Talk vs. Type: Taking Another Look at Voice Recognition

By Marisa Torrieri for Diagnostic Imaging

Even if you previously passed on using voice recognition technology, with EHRs and the movement toward integrated healthcare, it might deserve a second look.

Like many of his peers who have also practiced medicine for 25 years, family physician Fred Jorgensen doesn’t type as quickly as he can write or speak. So when the medical director for Cleveland-based Center for Family Medicine was looking to cut costs about 12 years ago, he and his colleagues adopted an early version of voice recognition technology.

“We were basically faced with either laying off staff or getting rid of transcription costs,” he recalls.

Though the immediate savings of $25,000 per year — the cost of using a transcription service shared by “just a handful of doctors” — was a big driver of his adopting the technology, Jorgensen says he’s happy he made the change for another big reason: quality.

“As a physician at this stage of my career, do I really want to take the time to learn to type 40 words a minute, or whatever I can learn, when I can speak at maybe 120, 130 words a minute? If I had to type, the quality of my note would be less because I would just not type as much,” says Jorgensen. “So details about a patient’s story, why they were here, details about their illness, and details about them would not be as thorough.”

With electronic patient notes like those produced by Jorgenson in high demand — thanks largely to CMS’ meaningful use incentive program for EHRs — voice recognition software is getting a closer look by physicians’ practices. According to our 2011 Technology Survey, 23.4 percent of 918 respondents said they use voice recognition technology, up 1 percent from 2010. In the next 12 months, healthcare industry experts expect an even greater growth spurt, as more physicians adopt EHRs, and more EHR vendors incorporate voice recognition technology into their systems.

“The healthcare technology sector represents a robust market due to the growing demand for general applications, which include EHRs, computerized physician order entry systems, and nonclinical systems,” says Patricia Kutza, a research analyst with BCC Research. “The key drivers for this market’s growth are the changing government regulations and government initiatives related to decreasing healthcare costs.”

And with time, money, and quality of care at the heart of physicians’ EHR-related woes, those who once turned a blind eye toward voice recognition software might want to reconsider.

Voice recognition today

Like EHRs and mobile devices, voice recognition technology has evolved over the last ten years; today it’s much more sophisticated and accurate. This may be attributed, at least in part, to better microphones that come with voice recognition software packages. Microphones don’t pick up as much background noise, so what is really being said is actually recorded.

“The voice recognition industry has made great strides in the last decade,” says Kutza. “Ambient noise, the nemesis of speech-enabled applications, is being managed much better due to the development and implementation of smarter noise reduction technologies.”

In addition, the technology can do more things.

Physicians talking at their EHRs will find that voice recognition technology also responds to vocal commands so a physician can navigate her EHR’s fields without lifting a finger. “What the physician is typing into is a series of discrete fields using keywords to tab from one field to the next,” says technology analyst West Richel, adding that companies making the technology are working on software they hope will not only transcribe the spoken text, but also figure out which fields in a patient’s chart to fill in.

Nancy Lindo-Drusch, a physician who works at a small practice in Appleton, Wis., says the voice-activated EHR navigation saves as many as five minutes per visit.

“It learns you,” says Lindo-Drusch. “The more you use it, the more it gets those big words. If it makes a mistake, you can [program a change].”

In addition to having more internal features, the technology is adaptable to a broader range of external devices, such as smart phones.

Still, even physicians who use it such as Southampton, N.J.-based solo practitioner and family medicine physician Max Burger, have their issues with it.

“One of the things about voice recognition is you still have to edit,” Berger says.

Ditch the transcription?

The verdict is in: Voice recognition technology is more efficient and also much less expensive than transcription service, which can run $12,000 or more per practice, annually.

But the cost savings of opting for the technology don’t come without a few adjustments that physicians will have to make.

“The main difference between a voice-recognition product like Dragon [Medical] and transcription is with transcription, I dictate, hang up the phone, and I’m done with it,” says Jorgensen. “With speech-recognition technology, when I speak, I actually see what’s in front of me, and in most cases, I’m going to correct it. So I’m not only the dictator, but I’m also the ‘correctionist.'”

But even if your practice has the funds to continue transcription service, there’s the issue of turnaround: It still can take a day or more to receive a note back from a transcriptionist, which means a slower work flow and completion of certain tasks (such as referrals to specialists), Richel, who works for IT research firm Gartner, notes.

And though voice recognition technology is cheaper than a transcription service, it isn’t free: Cost typically ranges from $99 for a basic system to $5,000 or more for a sophisticated system, says Richel, who doesn’t recommend the cheapest options. The more expensive the system, the broader range of vocabulary it recognizes. Cost also varies based on how many workstations in a physician’s office will be equipped with the technology.

“As you get beyond the one-physician practices, you don’t necessarily want the software to be associated with a specific device,” says Richel. “You may want it to work wherever the physician is.”

Speaking tips

When shopping around for voice-recognition software, Jorgensen says communication is key —you’ll be talking to whatever system you pick several times per day. “The holy grail, the single most important thing at the core of the technology, is accuracy,” says Jorgensen. “When I speak, does it type what I say?”

And if you speak into a microphone the way you speak to your patients, you’re bound to make more mistakes, says Jorgensen, adding that physicians need to be conscious of three different things:

1. Microphone placement. You want to make sure you have your microphone on the right spot on your headset or in your hand.
2. Training environment. Ideally you will train the voice recognition software in the environment you’re going to be using it in, so the system can better eliminate and adjust for the typical background noise, wherever that may be.
3. Speaking. You have to speak a little differently in terms of not running words together ― as a system can’t always separate the acoustic signal of what you’re speaking if you don’t enunciate each word clearly. If you normally speak very quickly, you’ll tend to run words together. Many voice recognition systems coming out today include dozens of complex medical terms.

Remember to keep your voice recognition expectations reasonable or you will be disappointed.

Says Jorgensen, “If you expect it to be as accurate as dictating and having a transcriptionist type your notes it’s probably not going to be that accurate the first time around, because you don’t have a human being listening to the words and typing them.”