What's going on behind the scenes, however, has changed a great deal: the recorder is a computer application, which usually outputs a digital .wav file. This file is sent over secure Internet connection to a transcriptionist, who could very well be in the Philippines or India, according to a report in the Feb. 14, 2009 issue of The Lancet. The practice gets back a document that is not paper, but more likely a digital text file.

Because of these changes, practices should reevaluate their transcription methods and services if they haven’t done so recently.

Outsourcing and Voice Recognition
"The overseas outsourcing that started about five years ago has really changed the landscape considerably," says Jeffery Daigrepont, a senior vice president of the Coker Group. "Those companies work while we're sleeping, so it's improved turnaround time. And labor costs are significantly lower."

Despite -- and perhaps in part because of -- faster, cheaper transcription services -- most practices have avoided even considering speech recognition applications (most notably Dragon Naturally Speaking), even though speech recognition accuracy has improved considerably over the past five years.

"I actually don't know many practices that use voice-based transcription services," says Kevin Pho, who writes the popular medical blog, KevinMD. "It's difficult to say why. There's a lot of pressure on physician practices to adopt computerized or digital medical records, and I think that added expense of a voice recognition system really adds on to the cost."

Pho makes no claim to being an expert in either transcription or voice recognition technology. However, his sentiment that voice recognition systems face "a huge obstacle" because of the likelihood (or perception) that the transition to such a system would create a "huge disruption in workflow" appears to be widespread and will be difficult to overcome.

Efficiency is a major selling point for speech recognition vendors. A recent press release from Nuance Communications featured the subhead, "With Dragon Medical Speech Recognition Software, doctors create highly accurate, real-time electronic health records simply by talking." The release touted the software's ability to help doctors create accurate and detailed medical reports, spending "up to 25 percent less time documenting care, when compared to using a non voice-enabled EHR."

Daigrepont agrees that accuracy of voice recognition software has greatly improved, in part due to large, specialized medical dictionaries included with the software. But voice recognition software must be "trained" to recognize an individual's voice, and this training can take a fair amount of time -- time physicians usually aren't willing to sacrifice. When they do take the time, says Daigrepont, the results can be excellent. But many expect the software to work off the shelf, which results in an unacceptable level of errors.

"The problem with the mistakes is that it [the software] doesn't necessarily spell the word wrong." says Daigrepont. "It inherently confuses words. We had an ophthalmologist, for example, who was using it, and every time that this poor doctor said, 'nice eyes,' the program accidently converted it to 'nice thighs.'"

While office workers or transcriptionists can be trained to edit voice-recognition output for mistakes, it can be difficult to discern errors when they are removed from their context. Sometimes a doctor's comments about a patient's thighs are entirely appropriate.

Which was why one of the key summary points in the preliminary findings reported for the 2007 Bentley College survey of medical transcriptionists was "Speech-recognition technology is only a transcription tool, and not a transcription replacement. The job involves too much professional intelligence and sense-making ability to have it migrate completely to speech recognition."

It is crucial, however, that speech recognition is mentioned as a transcription tool, as many are using the software to make the overall process more efficient. Nuance has said that transcriptionists "are able to transcribe an average of 30 percent more quickly through the use of the technology." 

Talking About the Bottom Line
While global trends in transcription and speech recognition are interesting, and very useful to think about, each practice is unique and therefore must take into account many factors when assessing its current transcription system, and whether changes to the system should be made.

Daigrepont says that Coker suggests practices consider the following when evaluating this:

• Is the current method of transcription the most cost-efficient one? Is the practice taking advantage of templates and other structures to streamline the process?
• Is outsourcing transcription less costly than having it done in-house? (For smaller practices, it usually is.)
• Can the transcription company return text directly into the physician's EMR?
• In what way(s) can the transcription company accept audio files, and in what way(s) can the original files be returned? Often physicians using EMR would like to "attach" the audio file to the appropriate, corresponding record for future reference.
• If transcription is outsourced, what is the average turnaround time? Is this acceptable and cost-efficient for the practice?
• What is the most efficient way to index transcription so it can be found going forward? In answering this question, the use of, or shift to, EMR and HER are crucial considerations.
• If the plan is to change the method of transcription, how much of a physician's personal time will it take?
• How will the transcriptions coexist/be integrated with related electronic and paper files?
• Will changes in front-office personnel be necessary?

When these factors are considered thoughtfully, it's possible that your practice can save both time and money on transcription, without significantly hampering physician workflow.