When one of two largest hospital groups in the Austin, Texas, area decided to implement PACS, CIO R. Todd Thomas of the Austin Radiological Association (ARA) had some choices to make for his night coverage. Although ARA had mostly blanketed the area with Synapse PACS (FUJIFILM Medical Systems USA, Stamford Connecticut), to which many of its radiologists were accustomed exclusively, the new platform from a different provider would account for 40% of the group’s night and emergency/off-hours interpretations.
Moreover, ARA radiologists were already juggling an influx of fax-transmitted patient information, paper requisitions, an in-house instant-messaging client, and several cross-platform dictation modules. Instead of bemoaning the longstanding interactivity concerns that complicated his life, Thomas saw an opportunity to rethink the entire off-hours ARA reading system.
“It’s strictly there for when our clinics close down and all the hospitals send their cases in for nighttime interpretations,” Thomas says. “During the day, we have radiologists staffed in the hospitals, and they’re either reading off the other PACS or Synapse.”
Thomas tasked Jason Wilson, software engineering manager, with making the new apparatus sing. Together with ARA radiologist Michael Gunlock, MD, who has an interest in software development, the duo (and team members Joseph Garro and Scot Pfuntner) set out to contrive a new interface, workflow, and application to solve the problem. They named their new solution Box Workflow.
“The first challenge that we had was to make it as simple for the radiologist as possible,” Thomas says. “The existing software was based on a very paper-driven requisition system; any manual system like that is prone to errors.”
Under Box Workflow, exams now enter the system through an automatic data-mining process that Thomas and Wilson refer to as the control cockpit. “Since we’ve got control of the data from the beginning to the end, we know exactly where the data came from,” Wilson explains.
Standardized Dictation Capture
A critical innovation of Box Workflow is its standardized dictation-capture function. By working closely with the transcription houses used by ARA’s affiliated hospitals, the team developed a recording module that can be integrated with two types of microphones. “Once the dictation module was completed, we found a way to present to humans only the stuff that absolutely needed to be presented,” Wilson says.
After capturing the dictation, Box Workflow packages it with patient metadata and transmits it to the appropriate transcription house, where its systems automatically import everything for the transcriptionist. This process eliminates errors that would occasionally arise from physicians unknowingly dialing into the wrong system. It also dramatically improves the fidelity of recordings over that of previous phone-based dictation.
“The feedback we’ve gotten from the companies is that the quality of the dictation has gone up using the module,” Wilson says. After the dictation is complete, radiologists may elect either to call in the results directly or to assign the case back to the system. Once the results have been reported, the radiologist or assistant marks the case as completed, and the system archives the study for future reporting or additional follow-up steps.
Aside from seamlessly addressing a variety of workflow issues, Box Workflow also provides useful managerial intelligence in three primary reporting areas: time-stamped interpretation completions, reading turnaround time, and radiologist-specific RVUs. These have helped Thomas and the management team make staffing decisions about high-volume interpretation hours, while providing ARA assistants with better predictions about when they are likely to need backup staff, Wilson explains.
“We’re also able to make sure that we’re meeting our service-level agreements, and we can show the hospitals exactly what they’re getting, as far as turnaround time and response are concerned, from each case they give us ,” he adds. “We’re a lot more in control of our workflow, in terms of where our cases are going, and maximizing that throughput that really is essential to patient care.”
A .NET Shop
ARA is primarily a .NET programming shop, Wilson says, and most of the Box Workflow systems are built on Microsoft® .NET. The front-end clients for radiologist users are based on Windows Forms desktop applications; the assistants’ clients were built in Web applications for the sake of portability. Its back-end programming is based on Microsoft SQL databases, and the Windows Communication Foundation provides real-time data updates (the second-generation iteration of Box Workflow will use Windows Presentation Foundation).
“When we were designing the product, we knew that our business environment is fluid and dynamic, and we wanted to build everything in such a way that it was easy to make changes,” Wilson notes. “Everything can be spun off to be an individual application, if needed, and we’ve done that a couple of times. When some bit of functionality was needed in an area, we’ve been able to take that module out as an executable program.”
One example of a module spun-off as a separate application involved the module that allows a user to search through dictations processed by the system. Assistants use the new application to answer questions about a given case when a radiologist might not be available.
Thomas believes that the solution might likewise morph to accommodate changes in the dictation module as ARA’s client hospitals deploy a new dictation system, but he doesn’t foresee any interest in outside marketing of the software. “I think there are different flavors of these radiology workflow tools, depending upon what challenges the group in question was trying to solve,” Thomas says, citing the development of similar in-house products by groups like Inland Imaging (Spokane, Washington); Riverside Radiology and Interventional Associates (Columbus, Ohio); and TeleRAD IT (Royersford, Pennsylvania).
“I think Box Workflow will be refined over the next couple of years, but there’s no intent to try to sell it to a large health-care vendor or create a separate company and sell it through that route,” Thomas predicts. “It addresses a very specific need for us, and I think the group is content to improve upon what has already been developed for at least the next 12 months.”
Matt Skoufalos is a contributing writer for Radinformatics.com.