Description: This is a report on the development of the VRad Custom Structured technology of the VIR company which has been widely used in the clinic system.
VIR is the world’s largest outsource radiology physician services and telemedicine company, we have over 500 U.S board-certified and eligible radiologists, over six and a half million diagnostic imaging studies annually from 2,100 healthcare facilities across all 50 states and around the globe. Our patented clinical and imaging global platform is the largest and most advanced teller radiology, PAC is moving over 1.9 billion radiology images each year at VIR.
Getting the right study to the right radiologists in the shortest period of time is mission-critical, delivering industry-leading turnaround times for emergent trauma and stroke cases means optimizing workflow for greater physician efficiency, and clinical quality v-rod is able to provide both speed and quality for exceptional patient care because of our patented workflows including our custom integration of voice recognition and structured reporting both which I will now demonstrate.
Our medical leadership team in partnership with VIR adds, IT experts have developed a sophisticated approach that reflects our two inviolable rules for voice recognition application. First a radiologists eyes should always be on the diagnostic images and not on the reporting screen, taking one’s eyes off the diagnostic images reduces a radiologists productivity and compromises his or her accuracy.
Second a radiologist should never touch his or her keyboard typing means violating our first principle and greatly increases the risks of inefficiency and inaccuracy while the vast majority of radiology practices today also use voice recognition, their approaches differ fundamentally from VReds and are not appropriate or efficient for optimal patient care to voice recognition in our diagnostic workflow.
Let me first demonstrate the two basic approaches used by most practices direct dictation and pre-populated reports, first a physician can dictate directly to the voice recognition system while this is relatively accurate, it is unfortunately that an insufficient and cumbersome method for the creation of complete diagnostic reports.
First it requires the radiologist to dictate every single word of any given report, this is an obviously inefficient use of precious clinical time and tends to introduce word substitution errors which can be very difficult for the radiologist to identify. When proofreading a report, this approach of direct dictation also fails to leverage the inherent efficiencies designed into most voice recognition software.
I would note that while we have created a more sophisticated voice recognition workflow, we still do on occasion depend on direct dictation for specific additions into reports even here, however, we have a customized approach for improved accuracy. When VIR adds radiologists use direct dictation they’re dictated, texts will be populated in red allowing the physician to see exactly where text has been added so that their attention can be immediately drawn to those lines in which word substitution errors would be most likely to occur.
VRed is also called voice recognition dictionary and has reduced the number of total possible words selection by about 10%, all words that could not possibly have a medical application, such as engine, yeoman and tuna have been removed from our voice recognition lexicon. It is a subtle but important quality assurance technique that helps keep our physicians eyes on the diagnostic images and off of their dictation window.
Pre-populated reports is the second of the basic voice recognition approaches most often used by radiology practices, for example with the use of dictation macros, any two or three word trigger phrase can be mapped to any amount of text that the radiologist would like to use in his or her report.
The logical temptation this approach is to create whole reports which then in turn unfortunately need to be edited, click head normal, this approach seems to be very efficient at first glance, unfortunately, over 90% of all the reports created in this fashion will need to be edited and that violates VReds fundamental rules for voice recognition as editing requires the radiologist to focus on the reporting window rather than on the diagnostic images.
Initially, in developing our approach, we tried to ease the editing burden by modifying these pre-populated macros, in this example, we have created macros where there is a single statement per line, our radiologists are able to identify those lines that need editing.
We’ve also taken advantage of rich text, formatting and pre-programmed mouse buttons, a double click on any word will select it and a triple click on any line will select that line for the radiologist to replace, click atrophy moderate, click ventricles mild, click maxillary sinusitis veer ads, use of programmable mouse buttons enables our physicians to edit with some facility without touching their keyboard or taking their eyes off of patient images .
We have programmed backspace and enter as default mappings to the thumb buttons of our mice, therefore a radiologist can select any word and remove it, select a line and remove it or add lines or remove them as needed.
One additional problem with this approach is that a radiologist can frequently place contradictory statements, within any given pre-populated report. There is inter pedunculated consistent with subarachnoid hemorrhage, in this case the radiologist has made a specific imaging finding and placed it in the report, however the physician may forget to remove contradictory language in the pre-populated report which obviously leads to a great deal confusion and ultimately an unacceptable compromise of patient care, in addition with pre-populated statements, a radiologist is essentially uncertain that all included structures have actually been evaluated.
This leads to second-guessing and the requirement of line by line review resulting in a lot of back-and-forth or tennis tournament bobbing of the head from the diagnostic images to the radiologist reporting screen again.
This is not optimal for efficiency or accuracy with five million studies a year, seconds all add up and with referring physicians and patients waiting for our final and preliminary reports, seconds mean a great deal to clinical outcomes, VRed had to address the drawbacks to direct dictation and pre-populated reports in order to mitigate their time and efficiencies and potential impact on errors and report quality.
So our approach a library of single sentence macros that allows our radiologists to populate a report in a thoroughly customized fashion that leverages their unique approach to structured evaluation.
A VRed diagnostic report is a travel log or real-time documentation of our radiologists progress through the search pattern applied to that given diagnostic image set, this approach completely mitigates the potential for typographical or word substitution errors, it removes the need for any kind of punctuation and it removes the potential for internally contradictory statements.
The pre-population of a statement regarding a structure that the radiologist has not actually evaluated in the course of interpreting that study, so how has this in fact impacted our practice? When we initially implemented VRed’s voice recognition approach, our radiologist demonstrated an average 10% improvement in productivity and a 30% reduction in error rate.
Collaboration between our clinical and operational experts has driven innovation into the Vred workflow that has proven over time to be both efficient and a means of enabling greater quality by removing distractions and frustrations that are inherent in the standard applications of voice recognition, in radiology.
As an adjunct to VIR, Ed’s unique voice recognition approach, we have also integrated a structured reporting program created by one of our radiologists who also happens to have a software engineering degree, the software has been highly customized for our workflow and creates even more benefits for our radiologists referring clinicians and ultimately the patients we serve.
For example, the software program automatically populates the dictation window, when a radiologist opens a case, it imports in real-time, the relevant study information that will be used later in the report header, greatly improve our header accuracy which is ultimately related to improved reimbursement for Vred and for our clients.
Our radiologist now needs spend no time dictating clinical history, elements of the studies, such as contrast administrator or reformatted images the exam title or even the presence or absence of comparison studies.
Again all of this information is pulled in automatically and at the end of the dictation process will automatically populate, the report header in addition, the custom program creates a uniformity of report formatting throughout our practice. A goal that is out of reach for most radiology groups Universal formatting and report appearance is difficult when dealing with more than one radiologist, Vred has over 350, each one has slightly different search patterns that go in slightly different orders, each one uses slightly different diction and every one objects to the pre-population of diction within the findings or subjective interpretive portion of any given radiology report.
Our voice recognition and report workflows allow VRed to circumvent those difficulties essentially pre-populating the dictation window with initial patient and study information allows the radiologist to begin the interpretation of the study without having to read Icke Tait.
The radiologist can begin the report simply in the interpretative portion of the report using veer ads single sentence macros, after a VRed radiologist has gone through the entire search pattern appropriate for that given study type, they simply press a mapped mouse button to process the report through our custom structured report program.
If there is any missing information for that study related to our inability to pull it in via hl7 interfaces DICOM headers or order management system entry by the technologist submitting the report the radiologist is alerted to that fact and can simply populate that information by clicking a checkbox.
The report is now reformatted, you can see the standard header information with all relevant study and patient information imported in such a way as to make the header reliable uniform and optimized for reimbursement, in addition all the statements that the radiologist has submitted with regard to the structural evaluations performed throughout the case are now identified and categorized according to organ system.
This allows VRed to create a uniform appearance to all of its reports without actually requiring any physician to alter his or her own clinical approach, so while their diagnostic search patterns remain unchanged, the report received by our clients referring clinicians are uniform in appearance consistent complete and well-organized regardless of which physician out of our large practice has actually provided.
One of the more important elements of our structured reporting workflow is its error checking intelligence, in this particular case, you can see that I included a statement about the prostate gland in a female patient, the program picks up on that inconsistency and gives a warning to the reading radiologists.
The workflow software will also find errors related to left and right sided inconsistencies, it will actually assess the report for diction that is required for appropriate reimbursement for that given study type, it will even give warnings when it detects that critical findings are described in the report and it will initiate communication with the referring clinician, in those cases and other specific instances.
Such as our customized stroke or trauma protocols, even if diction specific to a critical finding is absent, our radiologists can always initiate communication with our client sites with a single mouse click our 24/7 operation center, which handles all the non clinical communications for them.
Once that request is made, all 350 V read physicians are trained in using both our custom voice recognition and structured reporting. Workflows training is straightforward and requires only an hour and a half of demonstrations to have everyone relatively facile and efficient in applying this innovative diagnostic reporting approach.
The entire V read medical leadership team is obsessed with eliminating all non clinical tasks within our workflow, anything and everything we can do to maximize the time our physicians eyes spend on the images is better for our patients and our practice eyes on the images is an essential part of our philosophy and helps us prioritize our IT investments, simplify our workflows and let our radiologists do what they do best be doctors.
Thank you for reading.