Beyond Automated Transcription — Templates and Structured Reports Maximize Speech Recognition’s Value
By Beth W. Orenstein
Vol. 15 No. 5 P. 12
The days of radiologists dictating their reports and sending them to transcriptionists are becoming a thing of the past. Today, it’s all about speech recognition.
“Speech recognition for transcription has been around for quite awhile in radiology,” says Russ Cardwell, vice president of imaging solutions for M*Modal. It’s not only expected now but also employed to reduce costs and speed report turnaround times, he says.
While speech recognition programs have become the new norm, the transcription process itself hasn’t changed, says Steve Retz, manager of the eHealth documentation solutions business at 3M Health Information Systems. Radiologists still record their findings and interpretations, only now the computer transcribes the dictation and sends it back to the radiologists to edit and sign, he explains. “Those fundamental steps are still there,” he says.
What’s changing is that speech recognition programs and related tools have been developed to condense the time it takes to complete radiologists’ consultative reports, according to Retz. Automated features and the ability to include images and/or references improve the accuracy and value of the information included in those reports. “The result is that the ordering clinician gets the consultation much quicker and can act on it sooner,” Retz says. “Instead of it taking weeks or days to get results, now it can take hours, which improves patient care.”
The evolution of speech recognition and related technology in radiology dovetails with the movement from volume-based to value-based care, adds Christy Murfitt, director of solutions marketing at Nuance Communications, who has observed the evolution firsthand while working for the company for the past 14 years. The latest speech recognition solutions, such as Nuance PowerScribe 360, can improve the timeliness and quality of the communications between radiologists and referring physicians, she says. “And for the radiologist to be a partner in the whole care cycle, it requires clear, timely, and quality communications with those referring physicians.”
One advance that enhances the value of speech recognition software is combining physician dictation with structured reporting templates. The templates help ensure quality because they outline “what you need to include in this report and provide shortcuts as to how to do that,” Murfitt says.
Before the Affordable Care Act (ACA), radiologists could earn financial incentives from Medicare by including at least three of 74 individual quality measures in the records of at least 80% of their cases. Established in 2007 and extended through 2009, the Physician Quality Reporting Initiative had few takers. The ACA turned the initiative into the Physician Quality Reporting System (PQRS) and the incentive into what amounts to penalties. Beginning in 2015, radiologists can be docked 1.5% for noncompliance; the penalty rises to 2% in 2016 and beyond. (The penalty is applied to all Medicare Part B billings for that physician, not just the charges where PQRS data were reported.)
Structured templates can help with PQRS compliance, Cardwell says. “If an element is missing, the system alerts the radiologist. Say a complete abdomen ultrasound is ordered; nine findings are required for the report to be billed as a complete abdomen ultrasound,” he explains. “If one element is missing, the software tells you only eight were reported on, which would allow only for a partial ultrasound billing.”
He says the system doesn’t tell radiologists that the exam should be billed as a complete abdominal ultrasound, but if that’s the exam that was ordered, the system makes it clear what elements must be included for proper reimbursement. The system can help productivity because exams don’t need to be sent back to the original radiologist for additional reporting when someone notices the missing element or to a second radiologist to reread so that all the elements can be included, according to Cardwell.
He says one of M*Modal’s clients associated his or her templates down to the exam code, thereby enabling the correct report template to load for every case. This enabled the customer to print on the report, “This qualifies for PQRS #146.” Later, the customer could attribute approximately $180,000 in additional reimbursement in combination from his or her work in associating that exam template and the use of M*Modal’s speech recognition products, he says.
Templates typically are customized from one radiology group to the next. “We provide the application, the tool for them to create the template,” Cardwell says. “Typically, the harder part is to get the 10 radiologists in the group to agree on the text contained in the template. They can agree by practice or a single user can have it set up a certain way.” M*Modal’s speech recognition software provides flexibility “because we know that getting more than one person to agree on a template is not always easy,” he adds.
Sophisticated users of the speech recognition add-ons may have as many as 200 different templates to choose from, which doesn’t help workflow. To solve that issue, speech technology solutions enable radiologists to quickly pinpoint content-relevant templates. For example, M*Modal makes available to its users “smart templates” or content-sensitive templates. “If we know by the exam code it’s a chest X-ray, the radiologist will only pull in the default template that matches the exam code and any other related templates that would match a chest X-ray,” Cardwell says. “So if there’s an ankle fracture, we’re not going to pull that template in. Instead of trying to match potentially a list of 200, we are going to default the right one to that exam code, and we’ll make available the other ones that match that modality and body part. We have the option to show them all, but this simplifies their user experience.”
The structured templates also include space for any information required not only for the particular exam to be reimbursed but also for accreditation. Starting this July, The Joint Commission that accredits hospitals and ambulatory health care organizations will require radiation dose utilized for CT exams to be recorded in all patient records. With the latest speech recognition programs and add-ons, radiation dose can be integrated directly from the imaging equipment into the radiologist’s report.
“Being able to pull the dose directly from the modality and include it in the report saves time for the radiologist and improves the document quality,” Cardwell says. Murfitt agrees: “The old way is to get the dose measurement off the machine and then dictate it as part of the report. Integrating it directly into the PowerScribe report eliminates what had been an error-prone process where someone writes it down from the machine and hands it over. It also saves time for the technologist and the radiologist.”
Another advantage of combining speech recognition and templates is standardized records. For example, all breast exams have a dedicated BI-RADS entry, which standardizes breast imaging findings terminology and report organization along with the classification system for mammography, ultrasound, and MRI of the breast. “The templates guide the radiologist through the information that needs to be reported, making it easier for the referring clinician to find and retrieve what they need from it,” Murfitt says.
Some radiology reporting software also allows radiologists to capture a portion of the image to which they are referring and include it directly in their reports. “That’s something that radiologists really struggled with in the past,” Retz says. “Now the radiologist can say to the referring physician, ‘Here’s an image of the patient’s collar bone, and you can see the fracture.’” Being able to include images also helps if the radiologist wants to share his or her findings with a colleague for a second opinion, he says.
How well radiologists seamlessly integrate voice recognition workflow with other systems such as PACS is the future, says Andrew Chow-Woon, eHealth solutions manager for Agfa HealthCare. The need is for unified systems that are based on real-world workflow rather than voice recognition. It helps that voice recognition systems allow radiologists to send or import relevant information from diagnostic systems and embed it in their reports, he says.
For example, with today’s voice recognition systems, radiologists can measure tumors on the screen, right click, and have the information inserted directly in their report. The measurements automatically are populated into the report, reducing the possibility of errors. “If all you have to do is right click and insert it into the report, it takes away the possibility of errors that can occur with manual entry,” Chow-Woon says.
With the latest software, radiologists can import not only images but also information they’re using to interpret those images. With some programs, the radiologists can hyperlink words in their reports that allow their colleagues to research their comments, if desired. For instance, if a radiologist is looking at a chest X-ray and detects small lung nodules, and he recommends that the patient’s physician follow the Fleischner Society pulmonary nodule criteria for follow-up, he can hyperlink to the recommendations so the physician can see exactly when follow-up imaging is needed based on the nodule size and the patient’s age and risk. “This is the next big wave that has started already with image sharing,” Murfitt says.
Speech recognition software also is being combined with critical findings so that if a report contains a finding that needs to be acted on immediately, it doesn’t get lost. “A couple of times a day, radiologists will come across critical findings that require urgent communications to the attending physicians,” Murfitt says. Nuance PowerScribe’s critical test results management solution triggers a cascading communication protocol, including an alert that immediately notifies the attending physicians and documents delivery and receipt. If the alert isn’t acknowledged within a time frame set by the institution, it automatically will be forwarded to administrators or someone at the next defined point of escalation, who will ensure that appropriate providers respond.
M*Modal’s system has dozens of critical findings that it automatically looks for in real time. “But after you get past the first five or seven that everyone agrees to, there is great disparity,” Cardwell says. As a result, organizations can choose which critical findings to alert in real time or which to not alert, he says.
3M’s solution has a critical phrase recognition app that allows the radiologist to highlight keywords in the document that are critical and shouldn’t be overlooked. “The radiologist can determine what those keywords are,” Retz says. “It’s configurable.”
The radiologist uses it not just to highlight words such as “mass” or “malignant” but the small words that accompany them and that are critical because of their significance and meaning. For example, if the radiologist is reporting “no mass,” the word “no” is highlighted along with the word “mass.” The addition of the highlights enables the radiologist to double-check the report for accuracy before sending it, Retz says.
Samir Patel, MD, chairman of radiology for Elkhart General Hospital in Indiana, a 325-bed hospital that serves more than 19,000 patients per month, finds that taking full advantage of all the reporting and functionalities of its Nuance PowerScribe 360 platform has enabled his department to devote more time to patient diagnoses and management. Since the system was installed in 2012, it has helped to reduce report turnaround times from three or four hours to less than 15 minutes, he says. Also, it has eliminated more than 1,100 hours of transcription time per year, saving a considerable amount of money.
Patel likes that PowerScribe 360 also improves the department’s workflow. When beginning a report, the radiologists can enter data into templates. Once done, the radiologists need only to hit the icon and review the information. “Most of the time, we have no edits and go right into dictating the findings and impression,” he says. The streamlined process reduces turnaround time. “It’s about all the things you don’t have to dictate,” he adds.
By utilizing structured reporting and autoloading templates that are mapped to specific procedure codes, radiologists know exactly what they must include for their reports to be complete and to qualify for reimbursement. “With PowerScribe 360, the radiologists know they’re using the exact language the Centers for Medicare & Medicaid Services is looking for,” Patel says.
Customizing Speech Rec
Cardwell finds that different radiology groups are at different stages of adoption of the available speech recognition technology and add-ons. Some adopt speech recognition primarily to reduce transcription costs, while others want to perfect it. “Then they start asking, ‘How do I standardize some of my report text so my referring physician can more quickly build a treatment plan for that specific patient?’ That’s when you start to see more sophisticated use cases.”
The technology is available, Cardwell says, but usually it’s a matter of the amount of time and energy the group wants to leverage report templates. A customer’s exam code dictionary could include several thousand different codes. To associate them one by one to a template takes time. However, the benefits of productivity and improved report quality will pay dividends every day if they do, he says.
Murfitt says the key to accuracy and productivity is using proven technologies to streamline workflows while providing the foundation for the future, particularly patient-focused and quality-driven care. “As we look to the future and leverage modern technologies such as mobile communications and the cloud,” she says, “communication and collaboration becomes easier and more efficient than ever.”
— Beth W. Orenstein is a freelance medical writer based in Northampton, Pennsylvania. She is a regular contributor to Radiology Today.