Transcription companies unvexed by VR advance
Despite strides in voice recognition technology and announcements such as that of Microsoft's long-awaited Speech Server, some of the nation's largest transcription providers say they're not worried at all.
"We don't see it as a threat," says Spheris CEO Steven E. Simpson. "We might if we didn't see ourselves becoming a leader, but that's not the case."
Indeed, Spheris is an example of a company – make that two companies – that looked into the future and liked what it saw. Late in 2003, Spheris emerged as a top-three transcription company following the merger of physician group-facing Total eMed and EDiX, the former IDX subsidiary that provided transcription services mostly to hospitals. "There was a great opportunity to put these two companies together and create a stronger Number 2," Simpson said. "And we are advancing ourselves on the technology front."
Number 1 in the industry is MedQuist. Chief Technology Officer Ethan Cohen says his company, like Spheris, sees opportunities for growth through the new technologies rather than competition. Over the past three years, MedQuist has acquired smaller players including Digital Voice, Inc., Lanier Healthcare and Speech Machines. Now the company is converting its entire operations to a Web-based speech enabled typing application built around the Philips speech engine. The deployment is expected to be complete within two years.
While he acknowledges that speech and voice recognition applications are getting better, Cohen says, "the challenge is not understanding what the clinician said, it's what the clinician meant… This requires cognitive interpretation, and technology isn't going to get it. There will be a need for medical editors for years to come."
Both Spheris and MedQuist are also skeptical about the use of back-end speech recognition, even when the technology is mature. The accuracy of back-end speech recognition is still not good enough in most medical specialties to allow for cost-effective self-editing
In contrast, MedQuist thinks front-end speech recognition – already applied with success in radiology – could prove effective in certain specialties such as cardiology and pathology. "front-end speech recognition is generally targeted at medical specialties where the dictators use fairly limited vocabularies and a fairly consistent sentence structure. "In these environments,' Cohen notes, "front-end speech recognition with physician self-editing is often a cost effective solution that allows for rapid turnaround time."
The key issues remain cost, time and integration into physician work-flow.
"Our position is not to slow down or impair clinicians to make a transcriptionist's job easier," Cohen says. "Requiring the clinician to make major behavioral changes when dictating is difficult and can be detrimental to the overall process."
(Ed. note: This story was updated on May 24, 2004, to clarify MedQuist's position on back-end and front-end speech recognition)