The trusted source for
healthcare information and
By Julie Crawshaw
Richard P. O’Brien MD, an emergency room physician at Moses Taylor Hospital in Scranton, Pa, started using voice recognition software in 1996, when the user was required to speak slowly and pause slightly between words. The current generation of that software, though, allows users to speak continuously, and O’Brien says that anybody who is willing to work with the system can get good recognition.
"It’s no different than a video game," he says. "You learn how to play it, when it’s likely to mess up and how to edit things quickly." He estimates that most people can function with a voice recognition system in six weeks, and become really good at using it within six months. "When I first started, I was using it to do two charts a day and handwriting the remainder," O’Brien says. "Now, I can do complicated charts in five to eight minutes."
Nearly one-third of the information systems professionals at a recent Healthcare Information and Management Systems Society conference in Orlando, Fla., said they would likely invest in voice recognition technology during the next 12 months. O’Brien, who started with the very first Windows-based system, now uses Clinical Reporter from Lernout & Hauspie (L&H) of Belgium, which recently bought out Dragon Systems, another VR software company popular with physicians. L&H offers voice recognition systems customized for 13 medical specialties, including emergency room and neonatology medicine. Their systems recognize a medical vocabulary of approximately 250,000 terms that include medication names, medical procedures, diagnoses, and diseases.
Users can also create or import multiple custom vocabularies for different specialties or fields of interest. Though a system specifically targeted for critical care physicians is not yet available, O’Brien believes the emergency room vocabulary comes close.
O’Brien now considers his voice recognition system as useful to him as his prescription pad or stethoscope. "You have accurate, legible documents available immediately. If you’re integrated with the hospital information system you could have a paperless chart. And you can get information from previous records because they’re all digital," he says.
He also sees economic benefits in transcription cost savings and medical-legal benefits in better documentation because it’s more explicit. "Plus," O’Brien says, "I don’t consider it a particularly expensive solution."
In his prevoice recognition days, O’Brien dictated his procedures and findings into a cassette recorder. The tape was then transcribed, which took about three hours to turn each hour of tape into a transcribed report. Now, O’Brien’s reports are available instantly. He speaks into a headset plugged into a Pentium 133 computer with 32 megabytes of ram. Templates keyed to American Medical Association standards guide his creation of appropriate medical documentation. He uses a 1440 modem to fax the typed reports anywhere in the hospital, or to the offices of his patients’ primary care physicians.
His reports now go to other physicians immediately and his billing is done faster because the chart is complete and signed when it comes out of the machine. Because nurses and other physicians have thorough, legible records as soon as they get the patient, they know exactly what O’Brien did and why.
Users Can "Teach" the Machine New Words
O’Brien says that when the system occasionally fails to recognize a word—such as the name of a new physician or drug—he "teaches" it to do so by typing in the word or spelling it verbally. "It only takes a couple of seconds to do that," O’Brien says, "and if you don’t get it right the first time, you can delete the error from the system’s vocabulary and redo it."
Teachable though it may be, O’Brien says that his voice recognition system is not without its quirks. When he wanted to say "new paragraph" to bump the cursor, what came back to him was "the car breath." So he taught the machine to bump the cursor by saying "Brittney Spears."
"I use rock stars’ names a lot because the system won’t confuse them with anything else I say," O’Brien says. "For example, I taught the system that the left side of a parentheses is called Crosby, Stills Nash and Young’ and a semicolon is Arrowsmith.’ I teach it something nonsensical that I’m not going to use otherwise."
L&H will soon bring out an update for automatically entering the correct ICD-9 and CPT codes, one for instructions in both English and Spanish, and one for a bit-mapped signature that will make hand-signing reports unnecessary. And for handling reports made by more than one physician because of a shift change, there will soon be a feature called "protected text" which will stop changes from being made by one physician to notes taken earlier by another.
1. Lerner and Hauspie’s Web site: www.lhsl.com/voicexpress/med.