Skip to content Skip to navigation

A First-Hand Experience to Better Clinical Documentation Using Speech-to-Text Software

August 27, 2014
by Rajiv Leventhal
| Reprints
With this program, N.Y. endocrinologist says no one is more efficient in the EHR than him
Brian Levine, M.D., dictating into the medical record

When it comes to medical data, most people probably think about numbers and text rather than voice. But in recent years, speech recognition has proven the ability to enhance clinical documentation in various ways, which is more important now than ever as the demand for accurate documentation with every patient care encounter is growing.

At the Manhattan-based New York Presbyterian Hospital (NYPH), Brian Levine, M.D., has been using speech recognition software for years, since his days in medical school. Currently, Levine— a clinical fellow in reproductive endocrinology and infertility—sees up to 50 patients a day in addition to operating two to three days a week. Needing to maximize his time with patients, Levine knows that saving a few minutes here and there each day could prove extremely valuable.

As such, reproductive endocrinologist Levine has adopted Dragon Medical software from the Burlington Mass.-based vendor Nuance. After only 90 minutes of training, Levine says he became comfortable enough with the technology to begin using it in his daily workflow. “Since I started using it, it has saved me minutes per day, hours per month, days per year,” Levine says. “That is found time, and in this profession, found time is really valuable.”

When it comes to speech recognition software, a big concern is the accuracy of the transcription. Nuance says that Dragon Medical is up to 99 percent accurate out-of-the-box, and includes medical vocabularies covering nearly 80 medical specialties and subspecialties. Levine backs this claim, and adds that when he first started to use it, the program asked him what his specialty was to tailor its vocabulary to his needs, and even asked where he was from just in case it needed to pick up a particular accent.

Levine was actually able to show me first-hand how accurate it was during a recent visit to NYPH. Using the program, Levine is able to dictate in real time into its electronic health record (EHR) in his own words – letting him instantly review, sign, and make his notes available for other clinicians. Levine even speaks into the program’s microphone—which is plugged into his personal laptop—to tell the program to open.

He literally has no need for a keyboard, giving the program commands from “select all” to “new paragraph” to “cut and paste” to “scratch that” when he wants something deleted. And there is no need to speak like a robot, as speaking at your normal pace works just fine, Levine says. “Anywhere you can click, you can type. And anywhere you can type, you can talk,” he says. Levine was able to dictate the below note in well under a minute, with no errors in accuracy.

“I can dictate a note before most people think about writing one,” Levine says. If there was a mistake, Levine simply would tell the program where the mistake was and how to fix it. Levine also uses the software to enter a patient’s order into the EHR’s computerized physician order entry (CPOE) system with a simple command. And if there ever a need to look up something on the web about pelvic pain, for example, or search for an ICD-9 code, Levine uses the program’s microphone for those things as well.

Levine has also found other ways to leverage the technology. During a physical exam, for instance, he can dictate the patient’s note while the patient is right there in the office. Without a program such as this, he says, physicians will enter the note non-verbally, and that a) takes more time and b) doesn’t prove as effective in getting the patient more engaged and informed. “The patient becomes engaged when you document by speaking with him or her right in front of you,” he says “They hear key things.”

Previously, with operative notes, Levine says that he and his colleagues would speak into the phone and have the note get transcribed by someone—likely in another country—only to have it sent back to be edited. This cumbersome process can take as much as 72 hours, Levine says. “Operative notes could be very sloppy for this very reason. And of course, there are transcription costs involved with third-party companies. With this program, physicians can pay per license or institutions can subscribe to a site-wide license.”

Undoubtedly, voice recognition software has provided the means to lower transcription costs, speeding efficiency and populating data for achieving meaningful use; Levine now can open a note and transfer the accurate text right into the EHR. To that end, Levine says that physicians will often cut and paste their notes from previous ones, which can be dangerous because the note becomes much harder to defend. In fact, earlier this year, the Chicago-based American Health Information Management Association (AHIMA) released a statement saying that the use of the “copy and paste” functionality in EHRs should only be permitted in the presence of strong technical and administrative controls. “But now, I can justify everything that I have done,” Levine says, noting that when it comes to being audited, comprehensive and accurate notes are important.

When asked about criticism regarding speech recognition software, Levine says the negativity boggles his mind. “Doctors in smaller practices are worried about their practices staying open, but this type of software could save the doctors that need to type with [one finger at time],” he says. “In fact, I think EHR vendors should be telling these doctors how helpful this program could be to them.”