Since radiologists first began using voice recognition more than 30 years ago, supporters and resistors have debated the software intensely—in clinical journals as well as across reading room cubicle walls. Today, voice recognition is a mainstay in many reading rooms, slashing turnaround time, cutting costs and improving radiologists’ accuracy and control over reports, argues David L. Weiss, MD, physician coordinator for imaging informatics and radiologist at Carilion Clinic in Roanoke, Va. And yet, despite dramatic improvements to the technology and an overwhelming convergence of the debate in favor of voice recognition’s benefits, many physicians remain resistant to the software, arguing that even though voice recognition appears to cut costs and trim turnaround, more work and time have just been shifted to radiologists. The experts rebut these complaints: voice recognition dramatically improves radiology reporting, they say, but you have to know how to use it.
1. “Training, training, training,”
is the battle cry of David Hirschorn, MD, director of radiology informatics at Staten Island University Hospital in New York City. When done right, voice recognition eats up time at the outset; but correcting the software’s mistakes early on means that the program learns what you meant to say and ensures that it won’t make similar errors in the future. “It’s like a puppy, if you don’t train it early, it becomes a lot harder to fix later,” quips John Kaschko, MD, medical director for informatics at Group Health Cooperative in Seattle. Moreover, “Follow-up training is really the key to optimizing all of the time-saving features on your system,” says William W. Boonn, MD, radiologist at the University of Pennsylvania Health System in Philadelphia. Boonn continues, “It’s where radiologists who are beginning to understand the systems can really ask the good questions about optimizing the features.” Weiss agrees. “With adequate training, it’s possible to completely or almost completely offset any additional time that a transcriptionist once saved the radiologist.”
2. No need to reread.
Thanks to huge archives of report templates (macros), offered by larger hospital systems and through RSNA, dictating a report becomes a matter of words rather than paragraphs. In addition, templates help ensure that all necessary clinical and administrative information is entered into the report. “Templates are absolutely key to efficient use,” says Weiss. “I use a template nearly 100 percent of the time.”
3. “Imagine you’re a baseball commentator, it’s no different,”
advises Hirschorn. Better yet, take the advice of the perennial, 55-year announcer for the New York Yankees, Bob Sheppard: “A public-radio announcer should be clear, concise, correct.” Speaking clearly and enunciating are essential, says Hirschorn. And because voice recognition guesses what you’re saying based on contextual probabilities, speaking in phrases, polysyllabic words and complete concepts, all at a constant pace and without pause, gives the software more information—a context—for deciphering your reports, offers Howard M. Landa, MD, chief medical information officer at Alameda County Medical Center in Oakland, Calif. “It’s one of the rare times where increasing complexity actually enables the system to work better,” Landa insists.
4. Emphasize interoperability.
“Integrating voice recognition with clinical decision support, critical findings communicators and peer review programs saves considerable time,” according to Boonn. Weiss uses a mouse with 17 buttons on it, allowing him to navigate between applications much more quickly, he says. Prior to buying a new system, Boonn and Weiss recommend finding a practice with similar needs and the same PACS as your own and asking for pointers on how best to integrate your systems.
5. Try a headset.
Many radiologists find there’s a sweet spot for attaining the clearest recordings, but they tend to unconsciously drift away from that hand position. There’s a much easier solution: switch to a headset. “Switching to a headset eliminated 50 percent of errors for me,” Weiss observes.