Widespread health IT adoption in the United States is lagging, despite the promise of EMRs to reduce medical errors, improve quality of care, and overall cost of healthcare. While radiology is already reaping the rewards of speech recognition, many feel the technology holds the key for increased EMR utilization as well.
The United States spends a higher portion of its gross domestic product on healthcare than any other country in the world, but ranks only 37th in its performance, according to the World Health Organization. Many experts agree that the best way to improve healthcare quality and reduce medical errors is to fully deploy EMRs.
President George W. Bush set a goal to provide a portable health record for every American by 2014, and the savings from EMR adoption could exceed $30 billion annually. While Bush’s goal might be considered unreachable, speech recognition could help speed adoption.
Healthcare currently represents 85 percent of the global PC- and server-based speech recognition market—a market estimated to be worth $170 million in 2007 and $207 million in 2008. British market research firm Datamonitor estimates that the North American healthcare speech recognition market is currently valued at approximately $160 million.
Radiology recognizes benefits
Within radiology, speech recognition technology holds the promise of being more than a tool to convert speech to text. By incorporating natural language processing tools and a controlled medical vocabulary, the mapping and distribution of administrative codes and terminologies may enable interoperability and clinician support as part of an EMR.
According to Joe Moore, chief information officer at Radiology Consultants of Iowa (RCI) in Cedar Rapids, speech recognition has made a huge impact on the practice—more than any other IT-based technology.
Moore says he was hired in 2004 to implement systems and technology, such as PACS and HL7 interfacing, to support the core business of professional radiology services. “All have led to improved workflow and efficiency, but nothing has impacted us greater than speech recognition,” he adds. In 2005, RCI, which generates approximately 450,000 studies per year, implemented MedQuist’s SpeechQ for Radiology front-end speech recognition solution across all 12 hospitals and its imaging center that interfaces with eight different RIS. Radiologists working within SpeechQ can access PACS from the program to view films and pull reports.
Moore notes that since implementing SpeechQ, the percentage of exams with a final report delivery in under an hour is now at 87 percent, compared to approximately 7 percent prior; the percentage of exams with a delivery of final report in less than two hours is now at 94 percent, compared to approximately 15 percent before. For emergent cases, 85 percent are delivered in less than 30 minutes and 96 percent are delivered in less than an hour.
“Our biggest naysayers are now our biggest advocates for speech recognition—they never want to go back to the old way,” he says.
An off-ramp to the EMR
John Athas, MD, president of Athas Radiology in New York, N.Y., confirms that most radiologists have finally embraced speech recognition, realizing that without it, work can be time-consuming and inefficient. “Early on, speech recognition was difficult and time-consuming to use, but there have been many improvements over the last five years making speech much more accurate and easy to use.”
As a teleradiology company providing final interpretations, with approximately 30,000 to 40,000 studies per year, Athas wanted a scalable solution that would fit his growing practice.
Using M*Modal’s AnyModal CDS Speech Understanding technology, integrated into NeuroStar’s Virtual Radiology system since August 2008, the practice plugs into any imaging system to establish a comprehensive worklist for their radiologists. “We can add readers, track cases and give the service we need to, in an organized fashion—because radiology is all about that now,” he says.
AnyModal CDS Live captures and comprehends clinical information from dictation and can populate the medical record—creating a kind of “off-ramp” to the EMR. Physicians can elect to self-edit and use the real-time service or back-end approach to send drafts for editing by a medical transcriptionist.
Neurostar essentially provides an overlay image distribution platform that can integrate with any system to route cases and provide a unified worklist for a team of radiologists. He says that what is impressive about the technology is the concentration on natural speaking and its ability to reduce cut-and-paste macros and prefabricated template reports. “Any speech recognition technology can generate a report or insert a macro and notes—M*Modal just does it better by allowing you to speak quickly and naturally, creating a more desirable and accurate radiology report,” Athas notes.
EMR + speech recognition is a valuable combo
With more than 250 doctors providing healthcare services in 24 medical facilities, Fallon Clinic is the largest multispecialty medical group practice in Central Massachusetts. Two months ago, they went live with version 9.5 of Dragon NaturallySpeaking Medical from Nuance and are currently conducting a pilot test to see how it integrates with the practice’s Epic EMR. Lawrence Garber, MD has gathered a mix of 10 “real world” physicians within the practice, including neurology, rheumatology, orthopedics, internal medicine and pediatrics, with a wide range of computer skills and language accents—to participate in the study.
According to Garber, just prior to seeing a patient for a visit, the physician typically reviews the patient’s electronic chart, including prior notes, test results and other clinical information. Then in the exam room, the physician can review and input information with the patient right there. Patients can see the screen with their electronic medical record, which is particularly useful for sharing graphs that show trends of tests results over time.
Subsequently, the physician documents the visit using the strengths of both Dragon and the EMR. “Dragon is remarkably accurate at documenting free-text notes from the first day a physician uses it,” says Garber. During the dictation, Dragon also can activate Epic’s macros to take advantage of the EMR’s power. For example, if a physician dictates “Insert current meds,” Dragon sends “.takemed” to Epic and the EMR will insert the current medication list into the note. “There are a lot of shortcuts that maximize the power of the EMR, yet at the same time, Dragon allows you to do free-text transcription in areas where the EMR is weaker,” Garber adds. “The EMR is great for some structured information, but you also want to be able to have a discussion about the history of an illness, and what the possible treatment plan might be.”
Additionally, utilizing speech recognition in conjunction with an EMR is going to save money. Garber notes that it was costing the practice $10,000 to $20,000 a year per physician for medical transcriptions. With approximately $2,000 in startup costs and approximately $300 per year after that, the return on investment is “a couple of months,” compared to other investments a practice might consider.
“One of my missions is to get docs to use the EMR,” Garber says. “Everyone talks about barriers to adoption and really, it comes down to two factors—EMRs are expensive and they take up scarce physician time to implement. What I am seeing is that by using Dragon for speech recognition, you can quickly recapture the savings you need to pay for the EMR while reducing the amount of time it takes to become proficient.”