Few processes in radiology are as dreaded as switching from a legacy PACS to a newer replacement. Often approached only when the former PACS is on its last legs, the transition between systems requires diligent selection from an ever-widening field of solutions, lengthy migration of complex (and sometimes flawed) data, and retraining of all affected staff. It’s also, however, the next step that many hospitals and practices have been awaiting to bring 21st-century clinical and business capabilities to their imaging operations. As Tony Linkmeyer, director of medical imaging at Wilson Memorial Hospital, Sidney, Ohio, recalls, “Base systems now are three generations ahead of where we were,” with a PACS implemented in 1998. He adds, “We thought we were experts. We never even thought to ask whether the PACS we were considering could do certain things, and we found out about so many features that weren’t even on our radar.” PACS technology has come a long way since its youth; as a result, the selection process is more complex than ever before, often requiring buy-in from several parties: radiologists, technologists, IT staff, administrative staff, and managers. In the case of Nebraska Medical Center in Omaha, Michael Battreall, director of medical information systems, brought in a wide array of end users. “We wanted folks from each of the different disciplines within the radiology department and the hospital,” he says. “We had people from outside of radiology, like the emergency department, taking a look at the pieces that would be affecting them.” Alberto Goldszal, PhD, MBA, CIO of University Radiology Group PC, East Brunswick, New Jersey, recommends a different approach: While end-user reactions, desires, and aspirations must have considerable weight, he says, it’s best to let those who know their way around a PACS bear the brunt of the decision making. For University Radiology Group’s PACS selection team, Goldszal assembled two radiologist champions (typically at the CMIO level), the practice’s RIS administrator, and its PACS administrator, as well as representatives from the IT departments of the hospitals that the practice serves. “Power users and administrators are in the best position to make those decisions,” he says. “It takes time to accrue this kind of knowledge. In reality, these things are best learned by doing.” Strategic Decisions During the more than eight years that Wilson Memorial Hospital was using its legacy PACS, the hospital saw its growth almost double—and noted another important change. “When we looked at where our patients were coming from geographically (by zip code), we saw that our radius approached 100 miles—we had patients coming from Columbus, Dayton, and Lima, Ohio, and from Fort Wayne, Indiana,” Linkmeyer says. “We knew we had to distribute our images better if we wanted to grow and attract business from specialists.” Linkmeyer and his team started with 40 vendors, but they quickly were able to narrow that number down to six key contenders. From there, Linkmeyer eliminated potential vendors based on capabilities that Wilson Memorial Hospital knew it wanted (such as ease of integration and experience with various modalities), as well as on capabilities that the hospital hadn’t realized were available. “I was surprised how much everything had changed,” Linkmeyer recalls. “As we went through the process, it took several meetings with each of the vendors because we didn’t ask the right questions, the first time, about all of the features. We started learning about 3D, storage methodologies, cardiac cine, nuclear-medicine cine, and all this stuff we took for granted didn’t exist. It was an eye opener.” In the end, Wilson Memorial Hospital selected a PACS that enabled it to perform the kind of referring-physician outreach that the hospital needed without infringing on the IT staff’s protectiveness toward its network. “Our CIO was focused on letting people into the network without allowing them to touch it,” Linkmeyer says. “The image distribution of the PACS we selected has not disappointed.” For Battreall and his team, the decision-making process was more deeply affected by prior experience. “There were a number of features and capabilities we didn’t have with our current PACS that we knew we needed,” he says. “With the previous PACS, there were different databases for the radiologists and Web-based users, and trying to keep those in sync was difficult, so we knew we wanted one database. We had specific requirements regarding speed and how rapidly the exams opened. We were lacking tools for our orthopedic surgeons for their 3D templating, and we needed tighter integration with our 3D vendor so our radiologists would have a workspace where they didn’t have to get up and move from workstation to workstation.” Battreall was seeking a replacement for legacy PACS that had been in place at Nebraska Medical Center for seven years; he began with a short list of six vendors, and he established criteria with which to evaluate their products for an apples-to-apples comparison. “Capabilities we needed to have and things we didn’t do right the first time (or were lacking) went right to the top of the list,” he notes. “We tried to take a very systematic approach to the change.” Goldszal and University Radiology Group had been using a legacy PACS for 12 years when they began the transition to a full-blown version of the thin-client PACS they had been using for teleradiology. “[The legacy PACS] might have been the oldest PACS from that vendor on the East Coast, which is a testament to the bulletproof technology. Of course, on today’s level, it lacks some of the features and functionalities we see in applications developed later on,” he says. University Radiology Group’s team was seeking a solution that not only would improve quality of care, but would also boost the practice’s revenue by supporting its burgeoning teleradiology business. “It was a matter of strategic partnership,” Goldszal says. “We needed a PACS that could do equally well with the distribution of data among our clinics and the incoming data from our hospitals while providing full support for a worldwide teleradiology operation.” Storage was also a concern for the practice, and Goldszal attributes the weight that it held in his decision making to the fact that University Radiology Group had been through the PACS-replacement process more than once. “When you’re getting your second or third PACS, you start thinking about long-term needs, and you start creating economies of scale in your mind,” he says. Optimizing Migration When Linkmeyer’s team approached its legacy vendor about migrating data from the former PACS to the new platform, the response was “sour grapes: good luck,” he says. Linkmeyer then asked prospective PACS vendors for their recommendations, and almost unanimously, they suggested working with a third-party company that specializes in data migration. In the end, vendors’ responses to the data-migration question became a selection factor as well. “We knew we had to be prepared for what would happen at the end of the next eight years,” Linkmeyer says. “We asked every vendor, ‘If the relationship between us sours, how do I get my information off your system and onto something else?’” Linkmeyer found himself in the same position as many organizations that are trying to replace a turn-of-the-millennium PACS a decade or so later. “We had some really ugly old stuff,” he recalls. “We had tapes on shelves. Even the third-party company had a heck of a time doing it well.” Around six months after selecting the replacement PACS, in May 2007, Wilson Memorial Hospital commenced the migration process, two months prior to going live with the new system. “We only migrated five years’ data and left three behind, so it came to around 200,000 exams,” Linkmeyer says. “It took a few weeks after implementation to have it all migrated. We allowed them to work backward, so we knew some data would come trickling in late.” Goldszal is an advocate of this phased approach (see figure). “When it comes to data migration, people want to migrate everything and then turn on the new system, and it just can’t happen that way,” he says. “You have to maintain day-to-day operations with the old system while planning and phasing in the introduction of the new one.” He notes that it’s critical not to overburden the legacy PACS (a warning echoed by Battreall, whose team moved 1.1 million exams from its legacy PACS to the new system without involving the legacy vendor). “We were probably 85% of the way there when we turned on the new PACS,” Battreall says. “Our primary concern was that we didn’t want the old system to get beaten up to the point where it couldn’t get anything done.” Battreall’s group migrated all 70 terabytes of data in six months; Battreall attributes this rapid pace to the attention that Nebraska Medical Center’s staff had always paid to the data being input into the PACS. “The reason people run into issues is that they don’t take care of their data,” he says. “We made sure we had orders for all of our exams, and we took a very systematic approach to making sure what we stored was accurate. For some organizations, that can be an afterthought.” At University Radiology Group, Goldszal and colleagues adopted an algorithm that identifies images and reports that have matched accession numbers, migrating those data—which Goldszal calls the low-hanging fruit—first. “Then, you go after the exceptions and transfer those as well, and you keep doing that, filtering with algorithms, until you get to a point where you need a human operator to determine where the data belong or discard them altogether,” he says. “The good news is that the most recent data will be the cleanest, so you migrate those first, populating the archive at the fastest possible rate. Once you have two years’ data on the new archive, you can turn on the new PACS and have an on-demand process to prioritize other data you need. From the end-user perspective, the radiologist can always see the entire history of the patient.” Training and Implementation Just before implementation of the new PACS, Linkmeyer, his PACS administrator, and a member of the hospital’s IT staff traveled to the vendor’s headquarters for training, and upon their return to Ohio, the three proceeded to train radiologists and staff on a test server. “We considered ourselves administrative superusers,” he says. “The team leaders of each area were below our level.” Linkmeyer and the team started all users at an introductory level to make them more comfortable with the new system and then trained them in more advanced capabilities over time. “My PACS administrator babysat that thing day in and day out until everyone was comfortable with it,” he recalls. “Then, as people became more comfortable, we gave them more freedom and control.” Battreall’s group adopted a staggered approach to training, taking Nebraska Medical Center’s technologists live on the new system one week before the radiologists. “We didn’t want to throw the technologists under the bus,” he explains. “We wanted them to be acclimated before we brought on the radiologists.” Technologist superusers trained other technologists, while radiologists were given one-on-one training. Users outside the department were trained last, en masse. “You don’t want to start training users too early,” Battreall advises, “or they’ll forget what they’ve learned. We started with the radiologists one week prior to going live.” Goldszal’s team had an even more extended training process, complicated—or simplified, depending on your perspective—by the fact that University Radiology Group had already implemented a teleradiology-specific PACS, from its new PACS vendor, two years before replacing its legacy PACS. Goldszal says, “The teleradiology PACS works with a universal viewer, so when we went live with it, two years ago, a lot of our radiologists decided to use that viewer to look at the legacy PACS data. People started using it more and more, and we kept growing that.” For the obstinate few who stuck with the legacy PACS viewer, on-the-job training on the new system was given. “I’m a big believer in continued training,” Goldszal notes, “so our organization’s training across all applications is a lot more intensive than at a lot of other places. We keep constantly reinvesting in training sessions and knowledge-transfer events so that our staff remains really knowledgeable.” Wilson Memorial Hospital’s new PACS allows the administrator to grant graduated access to different users: advanced diagnostic viewing to radiologists, emergency-department physicians, and some other clinicians, and a more limited toolset for on-site clinicians and referring physicians. “There’s very little difference between the two except for the tools,” Linkmeyer notes. “When we get a new physician group, we simply ask its members if they want access and issue them usernames and passwords if they do.” Goldszal’s team took a similar divide-and-conquer approach. “The radiologist population is very clearly defined in its needs,” he says. “Don’t treat everyone else as the same type of user.” Compared with radiologists, he adds, “Technologists doing work prior to image acquisition may need to see similar things, but they’re doing different functions. Then there are the referring physicians, who may not want to see the image at all, and more image-intensive specialists, who are likely to look at the image and may even do some consultation with their radiologist colleagues. It’s easier to manage everyone’s expectations if you know what users want and tailor their accounts to give them just that.” Nebraska Medical Center’s new PACS offers customizable viewing to all referring physicians, although the team elected not to overwhelm them with too many options initially. “It’s very user-friendly, and you can change the look and feel to get it the way you want it,” Battreall says. “We kind of scaled down the customizability and gave them a canned template out of the gate. Most of the users aren’t that high level, and this way, they have a good start with the tools and can modify it from there, if they choose.” Completing the Transition All three organizations are adjusting to the new capabilities their replacement PACS have afforded them—and the new responsibilities that come with them. For Goldszal and his colleagues, the primary lesson has been to keep data clean for the next transition cycle. “It’s all dependent on really good behavior in data acquisition,” he says. “If you bypass those important milestones, you lose that ability to access the report down the line, so you phase it in and it pays off—your life becomes easier as you move forward.” Linkmeyer’s group also gained valuable experience when it came to the transition between two platforms. “Even though we thought we were ready for it, I just don’t think we were,” he says. “We use PACS without a RIS, and that presented a lot of challenges when it came to functionality and ease of use. Our new PACS vendor’s people would say, ‘Just have your RIS do this,’ and I’d have to stop them right there. In hindsight, I really wish we’d had a RIS with a PACS. Two years down the road, I don’t need a RIS anymore, but our IT department and vendor really put their heads together to make this thing work right with just our hospital information system.” Battreall adds that it’s important to know where you started so you can assess how far you’ve come. In the case of Nebraska Medical Center, implementing the new PACS led to a 25% reduction in turnaround time. “We wanted download speed, online access to images, and a reduction in reading time, so we constantly monitored those to make sure we got the results we were looking for,” he says. “When you make a change of this magnitude, you have to make sure you have measures in place before you make the change so you can see if you made any improvement after the fact.” Cat Vasko is associate editor of Radiology Business Journal.