Comprising the PACS team at the Veterans Administration Medical Center in Baltimore are (from left) Eliot Siegel, MD, chief of imaging, Khan Siddiqui, MD, medical informatics fellow, Steve Severance, research assistant, Ryan Moffitt, research assistant, Bruce Reiner, MD, director of research, and Steve Brower, research assistant.

The next time you meet with your enterprise's key decision-makers to weigh the wisdom of investing in PACS, and a foe of that acquisition attempts to portray that proposed purchase as too costly with too little return potential and too much downside risk, you may want to invoke the example of the Veterans Administration Medical Center in Baltimore, one of the earliest adopters of PACS.

In 1993, the Baltimore VA Medical Center installed an enterprise-wide PACS that, within 2 years, was extended to include three other VA hospitals and 10 outpatient centers spread across 50 miles; it entailed integration with a radiology information system (RIS), a hospital information system (HIS), and an emergency medical record (EMR)all at a time when there was no commercially available DICOM interface technology to smooth the way.

Yet, despite the potential for this PACS adventure to fail, the center's investment was quickly validated by just about every measurefrom reducing operating costs to improving the quality of care, from making radiologists and referring clinicians more productive to freeing up needed physical space in the radiology department so that new income-generating services could be added.

"We were the first kid on the block to take the plunge with PACS, but the more intriguing aspect is that we were able to be successful at taking that plunge," says Bruce I. Reiner, MD, director of research for the VA Maryland Health Care System in Baltimore (to which the Baltimore VA Medical Center belongs) and an associate professor with the Diagnostic Radiology Department at the University of Maryland School of Medicine. "I think a big factor in that success was our ability to anticipate problems. By anticipating them, we were often able to have solutions ready, so they could be dealt with early on, before they had a chance to grow into serious threats and sabotage the entire effort."

GOOD REASONS FOR PACS

The Baltimore VA Medical Center began considering PACS back in the late 1980s, when the technology was still in its infancy. Driving the enterprise to adopt it were several factors. One was rising concern over the costly inefficiency of work flow. Another was the difficulty of sending films to other VA hospitals in Maryland and then getting them back in a timely manner, a problem that forced the hospital to frequently repeat examinations. There also were growing concerns over the medicolegal liability associated with those lost or missing imaging studies.

However, the biggest impetus was the VA's announced plans to merge the Baltimore facility with three other Maryland VA hospitals into a single health care delivery system, with the radiology department at Baltimore alone providing imaging services to all. But before the merger would take place, the Baltimore campus was slated to shut down and take up residence a few miles away in a brand-new, state-of-the-art facility across the street from the University of Maryland School of Medicine.

"It was expected that the move and the merger would result in a significant increase in demand for imaging services," says Eliot L. Siegel, MD, chief of imaging for what now is known as the VA Maryland Health Care System and vice chairman of information systems for the Diagnostic Radiology Department at the University of Maryland School of Medicine (which supplies attending radiologist coverage to the VA hospital on a contract basis). "Coupled with the fact that we would be establishing a virtual' radiology and nuclear medicine department, we quickly began to see PACS as a potentially pivotal piece of technology to make all this possible."

It helped that certain senior administrators in the VA system had decided to pull out all the stops for acquisition of cutting-edge information technologies. Consequently, when it came time to buy a PACS, the Baltimore VA Medical Center radiology department was permitted to include the purchase price as a line item in the cost of construction of the new facility rather than having it come out of the radiology equipment budget.

Still, PACS was not an easy sell. Says Siegel, "There was some trepidation within the VA's hierarchy in Washington about taking the lead with what then was still a very new technology that hadn't been fully proven. Fortunately, we had a champion within the upper levels of the administration who was a visionary, and he helped make the case for PACS by convincing enough of the right people that the technology was sufficiently developed that we would be on safe ground with it. Even so, after authorizing the purchase, the hierarchy at the last minute had a change of mind. The only thing that saved the project was our insistence that we had no viable image distribution-and-archiving alternative to which we could resort were PACS to be nixed."

The PACS was acquired in 1991 at a cost of $7 million. It was delivered in 1993 and came with 40 clinical workstations. The system was designed around a centralized architecture in which images were stored using a high-speed central server. These images were retrieved, on-demand, via direct fiber-optic connections to the video memory in each computer workstation. In its original form, the system was Macintosh®-based. Then the vendor sold the rights to the technology (along with the VA's service contract) to another vendor, which, after a merger with two other companies in the defense industry, decided it no longer wanted to be in the medical systems business and so resold the PACS product rights to yet another vendor. The last one to receive the ball introduced a number of enhancements to the technology, including a shift of operating systems from Macintosh to Windows®. Accordingly, the VA Baltimore in 2000 exchanged its Mac-based devices for those developed by Bill Gates and company.

"There were some challenges involved in undertaking this last upgrade," Siegel recalls. "It was like switching to a completely different PACS. The network configuration migrated from a proprietary cyber-based network to a more standard Ethernet-based network. The software was different and the workstations were different, so users had to be retrained."

As currently configured, the PACS employs a central architecture, 288-gigabyte server on a storage area network, which provides 3 months of image storage and an image retrieval speed of approximately 1.5 seconds for an 8-megabyte CR image. After 3 months, images are stored on a pair of optical archives. The optical archives together have a capacity of 15 terabytes, or enough room to hold 10 years' worth of images in compressed form.

SMART MOVES

Looking back, Siegel says the single smartest thing the department did during implementation of PACS was devote a generous amount of resources to achieving full integration with the enterprise's RIS, HIS, and EMR.

"We felt that this integration would be key to maximizing the efficiency and clinical effectiveness of PACS," says Siegel. "We were right about that. Other, non-VA facilities that have installed PACS but not paid close attention to integration have not seen nearly the productivity gains as we have."

One of the primary effects of integration was the gained ability to transfer ordering information and radiology reports electronically "so that the radiologists would not be constrained by waiting for the paper imaging request forms and could operate in a more efficient paperless environment using automated work lists," Siegel says.

Integration with the EMR, HIS, and RISplus the various imaging modalitieswas a bit daunting for the reason that, in 1993, DICOM and all that it could make possible was still on the horizon.

"We had to create our own custom interfaces," says Siegel. "It took about 2 months to get those interfaces working properly once we had developed them. In retrospect, we probably should have tested them a lot more extensively before rolling out with them. Had we done that, we would have faced less pressure as we went about trying to resolve the problems that cropped up with them.

"We also made the mistake of not providing a graceful way to recover from errors that resulted from deficiencies in the interfaces we had devised for the modalities. The modality interfaces were not particularly reliable, so we had a number of times when images would not come across properly. Also, a lot of the modality interfaces suffered from unacceptably slow transfer speeds30 to 40 minutes to send a CT scan of the chest. We eventually solved the problems and now have very fast transfer rates, but, ironically, now that we're doing 16-slice CTwhich can involve up to 2,000 images in a single studywe're once again having the same kinds of transfer problems we had in the beginning."

Another smart move on the part of the radiology department was the way it went about redesigning work flow.

"If you install PACS and simply use workstations in place of viewboxes, you don't really save many steps because you're still reliant on paper-based operations with people having to manually move orders, reports, and images from one place to another," says Siegel. "You have to change the way the department operates in order to get a dramatic decrease in work-flow steps, which is both realistic and readily achievable. The first thing you must do is document your original work flow. Once you know all the steps in the pre-PACS environment, you can review them with your IT experts to get a good handle on where you can eliminate steps. That's exactly how we approached it."

In the course of implementing PACS, Siegel and his team made some interesting discoveries. Chiefly, while it was thought that turning off film enterprise-wide would cause a great deal of consternation among the users of radiology services, the opposite held true.

"To our surprise, there were virtually no complaints about film coming to an end," he says. "What we deduced later as the reason for this was that images could be so quickly and easily obtained via PACS."

Discovered, too, was the fact that only 22% of users needed to undergo formal training in PACS, a reflection of again how simplebut also how intuitivethe equipment is to operate.

A WORK IN PROGRESS

Meanwhile, Baltimore VA Medical Center continues to evolve its PACS.

"All we've accomplished up to this point is gotten ourselves to the party," says Siegel. "The party is the paradigm shift of making the transition from film to filmless operation and being able to take advantage of the replacement of manual communications with electronic versions. Now that we're here, there's a whole new world of fascinating opportunities. Such as the ability to use computer technology to enhance images, and to do techniques like dual-energy subtraction for telesynthesis to optimize the quality of general radiographic studies. Also, to use computer-assisted diagnosis as a way to increase sensitivity and specificity.

"The first generation approach to PACS was to make digital images look like film. The second generation approach was to create stacked images where you could look through a set of CT images in a cine data set, which has been demonstrated to result in higher accuracy and faster performance. The next generation is going to involve thinking about image information not as individual, discrete frames, but as volumetric data that can be interactively navigated in any plane using a combination of multiplanar and 3D reconstruction techniques to take a 2,000-image data set and read it much more rapidly and accurately."

Also still to come at the VA is the addition of speech-recognition technology. Says Reiner, "We've not yet integrated speech recognition into day-to-day operations. We have, however, been conducting a number of research projects with it. One reason we've not been in a hurry to adopt speech recognition into our day-to-day operations is that we have a dictation system in place that gives us extraordinarily fast turnaround time on reportsless than 2 hours after completing the read. This system is built around a team of transcriptionists working from their homes but patched into the hospital via digital connections that allow them to log on and hear our voices dictating the reports and then to transcribe directly into the EMR. We're going slow with speech recognition because we're concerned the changeover will resultinitially, at leastin a decrease in radiologist productivity while they go through the learning curve."

VA hospitals, Reiner points out, operate for the most part under a capitated payment system, which has resulted in a flat technology acquisition budget for his department over the past few years. As a result, an investment like PACS emerges as something all the more valuable.

"The only way we've been able to continue providing cutting-edge technology and rapid response time has been by bringing in PACS and correctly integrating it with our day-to-day environment," he says. "It's required a lot of work, a lot of dedication, but it's been absolutely worthwhile."

NOTE: The second installment of this two-part series will appear in the February 2003 issue of Decisions in Imaging Economics.

Rich Smith is a contributing writer for Decisions in Imaging Economics.