Best Practices in Speech Recognition

hiit040608Speech recognition software promises some pretty tantalizing results. Reduced costs, lightening fast report turnaround and a standard format for reports top the list. A hospital or imaging center can adopt a strategic approach, using speech to attract physicians and redeploying transcriptionists to added value duties that lighten physicians’ workloads. But simply investing in technology without implementing structures to foster success may not yield such positive results. This month, Health Imaging & IT visits with a few speech pioneers to uncover the secrets to their success.

“The issue is always acceptance among radiologists,” opines Edmond Knopp, MD, section chief for neuroradiology at New York University Medical Center in New York City and a user of Commissure’s RadWhere Suite speech reporting solution. Sites can facilitate acceptance by using a carrot or a stick approach.

NYU used a stick approach, making the decision to transition from an earlier speech solution to RadWhere and converting all users in a single morning. The ability to import macros from the previous product simplified the transition, says Knopp.

For most facilities, success does not come overnight. Emory Healthcare in Atlanta relied on a stick when it deployed Dictaphone Corp.’s PowerScribe solution seven years ago. “We went from six transcriptionists to one by requiring physicians to pay for transcription,” explains Dena Wheeler, radiology applications analyst. At the same time, Emory employed a more gentle touch with ongoing training and support. Wheeler provides one-on-one, on-the-spot training for the hospital’s 173 users. “If hospitals are going to implement speech, they need to have someone there to support it. Users shouldn’t have to wait a few hours for someone to respond to a call,” says Wheeler.

In fact, Ed Roman, CIO at Moses Taylor Hospital in Scranton, Pa., identifies ongoing training and support as one of two keys to speech recognition success. The other key? A flexible back-end/front-end approach that gradually transitions physicians into speech as the system is trained, Roman says.

Moses Taylor uses Dolbey Fusion Expert powered by Philips SpeechMagic. During the first deployment stage, the system compares the typed report to what it thought it recognized. After the transcriptionist corrects the document, the system stores and learns the corrections. Once the system reaches an 85 percent recognition rate for a physician, the doctor moves to front-end recognition with a speech recognition system administrator providing one-on-one training and support.

Other sites find that the proverbial carrot is more persuasive. For example, some determined hospitals share recouped transcription costs with speech converts.

The macro factor

Most speech solutions incorporate macros to create shortcuts and templates and automate some processes such as normal reports. “Use them,” affirms Wheeler. Wheeler created a barcode for some macros, enabling users to complete 20 normal reports in five minutes.

In fact, Knopp recommends potential buyers pay close attention to advanced features and capabilities like macros. Smart templates can save time by filling in standard report components; the radiologist dictates abnormalities. “This also standardizes final reports to make it easier for referring physicians,” Knopp explains.

As always, however, technology is only half of the equation. The speech recognition system administrator at Moses Taylor builds shortcuts and templates for physicians. “Most physicians don’t have the time to play with the system and learn these advanced features. When we completed site visits, we quickly realized that the successful sites made ongoing use of a system administrator,” explains Roman.

Other features that can boost acceptance include a high rate of accuracy, stability and a responsive vendor, says Knopp. Buyers also should analyze their users and needs and try to find a solution tailored to those needs. A radiology-centric solution, for example, won’t work very well for pathologists.

The final word

There is no absolute recipe for speech success. But those who have “talked the talk” recommend:

  • Evaluate all of the features of speech recognition solutions under consideration.
  • Carefully weigh the pros and cons of a gradual or rapid rollout.
  • Decide on a carrot or stick approach for users.
  • Provide users ongoing training and support, with an eye toward immediacy.