In every sense, the distinguishing element of an integrated delivery network (IDN)-PACS deployment is size.

“That may sound overly simple, but the main difference is it’s just bigger,” says John Couris. “It’s more complex. You have to cross a lot of organizational and cultural boundaries.”

Couris is vice president for ambulatory care and medical imaging for BayCare Health System, an alliance of three major health care providers in Tampa/St Petersburg, Fla. BayCare is proceeding now with a master plan to link imaging and other medical data in a centralized network that can be accessed by any of its nine member hospitals, plus its clinics and doctors’ offices.
BayCare is not alone in undertaking an IDN-PACS deployment. Routinely, large health care enterprises are committing millions of dollars to create and install IDN-PACS systems. The costs are huge, but so are the payoffs in potential economies of scale, staff cutbacks, improved patient care, and reduction in film expenses.

With mile upon mile of fiber-optic cable to be laid, hundreds of radiology workstations to be brought online, double-digit terabytes of imaging data to be archived, and dozens of software interfaces to be created, an IDN-PACS deployment starts as a giant technical puzzle. Physically, it involves the connecting of hardware and software in a maze of electronic interdependency that allows the movement of high-resolution images and associated data. But nearly the equal of the technical piece is the human piece, says Couris. So do the other experts consulted for this story. In the effort to deploy hardware, the people involved often get overlooked.

For instance, in a big landscape of competing desires, which specialist will receive what level of viewing monitor? Which of the competing OR equipment requests from in-system hospitals will become the model or the rule for the others? How will reluctant initiates to a filmless environment be brought on board? How will everyone be trained to use the network? These are the questions that keep IDN-PACS planners awake at night.

The Technical Piece

A picture archiving and communications system (PACS) allows images digitally captured on modalities to be moved electronically to viewing stations to be interpreted by radiologists on monitors. The same images can then be routed to referring physicians or others who need them. PACS in an IDN is a less defined construct. What enables PACS to encompass multiple institutions is essentially the conduithigh-speed broadband cable, ordinary phone lines, the Internet, plus the connective and translation devices (servers and interface boxes) needed to allow the images (and a lot of other medical data) to go from one place to another for purposes of patient care or record keeping. IDNs are attempting to set up networks that will handle not only PACS but radiology reports, laboratory reports, essentially complete EMRs (electronic medical records). It is not just the number of sites but often the distances between the sites that also make IDN deployments big.

Lori Rumreich, MBA, is director of The Education and Research Institute at the Indiana University School of Medicine, Indianapolis. She is also an associate faculty member at IU’s Kelley School of Business. Her primary assignment is to the radiology department, where there are about 75 radiologists plus residents and fellows.

Rumreich and her colleagues have developed a 3-day PACS school for administrators and doctors interested in the technology. Participants receive what Rumreich calls a “PACS toolkit,” which includes a CD with a sample business plan and a sample RFP (request for proposal) to get the long installation process started.

Rumreich herself is involved in a major IDN-PACS installation that involves the IU Hospital and two others in Indianapolis, Methodist Hospital and Riley Children’s Hospital. The three hospitals joined forces in 1997 as Clarian Health Partners. The three have a total of more than 1,300 beds and admit more than 55,000 patients annually. Their outpatient centers treat nearly 920,000 patients per year.

Kim Garriott is PACS project manager for the Clarian installation, working together with Rumreich. Garriott says the Clarian PACS is centralized at Methodist Hospital. The images come from the modalities at the various sites and then are redistributed. The images and data go over a wide area network (WAN) that uses dedicated transmission lines. Data can also be accessed over the Internet through VPNs (virtual private networks) that have security features.

Rumreich says Clarian chose in 2000 to go with one of the biggest PACS vendors after putting out an RFP and studying the responses. “There are lots of small vendors, and they tended to be more flexible and their costs were lower, but the large vendor we chose had a large service network with a field service engineer based locally. Service support was one of the key drivers in our RFP.”

Both Rumreich and Garriott note that the three Clarian hospitals are very different in culture. IU is a teaching hospital, Methodist is a trauma center and a general hospital, Riley is a children’s hospital known for the ability of its doctors to do transplants. Like most IDN-PACS installations, Clarian’s was done in phases. It started in 2001 and took about 18 months to complete, Rumreich says. While the technical part of the installation went smoothly in most respects, there wereas is typical with these mammoth projectsstumbling blocks and frustrations.

Rumreich says many more nonradiologists than anticipated demanded high-end viewing stations. “We hadn’t planned radiology monitors for these clinicians, but they all wanted them.” Garriott concurs. “We went to more than 120 of our outpatient clinics and individually interviewed doctors. If we had been able to have the luxury of universal standards, we could have told them what they would get. Most of them were very reasonable, but some felt that they needed the same quality viewers as the radiologists. In the end, we let them decide.” The difference could be as much as $40,000 per unit, she says. Clarian eventually installed 168 high-end workstations systemwide, Garriott says. Web viewing applications were routed to an additional 8,000 PCs. The workstations were a significant expense, even with a PACS budget of $15 million, says Garriott.

Garriott says other site inconsistencies added costs in time and money that had not been preplanned. “We learned that we couldn’t assume that what we implemented one way at one facility would work the same way at another,” she says. She uses the example of OR installations. “At one site they wanted dedicated dual-monitor PACS in every OR suite. At another site they wanted portable workstations in the ORs.”

Another problem, says Rumreich, was integrating the PACS with the Clarian RIS (radiology information system). Both systems came from major vendors, but at the time of the Clarian installation the two systems had never been [interfaced], says Rumreich. “Issues came up, and [interface] decisions had to be made on what the end-user would get. That’s what takes time. You have to decide. But who makes the decision? That’s why you must have the project manager, and you have to identify all your stakeholders.”

Ironically, another PACS shortcoming has Clarian still printing film. The film is not looked at, but it is printed and filed away as a backup in case the PACS crashes and the archive is somehow erased or destroyed. “A redundant archive was never taken into account,” Garriott says. “It turned out the PACS vendor couldn’t send images to two different archives.” Some departments that do not rely on prior images like emergency and ICU have chosen to operate without film redundancy because the whole process of getting film into their flow is too cumbersome, but for the rest of the hospitals’ patients the film is run off and put aside, Garriott says. She says the vendor’s next PACS software upgrade should be able to print to two archives, but that software is not expected until sometime in 2004.

Another big, unanticipated expense was transferring old images from legacy PACS systems at the various sites to the new IDN-PACS system archive, notes Rumreich. She says writing interfaces to put the old images on the new IDN-PACS was successful, but it took time and money. “That was outside our initial project plan and took months,” she says. “The rule of thumb for an interface is $10,000 to $15,000.”

Despite the impediments, Rumreich and Garriott both say the IDN-PACS goal of delivering images to every inpatient and outpatient unit in the system has been or will be met. Both say staff and doctors “love” the IDN-PACS for the information it can carry and the efficiency with which the information is delivered.

Patient identification

An important aspect of any IDN-PACS deployment is a system for patient identification. Richard L. Morin, PhD, is a medical physicist and a professor at the Mayo Medical School in Jacksonville, Fla. Mayo operates St Luke’s Hospital, a 289-bed Jacksonville facility that is hard-wired for PACS with the clinic/medical school, which is nine miles distant. Mayo has its main campus in Rochester, Minn, and a third facility in Scottsdale, Ariz. Rochester and Scottsdale are connected in some respects (specialty reads largely) with the Jacksonville PACS, but the bulk of Jacksonville’s patient data is not shared with the other campuses, Morin says.

“In our design,” Morin says, “we proposed a PACS that was one component interfaced with our RIS and our EMR. That level of integration was essential to us because we did not want independent entry of any patient demographic information. An error rate of 15% to 20% with manual data entry had been demonstrated in Rochester and elsewhere.”

To achieve its integration objective, Mayo had to create interfaces for its RIS and EMR systems to communicate with the PACS. “None of these interfaces were written when we started. There was nothing off-the-shelf available,” Morin says. Mayo rolled out its EMR first. “The idea was that the [radiology] order would be placed electronically and would go to the RIS. The RIS would schedule and the DICOM header would be built and tagged with the acquisition number assigned by the RIS. That way we would always have correlation between the reports and the images. When we are interfacing with the EMR, that is in HL-7 [a written language-based code]. The RIS can be either HL-7 or DICOM [a digital information code used by the modality] depending on the information transmitted. The RIS will have an HL-7 transmission to a broker, and the broker will download that data to the acquisition device [modality]. The accession number that the RIS provides will be dropped into Group 8, Element 50 of the DICOM header, and that’s how we know that image goes with that order,” Morin says.

Rick Perry is vice president for information services and CIO at BayCare Health System in Tampa/St Petersburg, and a colleague of Couris. Perry agrees that patient identification is a key element of the IDN-PACS deployment. He ranks it on a level with image transmission itself. “For us, it has been a cleanup and merging process [from the legacy systems],” he says. “We will use a sequential number that is smart enough to tag all the other numbers that John Doe might have. If he was in Clearwater but presents himself in Tampa, the system will ask demographic queries and be able to drill down to the prior images.”


Another key piece of IDN-PACS is the communications equipment that connects the many sites in the network. BayCare, says Perry, has PACS systems in six of its nine hospitals, but they are not yet connected. To connect them, BayCare is designing a “gigabit Ethernet across Tampa Bay,” says Perry. It will be capable of speeding as much as 200 megabytes of data per second from one hospital to another, Perry adds. “We’ve got the networking equipment installed and will soon begin testing.”

But designing the communications equipment piece for the IDN-PACS is clearly different from what has to be done with a single-site PACS installation. Single sites can rely on a modest LAN (local area network) of wired connections to get images to end-users. IDN-PACS installations require longer connections and more capacity. Transmission lines or wireless setups between sites have to be laid and/or leased from cable or phone companies. This is expensive.

Bill Lazarus is assistant vice president of information technology architecture and security for St Joseph Health System, a Catholic nonprofit network that is in the process of making regional deployments of IDN-PACS in its three hub systems, one in Texas and the others in Northern and Southern California. Altogether, says Lazarus, the three regions have 15 acute care hospitals with about 3,500 licensed beds. More than a million imaging studies are completed each year.

St Joseph is installing a WAN that consists of a “private fiber network” between its Southern California hospitals. The data moving along this “ring,” as Lazarus calls it, will be stored centrally at the home office in Orange, Calif. Hospitals in the Northern California hub will also be on a ring that will feed data to Orange. The Texas hub will have limited connection to Orange, but the Texas data will not be stored there. Texas, which accounts for about 35% of St Joseph’s volume, is too big to route into California. “They will have the same system architecture, the same type of central storage, the same design and workstation standards. It just wouldn’t have been a good decision to pull their volume and traffic through Orange,” Lazarus says. “The risk outweighed the benefit.”

Lazarus says the cost of the initial Orange WAN is about $300,000 per year. He says standardization at St Joseph’s three hub sites for IDN-PACS will fit in with a “larger initiative called Care Redesign” that will provide tools “to clinicians to automate a lot of what they do and improve accuracy by reducing errors.” New pieces, he says, will include computerized physician order-entry and clinical documentation. “We are modernizing our entire hospital IT [information technology] infrastructure,” he says, suggesting doctors might one day be making rounds with small handheld devices on which they could display EMRs and images.

The WAN must be planned carefully, just as carefully as the other elements of the IDN-PACS. Deanna Welch, MBA, RT, is director of imaging services for Intermountain Health Care (IHC), a 21-hospital network based in Salt Lake City. Welch says IHC began looking at PACS a decade ago and even then believed that having individual PACS systems at each of its hospitals “was never going to fly.” IHC chose a single vendor in 1995 and now has all but a few of its rural hospitals linked in an IDN-PACS. “We think we are the largest PACS install in the country,” Welch says. “We do more than 700,000 radiology procedures per year.”

Welch says different regions in the IHC network are linked on a regional WAN that feeds into a central WAN where data for the system are archived in Salt Lake City. Financials and miscellaneous communications also are transmitted over the WAN in company with images, she says. She notes that in the beginning IHC did not know how much WAN it would need. “You don’t want to overbuild. We had to buy some tools to monitor the traffic. We knew we’d get in trouble if we rolled out the PACS and it took up the WAN, so we developed strategies to make sure that didn’t happen.”

The People Factor

Once the hardware and the rest of the technical pieces are in place, the training can start. But the people piece should be initiated long before that. BayCare’s John Couris says that it is important to develop support teams from various hospitals and their departments before the IDN-PACS is ever deployed. “We decided for us [at the first site] phase one would be the ED, so we got the ED director, the ED nurse, and the radiologists sitting at the same table with the ED doctors. We got them involved in the decision-making process. They weren’t just filling out sheets of paper, they were at the table. That’s key. People think it’s obvious, but it’s not. In the end you have to be fast and you have to be really good, or you’re at a competitive disadvantage. You have to focus on cost and on service and how to deliver it quickly.”

Deanna Welch says one half of the people component involves training to the PACS, and the other half involves “redefining workflow in an electronic environment.” At IHC, Welch and her team employed what appears to be a common strategy for getting clinicians and staff comfortable with an IDN-PACS. Each department involved with imaging would have one person assigned as a “champion.” That champion would lead the others through training. “You also have to have dedicated PACS administrative people who are the driving force,” says Welch. “We had to sell the system in the beginning, but now it has kind of sold itself.”

To train users, there must also be trainers. At Clarian Health, Rumreich says, “We underestimated our own manpower needs. How many people it was going to take to train people, go to meetings, and interface with the vendor. We originally had a team of four people and we ended up with more like 10. We went with a model of developing superusers’ at each of our sites, and they transitioned into day-to-day support personnel.”

Rumreich and her colleague Kim Garriott also developed a PACS project team made up of radiologists, clinicians, and administrators. Rumreich calls them “stakeholders,” and they make key decisions. “PACS is not just a radiology system,” says Rumreich, “so if you’re politically savvy you’ll include all of the users. We created open houses’ to bring all the clinicians to radiology to show them the advantages.” Like others, Clarian also designated site champions.

But with different hospital cultures, Clarian found that even training radiologists to the PACS involved concessions. “One site said, All our radiologists will come to one room and you can train them.’ Another site said, Our radiologists are very specialized and very busy so you have to train them one-on-one,'” Rumreich says. “So we did one-room training at one site and one-on-one at the other.”

BayCare’s CIO Rick Perry says it is important in the installation process to “identify senior management who will sponsor and lead the project and hold them accountable. We do that We do that through a project justification model using service, outcome, and cost indicators. Ultimately, the senior management project sponsor is accountable for the success of the project based on the performance improvement observed through the data collected on the indicators.”

With big IDN-PACS systems the deployment and training never stop. New equipment is constantly being added somewhere in the system. Mayo’s Richard Morin calls it “the land of perpetual upgrades.”

“You get to the point where there’s obsolescence. Those are the biggest decisions we have noware we going to do the upgrades? It’s a very big deal,” he says. “You’re talking millions.”

George Wiley is a contributing writer for Decisions in Axis Imaging News.