.

"If the physician cannot devote the time or the energy to train the voice recognition system and, more importantly, is not receptive to this process, then voice recognition is doomed to failure," Bradd says, adding that older physicians or those not comfortable with computers may be particularly resistant.

Bonini admits that it takes a little time to get things right - something that can be frustrating for practices that expect instant results. Voice recognition increases efficiency and convenience in the long run, but there is an initial two- to three-week learning curve, plus start-up time spent "reading" to the software so that it can "learn" your voice and create an individualized profile to increase accuracy.
"Our one problem is with a physician who has a heavy South American accent," Bonini says. "He uses the voice recognition with about 85 percent accuracy as compared to everyone else's 95 percent. But this has more to do with his use of the system - rather than letting the software learn his voice and adapt, he tends to be impatient and catch the errors immediately and correct them himself, so the system never learns."

There also is additional time that must be spent on a daily basis. Bonini estimates that his physicians now spend an extra 20 minutes a day using voice recognition compared to the old days of using a transcriptionist.

"Before, the physicians could rattle off a bunch of half sentences and jargon over the phone to the transcriptionist and then hang up, figuring it was someone else's problem now. With voice recognition they have to enter and correct their own charts, so they have to learn how to speak slowly and more clearly, then make the corrections themselves," Bonini says. "But our physicians have learned to make up the time by working more efficiently."

And while voice recognition technology can be amazingly accurate, it's not foolproof. Forget that and you may be in for an embarrassing surprise. Case in point: One physician using voice recognition found that the words, "evaluating the patient," had been recognized by the software and entered on the chart as "violating the patient." The lesson? Proofreading is essential, and corrections will probably be necessary.

Getting started

Deciding whether voice recognition is a good fit for your practice is the first step. Despite all its benefits, voice recognition is not for everyone. Physicians who are used to doing things a certain way, or who are wary of technology, for example, may not be good candidates for voice recognition.

Ask yourself a few questions. Will you be able to get staff buy-in? Are your physicians and your staff computer-savvy and open to technology? Will they take the time to learn how to use voice recognition properly and continue to take the time to achieve accurate results? And do you already have (or plan to have in the near future) technology that is easily integrated with voice recognition software, such as an EMR?

If you think you are ready to try voice recognition, you have several choices. But do your homework. For example, some EMRs require specific voice recognition software, while others can integrate with several software types. Research and test different software packages until you get the right fit.


There also are accessories to consider. Microphones are important to the success of voice recognition. Bonini suggests investing in the best microphones possible, and testing several to make sure they are both high quality and comfortable for the users. After trying several models, Bonini's practice settled on a microphone that is hand-held and also functions as a mouse, but others might find a headset or tabletop microphone a better choice.

There also are additional software packages available to tailor voice recognition to individual specialties, like dictionaries of words and phrases that help voice recognition software "learn" the unique terms of a practice's specialty.

And while most who have implemented voice recognition will admit that it takes a little patience to get the ball rolling, it doesn't take a whole lot of computer know-how.

"I'm not a person with a big technology background, but within two hours of buying the software and microphones, I had installed the voice recognition software and hardware and was testing it," Bonini says. "We set up on all the PCs within the next day or two and within two to three weeks everyone was using it with extremely good accuracy. It was that easy."

Voice recognition technology is nothing new, but recent advances have finally made it a viable option for healthcare professionals. The technology will continue to improve, but the software packages on the market now are already proving to be efficient, cost-effective, and convenient alternatives to traditional transcription methods. The art of transcription may never disappear completely, but, according to Bradd, the question shouldn't be if you will implement this voice recognition technology into your practice, but when.

"There's really no reason for a physician not to move ahead with voice recognition software," he says.

Cassie Gainer is an associate editor for Physicians Practice. She can be reached at editor@physicianspractice.com.

This article originally appeared in the March 2003 issue of Physicians Practice.