A Better Rollout for Speech Recognition
By Selena Chavis
For The Record
Vol. 26 No. 1 P. 24
Attempting to integrate the technology with an EHR? For a smoother implementation, follow these eight tips.
Bombarded with an unprecedented number of requests to learn and adopt new technologies, many physicians are overwhelmed and frustrated. A growing number of studies reveal that in the race to achieve meaningful use incentives, many health care organizations either have not made optimal technology choices or have experienced lackluster deployments.
Simply put, the promise of EHR productivity gains is not coming to fruition the way the industry had hoped. “It’s like the Wild West out there right now,” says Matt Richard, vice president of HIT vendor Mighty Oak Technology, who points out that while incentives have brought a lot of vendors and professionals into the game, there is no real consistency from a workflow perspective. “You have to make sure users meet government requirements, but you also have to make sure you are not hindering a user’s ability to make a living.”
Results from a 2013 survey conducted by the American College of Physicians and AmericanEHR Partners that tracked the opinions of thousands of physicians across multiple specialties revealed that user satisfaction with EHRs fell 12% between 2010 and 2012. Meanwhile, a survey by Black Book Rankings suggested that 2013 would be the “year of the great EHR switch” because users are dissatisfied with the technology’s functionality and how it integrates into workflows.
In recent years, the promise of speech recognition to help the industry overcome productivity woes and create more efficient workflows within EHRs has been a focal point of activity. And while speech recognition has proven its ability to deliver on this promise in numerous environments, many bungled deployments have resulted in poor adoption rates and lack of return on investment.
“There are frustrations voiced from people who have implemented speech recognition,” says Wayne Crandall, CEO and president of NoteSwift. “Sometimes licenses end up sitting on the shelf because physicians find it too hard to use.”
Dale Kivi, MBA, director of business development for FutureNet Technologies, points out that most EHRs have been cost justified, in part, with the promise of speech recognition eliminating transcription expenses and processing delays. “The unfortunate reality is that most EHR approaches simply shift the documentation workflow from a dictation/transcription effort to a physician-centric effort that combines point-and-click activities with self-type or self-edited front-end speech recognition,” he explains. “Although this does indeed eliminate the transcription costs, the painful truth is that it also forces physicians to choose between being their own typists or speech recognition editor. In either case, the amount of time they spend on documenting encounters increases dramatically.”
Often, success with speech recognition depends on setting an implementation strategy. However, many organizations miss the boat because they try a one-size-fits-all deployment approach, according to Juergen Fritsch, PhD, chief scientist and cofounder with M*Modal. “Hospitals purchase speech recognition and assume that physicians will immediately adopt and use it,” he says. “Speech recognition is not the best tool for all tasks. It’s one of many tools: keyboard, smartphone, tablet, and so on.”
While industry experts agree that speech recognition has an important place in the future HIT landscape as a tool for improved efficiency, organizations need to draw from others’ successes and failures to achieve an effective rollout. To reach optimal efficiency, experts recommend eight best practices that can bolster plans to integrate speech recognition into EHR workflows.
1. Develop a workflow-based plan. Crandall says a well-developed project plan often differentiates organizations that have successful speech recognition deployments from those that fail in their attempt. Creating such a plan, he says, requires a full workflow analysis that can answer the following questions:
• What is the existing environment?
• What additional infrastructure is necessary?
• How will speech recognition help meet planned objectives?
“If you have all the details in a project plan and present it to all those who will be impacted, things seem to go so much smoother,” Crandall notes.
Fritsch points out that adequately addressing physician education and training needs is critical to a successful rollout. This process starts with educating physicians on the benefits of using speech recognition and then dedicating the needed resources to train them on how best to use the technology. “You get what you spend your time on,” he says. “If you don’t spend time on making them adopt and use it in the most efficient way, you’re not going to see high adoption rates.”
Crandall says time frames for training and go-live can vary widely—anywhere from a few weeks to several months—depending on physician availability, adding that continuity is important to adopting and adapting. “If a physician has only 30 minutes here and there to dedicate to training, you lose continuity,” he notes. “It’s better if several hours can be set aside at a time.”
Richard says a training plan should be thorough enough to deliver success from the first dictation. “If they don’t immediately say ‘Wow, this thing can work for me,’ oftentimes it’s going back on the shelf,” he says. “Make sure that first dictation is spot-on.”
2. Achieve up-front physician buy-in. Most organizations and practices feature a mix of physicians who either are open to new technology or resistant to change. The key is finding someone to sing speech recognition’s praises and attract others to the cause, Crandall says. “Physician champions are vital to any successful rollout,” he says. “You need practicing physicians to adopt and adapt, and others need to see this take place.”
Richard suggests that physician champions can come from any position, but the choice needs to center on a professional who wants to have control of an effective documentation strategy and is willing to get the job done.
“It always helps to have an open-minded physician, one who has an interest in new technology and how it can help with workflows,” Fritsch adds. “Physician resistance can be overcome with education about how technology can make life easier.”
3. Consider simultaneous rollouts. Because EHR workflows can vary, experts suggest that, if possible, health care organizations may be best served by deploying their EHR and speech recognition workflows at the same time. “You really can’t separate the two. If you want to have speech recognition work in the best way possible, you need to have EHR workflows working the best way possible as well,” Fritsch explains.
Richard points out that there are pros and cons to EHRs and finding where speech recognition fits into workflow and where it doesn’t is critical to a successful rollout. “When we go into a hospital, we look at the EHR and try to find out where the narrative falls in and whether there is anything we can do to add to the efficiency of how it works,” he says, adding that while speech recognition is designed around the narrative, EHRs tend to be more structured from a content perspective. If not properly assessed, this variance can complicate speech recognition use.
As the industry works to bridge the gap between structured content and narrative, the process will become a little more seamless, according to Richard.
4. Enlist vendor involvement. Vendors often have the best insight into what works from an implementation perspective because they have experienced the challenges and successes of numerous product rollouts. For this reason, Fritsch says it’s important for health care organizations to follow the vendor’s best practices. Some of these may revolve around how to train the system to learn voice patterns, how to speak, and when not to speak. “We never give someone a product without help,” he says. “We bundle our products with implementation and adoption services.”
Vendors also can help assess infrastructures to determine any gaps that may exist before moving forward with a speech recognition project, according to Crandall, who points out that when speech recognition is deployed within a local environment, the technology requires a fair amount of resources. In fact, in some medical practices, the technology may not meet the necessary standards. “Performance can be a problem sometimes or the technology simply won’t work right,” he says.
5. Set realistic expectations. Grandiose ideas surrounding speech recognition’s powers may not always translate to reality. Some organizations expect to see productivity gains and improved documentation accuracy in an easy-to-use format. Experts say understanding the technology’s limitations in each of these areas is important for developing a reasonable plan and rollout strategy.
“You can’t expect it to be flawless. The reality of speech recognition today is that it’s not going to be 100% accurate,” Fritsch says, pointing out that while the technology can result in improved productivity and efficiency, health care leaders will be disappointed if the wrong expectations are set.
6. Consider the context. According to Fritsch, some physicians tend to go overboard and want to use speech recognition for everything. The bottom line is that workflow should underscore any decision regarding the technology’s deployment. “You have to determine when and where it makes sense and where it doesn’t,” he says.
Pointing to the difference between structured text and narrative in the EHR, Fritsch explains that speech recognition often makes sense where narrative is used—a progress note, for example. On the flip side, it sometimes can be faster to use keyboard shortcuts for the structured text used in drop-down menus. The environment also is critical to the equation, Fritsch explains, pointing out that a quiet office naturally will get better results.
7. Choose an appropriate model. Front-end speech methods work well in some scenarios, while other environments are conducive to back-end speech processes where a second set of eyes ensures the content’s accuracy.
Kivi says the extra effort required for front-end, physician-centric workflow causes many physicians to start taking short cuts, such as cutting and pasting information from previous reports or not proofing their own speech-generated drafts. “The result is a notable decrease in encounter-specific detail for ongoing care combined with a dramatic increase in documentation errors,” he says. “We all learned in high school that it was a mistake to proofread your own efforts, yet physicians nationwide are now being asked to do just that.”
To combat this occurrence, Kivi suggests changing the workflow to include back-end speech recognition with dedicated editors to create traditional transcribed documents that are then sent through natural language processing to pull the discrete data and autopopulate the appropriate EHR report fields. “The result is physician-preferred document creation methods and compliant structured EHR documents,” he says.
8. Take advantage of enhancement tools. Because speech recognition technology primarily is geared to address output in narrative form, front-end speech processes can result in physicians taking time to copy and paste text into structured EHR fields. According to Crandall, this process can decrease productivity gains; however, there are tools available to help physicians minimize the impact.
For example, there are enhancement solutions on the market that capture free text narrative from speech recognition dictation and automatically populate it into structured data fields via terminology management features. According to Crandall, these solutions, which bridge the gap between EHRs and speech recognition solutions, have been shown to reduce clickable EHR interactions by up to 80%.
Fritsch says speech recognition can be a great addition to any organization if used correctly. “We’ve seen satisfaction levels rise by 50% or more when physicians are using speech recognition,” he says. “The most important aspect is to use speech recognition as an appropriate tool. Use it for the right thing. Don’t think it’s going to remove the mouse and keyboard from your desk.”
Despite an implementation process that requires serious attention to key steps and parameters, Crandall says speech recognition deployment is not rocket science: “It’s just time and commitment … and the right infrastructure.”
— Selena Chavis is a Florida-based freelance journalist whose writing regularly appears in various trade and consumer publications covering everything from corporate and managerial topics to health care and travel.