MaSSSoftech

About 10 years ago, health information technology experts were touting speech recognition software as one of the best things to come to health care. About five years ago, the talk was about why adoption of the technology never took off like it was expected.

Physicians' chances to earn incentives from the federal stimulus package have forced many to look at speech recognition as a way of making electronic medical record adoption more palatable to resistant physicians. And because improvements have made speech recognition more reliable and easier to use, some experts and physicians say speech recognition is finally ready to take off.

But will it? Experts say that at least some practices are willing to try it out again.
Megan Hastings, vice president of Health Directions, a Chicago-based consulting firm, said she has seen an increase in practices wanting speech recognition in recent months.

"When you get into a situation where the physician can't type, or [EMR adoption] is severely impacting their clinical work flow, we often see that institutions and practices will purchase [a speech recognition system] to help those physicians struggling to get on board," she said.

It's also helped that the technology has improved greatly from several years ago, Hastings said.

Larry Garber, MD, an internist and medical director for informatics at Fallon Clinic, which has more than 20 locations in Central Massachusetts, tried the technology in 2004. But he quickly discovered that it wasn't going to work. He stopped plans to roll it out to the entire practice.

"It wasn't ready for prime time," Dr. Garber pointed out. "Now it is. No question."

Dr. Garber's practice adopted the technology last year and found that the nagging problems such as speed and accuracy had been resolved. Using the speech recognition software not only has improved quality but also has saved the practice an average of $7,000 per physician per year in transcription fees. That never would have happened with the old technology, he said.
Technology more accurate

Keith Belton, senior director of product marketing for Burlington, Mass.-based Nuance Communications, which developed Dragon Medical, the most widely used speech recognition system in the medical industry, said he admits the technology was lacking in previous versions. But he said it now boasts an accuracy rate of 99%.

When Robert Frank, MD, a family physician with Aurora Advanced Healthcare in Milwaukee, first used the technology in 2004, he said many physicians felt it actually slowed them down. They were forced to go back and edit all of their notes because of the lagging accuracy, Dr. Frank said.

Newer versions are not only more accurate but also more robust, their vendors say. Systems now can work in concert with EMRs, for example.

Dr. Garber's internal study of how speech recognition was affecting efficiency at his practice found that EMR adoption alone didn't improve the availability of information.

The majority of physicians still were using traditional dictation services, which had about a four-day turnaround time, not including the time it took to do edits, once the transcripts were returned. The addition of speech recognition reduced that time to an average of 46 minutes.

Physicians now dictate clinical notes, which are transcribed directly into the patient's file, he said. Once the notes are complete, physicians check for accuracy, then approve them, and they are automatically saved to the record. The system also allows physicians to enter text, such as lab results or medicine lists, using a voice command shortcut.

That might make physicians faster at getting clinical notes into the system, said Mike Uretz, executive director of EHR Group, a Seattle-based consulting company. But if they want to qualify for incentive money for adopting EMRs, voice recognition used in conjunction with an EMR could be an impediment.

"The problem is when you talk into it, the data is not discrete ... it's still like a Word document," he said.

To meet the stimulus package's "meaningful use" requirements on EMRs, physicians will have to do quality reporting, which requires discrete data, Uretz said. Data are "discrete" when they can be categorized into specific fields.

Dr. Belton said it's likely that future versions voice recognition systems that work with EMRs will have this functionality.

Dolan is a business reporter. She can be reached at 312-464-5412 or by e-mail (pamela.dolan@ama-assn.org).