One of the more promising technology developments contributing to healthcare efficiency, outpatient settings notwithstanding, is speech/voice-to-text programming. After all, it’s designed to eliminate keyboard typing and can make it easier to populate electronic health and medical record files.
Speech/voice-to-text capabilities represent a key healthcare innovation with operating software that can trace its medical roots back to the early 1990s. Today’s speech recognition and digital dictation systems may include such advancements as superior sound fidelity, centralized, easy-to-use control/navigation buttons, larger memory cards and improved file security, among other noteworthy features. But to no surprise, they’re increasingly having to connect with EHR and EMR applications as healthcare facilities continue their slow migration toward digital – and virtually paperless – operations.
Outpatient Care Technology queried several executives at the leading manufacturers of speech recognition devices to gauge their outlook on market demand, product development and technology trends in outpatient care facilities.
What are some of the more recent noteworthy improvements in the speech recognition technology you manufacture and why do they matter clinically, financially and operationally?
Peter Durlach, senior vice president, marketing and product strategy, Nuance Communications Inc.’s Healthcare division, Burlington, MA
The improvements found in Dragon Medical 9 — and those being delivered in Dragon Medical 10, (scheduled to be announced in August 2008) — will improve quality of care, make care (specifically the documentation process) more efficient, and make medicine more profitable under current CMS reimbursement regulations.
Version 10 will enable doctors to dictate faster and achieve higher levels of recognition accuracy than ever before. Dragon Medical 10 is specifically designed to enable physicians to dictate a complete patient narrative directly into their EMR system; this capability is critical for physicians. Recent articles in the New England Journal of Medicine and other publications decry the rigid, impersonal template-based documentation of EMR systems. With the addition of speech, physicians are able to find comfort in the thoroughness of medical records created electronically without reducing patient notes to point-and-click templates.
The ability to dictate the history of present illness, review of systems, assessment and plan in a highly patient-specific format is good for peers and referring clinicians. It also translates to higher reimbursement as physicians have the confidence to code at the highest possible level. Our physicians tell us that increasingly Dragon Medical lets them make a 99213 a 99214. Over a few months or a year that can add thousands of dollars to the practice while the doctor is saving time. There are few other technologies of which we know that have such a simultaneously huge productivity and financial impact.
Additionally, we have made improvements that allow Dragon Medical to cater to the needs of any healthcare provider – no matter their accent, organization size, the EMR they use – or don’t use – and regardless of their medical specialty. Version 10 will create an entirely new ‘out-of-the-box’ experience for the many non-native English speakers who practice medicine in the U.S.
Steve Deaton, vice president, sales, Viztek Inc., Jacksonville, FL
The most recent improvement in the speech recognition technology we manufacture is tight integration with an engine that duplicates traditional dictation standards. Most speech recognition modules require that the speaker maintain an open window of the text being transcribed. This is distracting because it requires a radiologist to focus both on reading images and formatting text.
With Viztek's speech recognition offering, utilizing the M*Modal technology, users dictate into PACS without seeing any text. Radiologists are challenged by Viztek to dictate as if they are using the old familiar tape system, free to not declare punctuation or any formatting commands if that is preferred. The result is a speech recognition system that enables the radiologist to focus on the clinical component exclusively.
Transcribed reports should return to the radiologist with greater than 90 percent accuracy, and a synchronized audio file makes editing simple. Words that the system is unsure of are underlined in red, prompting the radiologist to review them specifically. This ensures clinical accuracy without sacrificing speed.
Following suit, the report header is filled in by the RIS/PACS automatically. This not only saves the radiologist time, but improves overall billing success by eliminating inconsistencies between charges and reports. The reports synchronize with the RIS to maintain consistent patient and study information through the entire process, even going so far as to include the CPT codes directly on the reports.
Viztek’s integrated speech recognition product also improves report consistency for a given facility. Radiologists dictate freely into the exams from anywhere in the world via the Web, no workstation-based voice profiles. The reading is done free of any concern for report formatting. The facility maintains control, however, by creating the overall template layouts. Radiologists are free to import their own canned reports, and they will automatically adhere to the format structure management has selected. Key words spoken by the radiologists trigger the system for section headers. The section headers and following text are placed on the page in the order and font selected by the facility management. A section such as ‘Technique’ can be declared to always be present and consistent, regardless of whether or not it is dictated. This consistency ensures rich-text reports from all radiologists.
Marcel Wassink, CEO, Philips Speech Recognition Systems, Vienna, Austria
Philips is exploring new clinical application areas of speech recognition, extending its capabilities and functionalities to address the entire issue of information capturing in healthcare. Our goal is to generate clinically actionable information that is searchable, sharable and can be used for statistical evaluation.
Therefore, the major developments aim towards semantic interoperability of healthcare IT systems to support seamless communications and information flow across the healthcare enterprise. How do we keep a family physician informed about an outpatient surgery for adequate post-intervention control? Can the hospital information system transmit information directly to a patient’s general practitioner? And the other way round: How do we get information from a rural physician’s office into the hospital? Today, the information exchange is still taking place on paper, which results in delays; or worse, in errors from incomplete or missing patient information.
This requires that information is available in a standardized format, which is why at Philips we involve major providers of standardization and coding technologies, to advance semantic interoperability. Speech recognition is now not only able to turn words into text, but turn words into SNOMED or ICD-10 Codes, which serve as the basis for communication exchange between healthcare IT systems.
Another major focus of our R&D activities is to give hospitals the freedom of choice by ensuring compatibility of speech recognition with a variety of IT infrastructures. This resulted in SpeechMagic being the only Citrix Ready speech recognition technology for healthcare.
Together with our partners, we also aim at bringing SpeechMagic to new groups of healthcare professionals: nurses, for example. Our partner VoiceViewer has developed a voice-activated tool for capturing patient information at the point of care, helping nurses to record vital patient data and efficiently upload and store it in the central patient information system. With our partner Crescendo, we brought to fruition North America’s first wireless speech-recognition workflow in an emergency room at the Montreal, Quebec-based Jewish General Hospital.
David Talton, marketing manager, reporting/radiology PACS solutions, Agfa HealthCare, Greenville, SC
Recent noteworthy improvements in Agfa HealthCare’s Reporting solutions include improvements in workflow and in RIS/PACS integration. Our customers demand a variety of workflows, including immediate signoff, remote signoff, correctionist, resident, addendum, etc. We strive to provide a flexible environment for our customers, especially those working in remote locations or from home. For example, we now offer PACS-based signoff of reports. A user may develop a worklist within IMPAX PACS to sign off reports and has the option to synchronize image display, or just look at reports (e.g., over a low-bandwidth connection). Specific workflows vary with solution configuration, and customers should consult with an Agfa Workflow Analyst to determine their specific options.
Rik Primo, director, marketing and strategic relationships, image and knowledge management division, Siemens Medical Solutions USA, Inc., Malvern, PA
[Speech recognition] has become a mainstay technology in the radiology reporting workflow. SR improves accuracy and report availability while reducing cost. ‘Shortcuts’ and ‘Voice initiated macro’’ are replacing many of the original extensive word-by-word dictation process. SR is also used to control the RIS and reporting workstation with ‘command-and-control’ keywords, instead of keyboard strokes. Last but not least, SR can already today completely replace manual transcription in self-editing workflow models. The radiologist makes his own edits in this model.
What about the next generation of speech recognition technology will improve workflow by connecting with other systems (e.g., EHR/EMR, HIS, RIS/PACS, financial and purchasing modules of ERP systems)?
DURLACH: EMRs are here to stay and their roll-out is growing, slowly but steadily. While one recent NEJM study estimated that less than 5 percent of small practices are using EMRs, bigger practices (25 physicians) are at about a 30 percent or more adoption rate. We have about 80,000 physicians using Dragon Medical today – about the same percentage as those using EMR systems. The adoption rate of these two technologies is moving in parallel and each affects the other directly.
Our strategy is to make each generation of speech recognition technology ever more tightly integrated with all clinical systems of record – an enabling technology to the larger system and its purpose. Today, we ‘speech-enable’ a number of functions within the EMR. Physicians can navigate by voice – ‘show me current labs,’ ‘dictate a script’ and other commands can execute multiple mouse clicks of the EMR, saving doctors minutes per each encounter. They can dictate a patient’s narratives into the notes section of any EMR. So they can either spend more time with each patient than they would using the EMR alone, or see more patients and drive practice volume and revenue. In either case, physicians don’t have to stay late anymore completing their notes.
We’re positioning speech recognition as a complementary – and in some cases a must-have – application. In addition to serving its purpose for digital dictation and data input, the next generation of speech recognition technologies will enable doctors to dictate freely with information auto-populating the electronic medical record based on intelligent recognition and system navigation capabilities.
The downstream implications of adapting speech recognition features to drive workflow for the physician while they are working within the EMR will improve the cycle time and effectiveness of billing, coding, and revenue cycle management. Dragon Medical-generated notes are more complete and signed faster so billing can shave days or even weeks of the electronic billing submission process. First-pass denials will drop as coders have the archaeology of the encounter at their fingertips in the patient narrative and don’t have to hunt doctors down to ask questions as the notes are now complete.
The next generation of speech recognition and the EMR will be so tightly integrated that the two will work seamlessly together, ridding the need for a doctor to only dictate where the cursor sits. A number of major EMR vendors are evaluating how their software development teams can use the Dragon Medical software development kit (SDK) so that speech is literally embedded into their application. There will be a broader set of built-in navigation features. Father down the road, the EMR vendors will formally consider how speech should affect the UI of their applications. That’s starting now.
DEATON: Viztek’s integrated M*Modal technology is pursuing the path of data mining and triggered events. Since reports are now truly structured datasets, not just text, it is possible for database queries and actions to take place automatically after a radiologist has approved a report. For example, if a mammogram study is declared to have a positive finding, the system can automatically print the appropriate letters and even evoke follow-ups for necessary procedures to be scheduled. Other systems can be notified for appropriate action, but only when desired. This eliminates the wasted time of personnel reviewing every report, promoting attention only to specific cases. Data mining will help for facilities to track trends in the healthcare environment by performing advanced searches against all previous reports. This information is stored in an advanced database, so other systems such as EHR, HIS, and RIS can gain access for searches.
WASSINK: Philips has always argued that integration is key for the success of speech recognition. We cannot look at speech recognition as an isolated technology as we are dealing with very sensitive, life-critical healthcare information. That is why we do not provide out-of-the box technology.
So, yes, we continue to address the connectivity and integration issues. The main focus is indeed the electronic health record – as we strongly believe that it will clearly revolutionize healthcare information management. In order to allow physicians to capture information in the electronic health record (or any other healthcare IT system) in an accurate, convenient and efficient way, providers must integrate speech recognition into their systems. This will also, in our opinion, speed-up the adoption of the electronic health record.
Integration into RIS/PACS is already standard practice and most of the major providers have deeply integrated SpeechMagic. The research about the integration with ERP systems is ongoing and it is part of our strategy to extending speech recognition applications to all groups involved in delivering efficient and safe healthcare. More than 200 companies worldwide integrate SpeechMagic with their healthcare IT applications.
TALTON: Agfa HealthCare’s reporting solutions are an integrated part of our IMPAX RIS/PACS solution. This creates a single user interface and single database that reduces administrative activities and provides a single set of report statuses. Since our reporting solution is built on the backbone of our RIS, various report distribution mechanisms are available in our Output Management System.
PRIMO: SR will play an increasing role, in combination with mobile technology, templates and multiple choice questions/answers as data input modality. Since patient data input is increasingly becoming a process where the information input is relying on the care provider, rather then on a transcriptionist, SR is slated to do very well as a direct and cost effective technology to input and update the EMR information, even at the patient bedside in the hospital.