The Latest Word on Voice Recognition Implementation

Neuroradiologist William Keyes, MD, uses the newly installed MedQuist SpeechQ for Radiology VR system at Inland Imaging in Spokane, Wash.Now just may be the right time to add voice recognition — after all, VR has matured to a point that it is ready for prime time and recent feature upgrades make systems more accurate and efficient than ever. But, is your facility prepared for what it takes to ensure successful implementation? Those with experience, from major health organizations such as Mass General to smaller community hospitals such as the 48-bed Goodall Hospital in Sanford, Maine, agree that what is  essential — no matter how tech-savvy, big, or small an organization might be — is a carefully drafted implementation plan.

For years, implementing voice recognition was limited by its technology and lack of support from radiologists. That is changing. In the past couple of years, VR systems have advanced to become an essential component of medical dictation and transcription, as well as an effective workflow management system. They are credited with saving time, money, and work. However, what remain issues for some hospitals and imaging centers are how to turn reluctant radiologists into fans and how to minimize the bumps in the road during the implementation process.

Do your homework

Implementing voice recognition begins with doing your homework, says Paula Crosbie, RHIA, director of health information management at Goodall Hospital. When she joined the staff at Goodall in 2005, she sought to bring radiology IT up to date, which included adding VR. “I read [about VR], attended conferences, talked to vendors and other users,” says Crosbie.

Armed with information and administrative support, Crosbie formed a 12-member multidisciplinary team last year. After looking at systems for over a year, they selected Crescendo Digi--Scribe-XL (back-end system) with a Philips Speech-Magic speech engine, with the option of adding MD Center-XL (front-end system) later this year. “Because we are small, we wanted a vendor who could give us customer service,” she says. Goodall performs about 35,000 imaging procedures yearly.

Joining Crosbie to lead the actual implementation was Cherie Edris, CMT, transcription supervisor, who came to Goodall five months ago. Her homework included the evaluation of workflow, and the standardization of formatting prior to a January installation date. Additionally, nearly all transcription has been brought inside, with outsourcing only as a backup. Transcription will see further changes when the front-end mode goes live in June.

Team spirit

Last year, on the other side of the country, Inland Imaging implemented VR in their three locations in Spokane and Seattle, Wash., and Phoenix, Ariz. Jeff Kelly, IT project director, knew cross-sectional support within the organization was essential. “We created a [16-member] task force composed of radiologists, IT, and operational staff to evaluate several systems,” he says. Inland has 58 radiologists and archives 450,000 images in their PACS.

“We had vendors demo the system to the task force. We had many discussions with their technical teams exploring their IT capabilities and willingness to work with us.” They selected MedQuist SpeechQ for Radiology, also powered by Philips SpeechMagic and made two site visits. “SpeechQ attended our IT task force meetings that met monthly. These meetings allowed the radiologists to provide feedback directly to those in SpeechIQ who can implement requested changes.” says Kelly. “We then had continual, weekly conference calls to address these issues and SpeechQ delivered.”

Know your needs

Implementing a voice recognition system at Massachusetts General Hospital was not simple, but neither was it new. The teaching hospital, with more than 200 radiologists, performs 750,000 imaging procedures annually. Keith J. Dreyer, DO, PhD, vice chairman of radiology, has been involved with voice recognition since its early days 12 years ago. He was one of 45 members of the Boston hospital’s project management team to implement VR. To prepare, they considered the size of the hospital, their workflow, who they serve, and their goal of using only a front-end VR system for all reporting. Key to selecting the right VR system, they also evaluated server hardware, client hardware, PACS and RIS training, and user training.

“Until the entrance of Commissure into the market, we felt the other offerings were incapable of meeting our needs,” says Dreyer. “Commissure [RadWhere] includes workflow orchestration so that we can read cases from numerous hospitals from various PACS and RIS all with one workflow. Also, their real-time decision support for the radiologists and referring physicians is a great aid.”

The one “weakness” MGH addressed was standardizing reports. With the high number of radiologists and the need to supply reports across the country, he says they wanted “the same report style for 400 different types of exams” to ease report reading for practitioners. RadWhere offers options in dictation so radiologists can complete reports with standard free text recognition, standard macros, RadWhere Smart Templates, and RadWhere Findings-only Dictation.

Go for the goal

At the teaching hospital, the primary consideration in adding voice recognition was to improve workflow. “I would only select a vendor that can supply radiology workflow management. Speech recognition is the system that should control which cases your radiologists read, not your RIS or PACS,” says Dreyer.

At Inland Imaging, the goal was much the same as MGH, and they used it as their motto throughout the entire implementation process: “Single work list, with single dictation solution, and single signature queue.”  Kelly says, “We needed to get to a single dictation solution that resulted in a single signature queue for the radiologists — all of this driven from a single PACS work list.”

Radiologists get onboard

Communication drove the buy-in process at Goodall. Crosbie says “communication was most important. We hung posters, talked to individuals and groups.” By doing so, she says buy-in wasn’t a problem. Radiologists learned of the benefits, including much-improved turnaround times.

“Buy-in can be a challenge with some difficult radiologists, but if you let them try a solution that offers more than just speech recognition they will be far more quickly convinced of its value,” says Dreyer. He knows of cases where radiologists did not wait for training and tried the system on their own. At MGH, training was done in groups of four or five radiologists at a time.

Continually selling the benefits of SpeechQ was what Kelly says lead to buy-in at Inland. He touted the “immediate turnaround time benefit to the patient and referring physician or ED. For others it was the use of normal templates to reduce their overall workload. Or the realization that if the radiologists self edits and signs the report, they are done at the end of the day.”

Time is ticking

Going live might best be done in stages. Kelly says, “For a large group like ours, it was necessary to stage the implementation process to provide adequate operational and technical support. We determined for us the best methodology was to implement by subspecialty. Over the course of nine months, each section went live.”

While much smaller, Goodall also implemented VR in two phases. For them, phase one involved education, change of work types, formatting changes. Phase two was training the radiologists to use the system and bringing them on-line. Training needs varied says Edris, from just 10 minutes to several hours.

Measure results

Even though they just went live with VR about three months ago, Edris says Goodall has seen remarkable results. “In the first two weeks, I made a total of six changes on three reports and they were small errors.” She adds that they have seen a 30 to 50 percent jump in productivity. In terms of dollars, Dreyer estimates MGH should save about $1.5 million this year thanks to VR. When it comes to saving time, Kelly says of Inland, “ER TAT [turnaround time] in many cases is only minutes from when the case is completed. Overall turnaround time plummeted from a day to several hours.”

The evidence is in

With just a few months to a couple years of experience using the newest voice recognition systems, hospitals and imaging centers are reaping the benefits of implementing voice recognition. Through the use of a well-designed plan, goals set at the beginning of the process are realized and doubting radiologists are turned into believers. 


Top 10 VR implementation steps
Setting the stage for implementing a voice recognition system is critical to success. Once your homework is done and a multi-disciplinary team is formed, following a plan could make the difference between a smooth road and one filled with bumps.
Assess your organization — people, IT, workflow, and finances. Take an honest look at your organization. Determine your strengths and weaknesses. Who are you and where do you want to go? What does your current technology include and are upgrades needed? Do you have the resources to move forward? 
Address weaknesses that may hinder implementation. From gaining buy-in from reluctant radiologists to standardizing reports, look at what needs to be done before implementing VR.
Eradicate fears through education. Scared off by the thoughts of having to be more hands-on, radiologists need to be educated on the benefits of the second generation of VR and shown the ease-of-use.
Set goals. Determine your primary goal, as well as secondary goals. They may include workflow management, improved turnaround time, and cost savings. Communicate goals throughout the organization.
Make the most of the RFP process. Here is where the homework and self-assessment come in handy. Work with vendors to gain information and meet your needs. In addition to live demos, contact their customers for feedback.
Utilize a timetable. It takes time to make effective, lasting changes and to maximize benefits. Set goals for one, three, and six months, and one year. A timetable includes when to begin the RFP process, training time, and a go live date.
Manage expectations. Even with the best plans, expectations need to be managed. Tell participants that it will not be all smooth sailing, but every resource will be used to ease the transition. Communicate the plans and timetable. Make plans to solve possible problems along the way.
Train the trainers. Vendor involvement drives the training, but it shouldn’t end there. A staff member(s) should be assigned as the superuser or go-to person for continued training and periodic check-ups.
Provide for personnel changes. Make plans to relocate or retrain transcriptionists early in the process to minimize concern.
Measure the results. Once the VR system is up and running, measure turnaround times, cost savings, and accuracy. Look for areas that need tweaking. Also, take pleasure in all that has gone according to plan.