Giving Physicians Choices

It's a familiar story throughout health care institutions in the United States - essential medical reports are delayed for days or sometimes weeks, awaiting transcription by an overwhelmed transcription service. Documentation is the life blood of health care - without documentation there can be no communication amongst the health care team and no payment for services rendered by health care providers to patients. In fact, documentation has become a quagmire in almost every hospital, medical specialty and clinical practice.

Facing the Documentation Challenge

Staten Island University Hospital (SIUH), a 785-bed teaching health care facility located in New York City's fastest growing borough of 450,000 inhabitants, is a recognized leader in innovative, technology-based medicine.

The department of radiology produces approximately 4,000 reports per week, with some radiologists generating more than 100 reports each day. Previously, a pool of 17 hospital-based medical transcriptionists handled these documents, supplemented by outsourced transcription services. The scarcity of transcription resources and the increasing costs associated with the burgeoning workload represented potential delays and hindered the departmental and institutional goal of providing excellent service to their customers, referring doctors and patients.

Having had some experience with speech recognition in the past, Dr. Shalom Buchbinder, who manages the department, and Dr. David Hirschorn, director of radiology informatics, carefully set about to evaluate the latest systems available - hoping that the new technological developments would deliver more suitable solutions for the medical field. "We wanted a speech recognition system that really works," Hirschorn said.

A key point was that the department had recently installed the IDX Radiology Information System (RIS), which was the driver for all activities in the department. SIUH thus expected the speech recognition system to be able to seamlessly integrate with the RIS, so that patient data could be automatically inserted into the reports generated and maintained in the RIS system.

Another issue was the need to cope with the radiology department's newly launched residency program that involved residents rotating through their department as part of their clinical experience and training. Integrating these individuals into the workflow is essential as they are included in the clinical reporting process and typically must have their work reviewed and signed off prior to sending reports out to external clinicians.

Aiming for Real-time Documentation

The system was implemented with a front-end workflow for 25 radiologists who now dictate their reports into SpeechQ, a speech solution from MedQuist and Philips Speech Magic, and who receive a real-time transcription of their dictation on-screen while reviewing the radiology images. This allows for immediate editing and signing of the recognized report using the speech application's electronic signature feature while the patient's radiology image is still displayed on the screen. The system then immediately transmits the report to the RIS. As soon as the radiologist has signed the report, it is available for the referring physician and others to view online together with the associated images.

The implementation of this system had an especially big impact on the emergency department (ED). To bridge the waiting time for reports, radiologists used to handwrite a preliminary report and fax it to the ED. Today, the ED physicians can access a report as soon as the radiologist has signed it electronically.

Now that reports are instantly available, calls from referring physicians to the department have dropped off dramatically. "Ninety percent of our reports are being done immediately and the remaining 10 percent come back to the radiologists in under 45 minutes from their transcription pool, ready for signature," said Dr. David Hirschorn, "which means our radiologists have more time available for clinical activity and are more productive, which, in the case of some of our radiologists, equated to an additional hour or more per day!"

Insert Call Out Box, Explain Front-end and Back-end Solutions

Back-End Speech Recognition
In the past, physicians typically used a back-end process in which the author dictated either by phone or into a handheld device and then passed this dictation on to transcriptionists for manual transcription. Back-end speech recognition applications mirror this type of workflow while adding speech recognition, enabling the transcriptionist to edit the automatically generated text instead of having to type the entire dictation.

Pros for back-end speech recognition:

• ideal for hospitals that don't want to involve physicians in the report transcription process
• doctors can work the same way as before (no change in behavior)
• the transcriptionists' productivity levels increase

Cons for back-end speech recognition:

• if reports are corrected by transcriptionists, then the doctors are still dependent on transcriptionists
• reports available after dictation, not during dictation
• if an interruption occurs during dictation, there is no immediate indication on-screen as to where to continue the dictation

Front-End Speech Recognition
Front-end speech recognition gives physicians more flexibility and control, but also more responsibility for creating the reports due to not using a transcriptionist. Instead of having a dictation recognized by the speech recognition server in the background, the text is presented to the physician in real-time, either while they dictate or immediately after they have completed the dictation.

Pros for front-end speech recognition:

• instant report production and availability
• clinicians are able to make corrections immediately - less likelihood of report errors
• usage of auto texts and templates ensures more standardized, precise reporting

Cons for front-end speech recognition:

• may involve additional report creation time, especially with physicians who correct documents themselves
• requires some change in clinicians' behavior

User Satisfaction

Sometimes the most frustrating aspect of implementing a speech recognition system is enrolling the speakers. This can often take vital time away from busy physicians, but the radiologists at SIUH were able to start using their system after only a two-minute enrollment. From this point on, the user's productivity and accuracy continues to increase since the system constantly adapts to each unique vocal profile.

Some radiologists use certain standard texts in many of their reports. These doctors can now use auto texts to insert entire pre-defined text blocks by voice command. The auto texts can also be designed as templates with "fill-in blanks" that are used to insert measurements, dates or other elements necessary to complete each specific report.

The Results

Within six months of implementation, more than 90 percent of the hospital's reports were being processed using speech recognition. The remaining 10 percent are returned within 45 minutes from the transcription pool that is now able to cope with the workload - no more outsourcing necessary.

• report turn-around time has been reduced to minutes
• reports can be signed off while images are still on display
• availability of reports promptly following an examination ensures higher accuracy
• improved transcription staff efficiency, resulting in savings of over $400,000 annually

In addition to the back- and front-end choices, physicians are offered a selection of templates, macros and auto texts that result in greater efficiency and productivity.