March 24 , 2008

As You Like It
By Annie Macios
Radiology Today
Vol. 9 No. 6 P. 12

Whether they use front-end, back-end, or combination approaches, these facilities report dramatically reduced turnaround times, cost savings, and often better care.

“Our facility does 275,000 procedures each year, and our radiologists use front-end speech recognition for 85% to 90% of reporting.”

Callout: Dictaphone’s Peter Durlach finds that around 85% of practitioners like to dictate, edit, authenticate, and complete the reports on the front end.

As more imaging facilities adopt speech recognition, they’re not all doing it the same way. Front- and back-end approaches to speech recognition are what vanilla and chocolate are to ice cream: largely a matter of preference. Users in facilities across the country have seen improvements in turnaround time, workflow, and cost reduction thanks to the way they’ve chosen to implement this technology.

“People talk about front-end and back-end technology, but it’s actually not quite so dichotomous,” says Peter Durlach, senior vice president of marketing and product strategy for Dictaphone Healthcare Solutions, a division of Nuance. “It all basically depends on how the radiologist wants to use the technology.”

Durlach generally finds that approximately 85% of practitioners like to dictate, edit, authenticate, and complete the reports on the front end. The advantages are that the turnaround time decreases from days to minutes, and radiologists can complete better quality reports because the images they are reporting on is in front of them.

Back-end speech recognition is often used in instances where radiologists don’t have the time or inclination to edit their work. “This is done by about 15% of users. Sometimes if there is a high volume of similar cases being worked on, the practitioner might dictate several reports in a row, then review each later,” says Durlach.

The decision between front- and back-end speech recognition technology for radiology hinges on which will allow for greater flexibility in how the facility and radiologists want to proceed, according to Durlach.

Cutting Turnaround Time
At Southern Ohio Medical Center in Portsmouth, Ohio, Agfa’s TalkStation software is used for 100% front-end reporting, according to Howard Stewart, the facility’s RIS/PACS administrator. Its speech recognition software is integrated with the PACS, and decreased turnaround time for reports is among the most evident benefits.

“The turnaround time for all studies is just under two hours, and [emergency department] cases are completed in 30 minutes. It’s been a boon to productivity,” says Stewart. He adds that in completing studies on the front end, the radiologists like that they don’t have to sign off on the report at a later time and can feel in control of the dictation. Workflow has also improved because they are no longer producing preliminary reports. “We have removed a whole piece of the workflow by eliminating the distribution of preliminary reports, so the transfer of documents is made much simpler. We only have to focus on the final report,” he says.

Most importantly, having radiologists dictate and review their own reports has reduced the number of clinically significant errors. Such errors are more prevalent when speech recognition is not integrated with PACS, according to Stewart. “For example, a radiologist might be looking at films for patient A but doing a dictation for patient B. By integrating with the PACS, whatever image is displayed is the case he’s dictating on, so there is no confusion,” he says.

Stewart admits that implementing speech recognition takes a lot of commitment from the radiologists, but says that ultimately, they realize the benefits in improved workflow and patient care.

Lynn Nadrowski, the applications system analyst for the University of Michigan Health System’s medical center IT division, reports that her facility has been using PowerScribe, Nuance’s front-end speech processing software for radiology, since 2000. All the systems divisions were phased in over time.

Among the benefits she’s found are reduced turnaround time, improved workflow for radiologists, savings in transcription costs, and the ability to use templates to produce a standardized report. “What we like best is the turnaround time for reports,” Nadrowski says. “A lot of facilities look to speech recognition to decrease transcription cost, but we’ve found the biggest impact to be in turnaround time. It’s raised the bar of expectations in some regards.”

Nadrowski estimates that what used to be a 48-hour turnaround time for more complicated reports has now been reduced to a few hours. “In the first few months of rolling out the system, chest reports were the first ones done on PowerScribe. Once referring physicians began receiving the reports so quickly, they asked why it took so long for other areas to produce their reports and expected them all to be done that quickly,” she explains. Once speech recognition was implemented in all divisions, quick turnaround times became more commonplace.

Link to RIS and PACS
Because the reports are uploaded to the RIS and then sent to an institutional data repository, Nadrowski says they have significantly reduced visits and phone calls to radiologists from people seeking reports or information. “Now staff can look for reports on the institutional data repository, which has been a real plus,” she says.

While reducing transcription costs wasn’t the main reason for adopting speech recognition, the University of Michigan Health System has drastically reduced its transcription staff since going live. They previously employed 12 transcriptionists and now use only three who work on reports for certain specialty areas.

“Our chairman made the decision when we went live that there would be no choice to allow reports to go to an editor,” Nadrowski says. Only a few specialty areas, such as neuroradiology and nuclear medicine, still have a need for transcriptionists; the vast majority of reports are generated using speech recognition and are edited by radiologists. Nadrowski says that by eliminating the ability to choose whether to use a transcriptionist, the adoption was quicker and the benefits remarkable.

Another positive feature, especially because the health system is a teaching institution, is the ability to use the technology to help standardize reporting. “Faculty can request that a certain report be standardized, so we build a template that physicians can use,” says Nadrowski. She notes that it is especially helpful when residents are on rotation because information is always in the same place within a report and in a standard form, helping referring physicians access it in a timely, consistent manner. “In divisions like interventional radiology, you might sometimes have a two- to three-page report, so having the information in a standardized format is very helpful,” she says.

Keeping Up
Radiologists at Washington Hospital Center in Washington, D.C., have also realized many benefits by using a front-end approach to speech processing. Gayle Thompson Smillie, CRA, RT, the radiology director, oversaw the implementation of MedQuist’s SpeechQ in January 2007. Since then, she has seen decreased turnaround time for reports, a large reduction in transcription costs, and minimal workflow interruption.

“Our facility does 275,000 procedures each year, and our radiologists use front-end speech recognition for 85% to 90% of reporting,” says Thompson Smillie. In addition, turnaround time for reports was decreased from 32 to 12 hours, something that pleases the entire medical staff.

Thompson Smillie says the product was easy to implement with little disruption along with good oversight and support by MedQuist. She was also pleased that the program continues to learn the users’ voices beyond the typical hour of voice training undertaken during implementation. She believes that this has contributed to improvements in patient care because reporting is faster and better.

On the financial side, Thompson Smillie reports a $400,000 decrease in transcription costs. “We saw a return on investment in less than eight months,” she says.

Thompson Smillie says the radiologists are satisfied with SpeechQ, and they participate in MedQuist’s users’ group to talk about various aspects of the technology and provide feedback for improvements.

Mix and Match
Toni Spencer, project manager and systems analyst at Providence Health & Services for the Little Company of Mary Service Area in Torrance, Calif., has been impressed with how far speech recognition has evolved since she first encountered it in the early 1990s. At Little Company of Mary facilities, physicians have been using PowerScribe for more than three years. Among them, there is a mix of front-end and back-end users, with four of the 15 radiologists preferring to dictate, edit, and authenticate completely on the front end; one who uses only dictation; and the rest using a mix of self-editing and transcription. The focus is on enabling the physicians to work in a way that complements their personal work style while increasing their performance.

While there are mixed preferences among the staff, the unifying benefit is a decreased turnaround time for reports. “It’s decreased from 32 hours to about eight hours. At first when the doctors were just beginning to use PowerScribe, they were resistant, but now they wouldn’t be without it,” says Spencer.

Adopting speech recognition enabled Little Company of Mary to reduce its number of full-time editors by two; the system’s three sites use five editors/transcriptionists instead of seven. “We’re a high-volume group, so to be able to decrease the number of editors was significant,” Spencer says.

Allowing the physicians to choose their preferred approach shows how versatile speech recognition can be, a feature Spencer enjoys. “One of the most common statements I hear is that you adapt to PowerScribe,” Spencer says. If a physician is particularly busy one day, he or she can choose to complete the reports using an editor, while in other instances, a doctor can completely self-edit and authenticate the report.

Spencer is convinced that facilities need to look at both the cost savings and the quality of service being provided. She also firmly believes that if you are providing a quality service for the clinicians, it improves your business and adds to revenue. “The radiologists are happy with the reports attaching quickly to images. Having that technology is part of the service we provide,” she says.

Spencer is also pleased that PowerScribe can integrate with the PACS. “Now that the physicians see what [speech recognition] can do to improve their workflow, they can’t wait to have it integrated with PACS. Images are available so quickly, they want to have reports available just as quickly,” she says. Once integrated with a PACS, PowerScribe will automatically begin when a physician launches an image.

On the Back End
Angela Mead, transcription supervisor at Floyd Memorial Hospital and Health Services in New Albany, Ind., has been using SpeechQ for approximately two years. While she says it’s designed for use as a front-end system, they adapted it for back-end use because it was a better fit for the preferences of the physicians, who still send 100% of their reports through transcription.

Mead lists the decrease in turnaround time as one positive aspect. “Prior to SpeechQ, we had eight transcriptionists, and turnaround time was 12 to 14 hours. Since implementation, we have reduced to 4.5 transcriptionists for radiology, and the turnaround time is down to four hours. Because of that, I was able to reallocate the additional transcriptionists to med/surg projects,” she says.

The software’s ability to adapt to its users also sits well with Mead and her staff. “If you have some challenging dictators—for example, fast talkers—this piece of technology understands them quite well. I’m amazed all the time how well the program understands even the most challenging speakers.” However, she notes that it is important for physicians to diligently complete the training up front to maximize the program’s success.

Additionally, her outsourcing costs for transcription have continued to drop month after month. “It was a huge goal to reduce outsourcing costs, and each month over the last year, the costs have decreased by $3,000,” she says. And she says that work is now being pulled back from the outsourcing company, which traces the number of minutes per day they were performing transcription tasks. Mead previously outsourced 260 minutes per day of transcription work and has seen that number steadily drop to 116 minutes per day. “It’s exactly what we wanted to happen all along, and it’s great to finally be at that point,” she says.

Mead notes that the radiologists at her facility chose back-end reporting because they do a high volume of procedures and didn’t want their workflow to be interrupted. They like that an extra screen is there allowing them to see the report as they prepare it. “Some radiology reports are very long, and the physicians used to have to take notes in completing them. Now they can simply refer to the screen as they dictate straight into SpeechQ,” she says.

Annie Macios is a freelance writer based in Doylestown, Pa.