|
|
October 16, 2006
Doctors are known to be creatures of habit, adverse to changes in their routines. However, it may be possible to convince them that speech recognition can be beneficial. Speech-recognition technology has the potential to reduce the time physicians spend writing patient notes; more importantly, it has the potential to reduce turnaround time from dictation to having the note in hand. Doctors need to know that speech-recognition technology can improve patient care because—through the use of front-end speech recognition—records can be available almost immediately. Back End, Front End, or a Marriage
of the Two “A hybrid model for dictation is the way to go,” he explains. “Giving the physician a choice on front end, back end [background], or a combination of the two makes for an easier adoption ratio. There will be a bell curve of users that aren’t so good at the technology and those who may not be good candidates for speech recognition in general.” As with any technology, there will be those who readily embrace it while others will balk at a change in routine. “With some of Nuance’s core products, the user can bypass transcription completely and that could be the primary mode of dictation—they would edit their text themselves,” Fallati says. “For others, it doesn’t make sense for them to be doing their own editing and you would want that dictation sent to the transcriptionist for completion.” The choice is up to the physician, the facility, and the HIM department. “When introducing a new technology, you need to offer flexibility to suit various physician styles,” Fallati says. In front-end speech recognition, the provider dictates into the system in which the words are displayed after they are spoken, and the dictator is responsible for correcting misrecognition and signing off on the document. In this system, the document never goes through an editor. Back-end speech recognition is the process by which the clinician dictates into a digital dictation system, the voice is routed through a speech-recognition machine, and the draft document is routed— along with the original voice file—to a medical transcription editor who verifies the accuracy of the draft, finalizes the report, and forwards it for signature. Lee Stephen, a programmer at Custom Speech USA, says back-end speech recognition should be easy for any physician with dictation skills. “The only difference is that the audio file is now transcribed first by a machine, not by a person,” he explains. “Front-end SR [speech recognition] is more difficult for some physicians [whose] word error rate is high and many corrections are required. For other physicians, where there is low error rate, there should be less resistance to adoption of speech recognition. However, it should be remembered that even a 5% error rate means an error about once every two sentences. If there is a short report with a dozen sentences, that still means that the physician may have to spend several minutes correcting the errors. For a busy physician committed to patient care, that is still a distraction.” Stephen believes physicians are most concerned that editing speech recognition errors will take time away from dealing with patients and other medical issues. “Consequently, a server-based system—where the speech recognition prepares a ‘rough draft’ for editing—and correction by a transcriptionist makes a lot of sense,” he says. Front-end speech recognition, Stephen says, is useful for completion of short reports in STAT settings where there is minimal correction required by the physician and/or there is a need for rapid turnaround time. “Where the report is longer or required turnaround time is longer, there is less need for front-end speech recognition,” he says. “Back-end speech recognition can be used, relying upon the transcriptionist editor to create a polished draft after correcting misrecognitions.” Fallati says, from a physician perspective, front-end speech recognition allows for self editing and that may be the best option to offer. “But, the easiest first step for implementation is the back-end model. It really doesn’t change the everyday dictation methods a physician is accustomed to and all a hospital has to do is bring in good transcriptionists,” he says. “Many times with the back-end programs, the physician isn’t even aware that he is utilizing a speech-recognition program.” Front-end and background speech recognition have different strengths, according to eScription Director of Marketing Lauren Richman. “Front-end is well-suited for certain specialties where physicians wish to do the editing of their documents themselves,” she explains. “Background speech recognition is well-suited for [an] enterprisewide application where the organization and its physicians wish to increase transcription productivity and reduce costs overall, with minimal change to clinicians’ behavior.” Making the Sale There are those clinicians who are clamoring for the technology because vendors have told them it will replace transcription. But those in the industry know that if there is a minute of extra time and effort on the doctors’ part, they will balk at the technology. There has to be a seamless transition for the technology to be accepted. Even with the technology, Fallati says, a two-minute dictation is still a two-minute dictation. “What needs to happen is that software needs to accommodate narrative detail in a more streamlined, contemporary fashion than plain old dictation and typing,” he explains. “We can embed data mining tools that can read narrative text, extract info, and that will reduce the amount of time a physician spends in documentation.” There are technologies available that allow a clinician to speak a “trigger” word that could potentially generate a paragraph of narrative. “A lot of doctors say, ‘My documentation is very consistently similar,’ and with the new technologies, they may be able to append a narrative piece on top of an existing template,” Fallati says. Discharge summaries, sometimes the bane of a physician’s existence because they are literally redictating material that was done previously, could be merged with the new speech-recognition technologies which would reprocess the history, integrate it with the discharge summary, and potentially save the physician hours of dictation time. “You need to give flexibility on the user side in order for adoption to be mastered,” Fallati explains. “You can’t force-feed a one-size-fits-all system into a facility. The physician has to be shown how these technologies benefit them and how it will reduce their behind-the-scenes time when they are not with patients.” Richman says physician reluctance often stems from a concern for quality of care and an aversion to learning how to use a new system. To sell physicians on the technology, she recommends demonstrating accuracy, consistency, and speed when addressing quality-of-care issues. “Physicians are often very busy and see the time requirements of retraining unacceptable,” she says. “If you are recommending a new system, develop a plan that makes learning how to use the system as simple and as quick as possible.” Back-end speech-recognition solutions can be installed nearly transparently, Richman says. “Physicians can begin using the new system without any retraining at all,” she notes. Necessary Skills What about those physicians who speak heavily accented English? Fallati says that if their voice is difficult for a speech program to recognize, it is likely it would be as difficult for a transcriptionist to translate. “Truly, though, a heavily accented voice is no longer the disqualifier that it might have been in the past. The technologies are more adaptable,” he says. The obvious benefits of speech-recognition technologies are the streamlining of turnaround time and the flexibility that can be built into the software. “The programs can be tailored to mesh with specific disciplines and key words can ‘cut and paste’ phrases or paragraphs from one document to another,” Fallati explains. “If a physician is given some ‘at the elbow’ training time, they will likely be comfortable with the new technology and adapt rather quickly.” According to Fallati, radiologists are heavily into using the front-end speech-recognition systems. “They are accustomed to this technology,” he explains. For back-end, server-based recognition, physicians can employ standard dictation skills. Stephen says transcriptionist editors need to be aware of the differences between machine and human transcription. For example, a transcriptionist may make spelling errors, but the machine never does. It always spells the word correctly, but the problem is that it has recognized the wrong word; for example, transcribing “art” instead of “heart.” In addition, these misrecognitions are random. “A back-end, server-based system needs a way to display potential random errors and flag them for review by the editing transcriptionist,” he explains. “One approach is to compare synchronized output of different speech engines, making available the audio-linked session files for each engine. The transcriptionist can listen to the audio to determine whether there is a misrecognition that needs correction. So the process emphasizes ‘word-check’ rather than ‘spell-check.’” Richman says systems can be designed so physicians can use them with little to no retraining. “However, if a physician is looking to improve his or her dictation—in order to improve the quality of the first draft document produced by background speech recognition—the most important skill that providers must learn is to organize their thoughts,” she explains. “Speech recognition has advanced to the point where as long as they don’t slur or mumble their words, physicians don’t have to change the way they speak for the background speech recognizer to understand them.” Richman also says physicians should provide all the demographic information at the beginning of the dictation and group their information based on the formatting standards of the organization. “Jumping back and forth between demographics,
diagnosis, medication, etc can reduce the quality of the draft produced
through background speech recognition and also slow down the review/editing
process by MTs [medical transcriptionists],” she says. “To
continually improve, physicians should review the finished documents
to see how their dictations have been changed and ordered to best understand
the documentation standards of their facility.” “Some disciplines are more conducive to speech recognition,” Fallati says. “Surgeons, for example, tend to walk out of the operating room, pick up a telephone, and start dictating the notes of the procedure. It may not work for them to have to walk down a hall, sit at a computer, and dictate notes. But others, like physical therapists or another practice that isn’t well-supported by the hospital’s infrastructure, are really showing readiness for this technology. They are the ones right now who are lugging home paperwork, dictating, editing notes themselves—speech-recognition software would be beneficial for them.” — Robbi Hess, a journalist for more than 20
years, is a writer/editor for a weekly newspaper and a monthly business
magazine in western New York.
|
 |