Providers report a demonstrable return on their speech recognition investments, such as staff reductions, improved report turnaround times, and increased physician satisfaction, according to Speech Recognition 2010: Vocalizing Benefits, a new report from KLAS.
Still, there is room for improvement, KLAS reported. Providers would like speech systems to learn more quickly, be less initially disruptive to workflow, and better recognize speech patterns of non-native English speakers. Once through the initial phase, however, physicians are embracing it, KLAS said.
The report includes feedback from 355 providers and covers both back-end and front-end speech-recognition technologies. Front-end technologies require physicians to self-edit their transcription as they dictate it, while back-end applications see those audio files sent to a transcriptionist or editor who cleans them up before sending the file back to the clinician for a final review. Clearly, front-end solutions provide the largest potential savings, but they also have the greatest impact on physician workflow.
Some are looking at back-end technologies as a stepping-stone to the more value-generating front-end solutions. "Once the system recognizes their voice on the back end, it's easier for them to adopt a front-end solution," said Ben Brown, general manager of Medical Equipment and Imaging Informatics at KLAS.
Nuance continues to be the "power player" in the speech recognition market, according to the survey. Nuance's eScription takes first place in the back-end section of the report, "a direct result of happy, loyal customers who are able to document direct savings," KLAS said. Dolbey Fusion Speech "performs well" against Nuance with its back-end speech system.
Nuance also "performs well" in the front-end speech segment, taking both first and second place with PowerScribe and RadWhere. MedQuist's SpeechQ is Nuance's biggest competitor in front-end speech software and "is also growing their stake in the industry," KLAS reported. Agfa's TalkStation "has been improving slowly and steadily, but suffers from inadequate support and poor functionality," and takes fourth place out of four.
According to Brown, the government's Meaningful Use program has elevated speech recognitions solutions right up to the C-suite.
"More and more CIOs and CMIOs are getting involved with enterprise-wide speech recognition purchases and decisions because, as we look at the demands that are coming down from Meaningful Use, one of the big themes is improving clinical documentation through structured data," said Brown. "And even though speech recognition doesn't produce structured data, we're seeing more and more of those decisions being made at a higher level."
Brown sees CIOs and CMIOs not only handling the technical sides of system selection and implementation, but identifying and working with clinical "super-users" who can, in turn, champion the technology's use among their peers.
Looking to the future, Brown said that evolving Natural Language Processing (NLP) technologies may someday turn speech-recognition-produced transcripts into the discrete data needed for reporting and quality analysis.
"We're starting to see NLP-type technologies pop up that can take free-text data and then structure it which, in the long run, can deliver structured data to the EMR, enhancing clinical decision support and clinical business intelligence," said Brown. "These are all things driving toward deeper clinical adoption and Meaningful Use."