Transcription is the easiest way for physicians to get information into the electronic medical record, and yet traditional transcription is often too costly. When it was time to upgrade our dictation system, our vision was to put technology to work and break the productivity glass ceiling for providers and transcriptionists. In January 2008, Seattle Children's Hospital became the first hospital in Washington and the first children's hospital nationally to implement computer-aided medical transcription (CAMT) from eScription by Nuance. CAMT leverages intelligent speech recognition software to convert dictation into highly formatted draft documents that the medical transcriptionist reviews and edits, rather than transcribing from scratch.

Since implementing CAMT technology, Seattle Children's has seen the following benefits:

·        Eighty percent of dictation is voice-recognized.

·        Chronic overtime in transcription has been eliminated.

·        Outsourcing has dropped from over 30 percent to 10 percent.

·        Transcription productivity has increased 61 percent.

·        Reports are available much faster and with outstanding accuracy.

·        Speech recognition engine continuously learns from corrections.

The voice recognition model Children's selected is a "back-end" technology. The drafts that are created by the automated speech recognition engine are reviewed and edited by transcriptionists, not by the dictating provider. This model retains the provider's workflow, requires little training, and accepts foreign accents and rapid speech as long as the provider is consistent. We use the system for all medical dictation and in Pathology. (We are using a front-end voice recognition technology in Radiology; the speakers edit their own dictation.)

From left: Cindy Lewis, CMT, transcription services supervisor; Paula Dascher, manager of transcription services and HIM technology; and Drex DeFord, senior vice president and CIO at Seattle Children's Hospital. courtesy Seattle Children's Hospital

The system implementation spanned four months from start to go-live. During that time Children's staff defined the dictation functionality and built the transcription framework. Redundant voice capture servers were installed on site, with an automated failover system. Now, if one server goes down, phone lines are automatically transferred to the other server and dictation can resume.  

In order to speed up the voice recognition process, the past five years of transcribed documents were sent to eScription. The company used that data to create language models for each physician and for each dictated report type. The automated speech recognition engine uses an algorithm that looks not only at the words used by the physician, but the combination of words commonly spoken by the physician. When the physician begins to dictate, the spoken words are compared to the language model to help the voice recognition engine interpret the dictation. Physicians who are consistent in their dictation are quickly recognized, even if they speak rapidly or with a thick accent. There is no licensing of the technology, and all providers who dictate are included in the voice recognition process.

Dictation workflow improvements

This technology came with new dictation functionality that improved the physician's workflow. Often a physician or less experienced resident will complete the dictation and disconnect. Minutes later the physician remembers a critical element that he forgot to say. When he calls back into the system, enters his ID number, the patient ID and the report type, the system responds, "You recently dictated on this patient. To continue dictating, press *6. Otherwise, to start a new job press 2." This functionality eliminates the orphaned addendum, which takes extra time to piece together with the original dictation.

The system also allows the provider to suspend dictation and complete it at a later time. Unlike other systems that have offered this feature in the past, this system reminds the provider to complete the dictation the next time he signs into the dictation system. As soon as the physician logs in to dictate, he hears, "You have (x) open job(s). To complete dictation, press #0. To continue dictating, press 2." This illustrates the user-friendliness of the system for the provider. It does not force him to complete the open dictation, but reminds him that it is there every time he logs on until he completes it.

We configured the system to allow the user to pause dictation for up to 45 minutes. This fits into the workflow of some of our providers better than suspending the dictation and returning to it at a later time.

Last, we implemented digital hand-held and hands-free dictation devices. We focused this effort in the Pathology department initially, and now have expanded that to accommodate providers who travel to conduct clinic visits across the states of Washington and Alaska. The hand-held dictation devices are docked via USB port to a computer, and voice files are downloaded across the Internet.