d02a.jpg (9155 bytes)Is your facility pining for a PACS?
Are your administrators relying on a PACS to increase productivity?
Are your financial folks counting on a PACS to save money?

Words of advice: GO WITH THE FLOW… the workflow, that is.

Anyone who has installed a PACS (picture archiving and communications system), is installing a PACS or contemplates implementing a PACS quickly learns that the process — from planning to placement — takes months, even years. You can never ask too many questions. You can never sift through too much information. You can never explore too many options.

When all is planned that can be planned, however, when all is discussed that can be discussed and when all is evaluated that can be evaluated, one factor stands out as the mark of a successful PACS implementation: the ability of the facility to appreciate the importance of workflow.

“In some environments, the object of the game, to install a PAC system, is essentially to achieve a filmless operation, as though that were in and of itself a goal. But I think the goal really ought to be to improve the workflow in the department,” says Eliot Siegel, M.D., chief of radiology, VA Maryland Healthcare System (Baltimore) and associate professor, University of Maryland School of Medicine (Baltimore). “In hospitals that appreciate workflow, what happens is the PACS just becomes a tool, one of the many tools to try and automate the whole process.”

Siegel, who has been in charge of Maryland’s VA hospital system since 1987 and on faculty at the university since 1986, notes that the VA holds the distinction of being the first hospital to discard film when it did so in 1993. All images are read via soft-copy, except for mammograms, which are read on film and digitized to PACS.

The hospital PACS, installed at the main VA in Baltimore, also serves the Perry Point VA, the Fort Howard VA and the Baltimore Rehabilitation and Extended Care Facility. The facility generates approximately 90,000 studies annually, many of which consist of multiple images.

When an institution does not use its PACS to automate the work process, it ends up with an electronic analog of its film-based operation, Siegel explains. Technologists still type into their CT scanners, paper requisitions continue to come into the radiology department and radiologists still type in information to call up a patient in the system.

In an institution that values the workflow concept, however, all steps are automated: An order from a referring physician flows directly into the radiology department in the PACS; radiologists find all the information they need rather than having to type it in themselves; physicians can read from a worklist of studies waiting to be read; and the technologists can scan patients from a list of patients whose studies have been ordered.

It’s the tuner, not the speaker
Katherine Andriole, Ph.D., suggests that workflow gets the short shrift because it lacks what Madison Avenue would call the “sex appeal” inherent in hardware components, particularly workstations.

“Some of the mistakes people make are that they look at the end of the system, that being the display station, and they look at the fancy things that the display station can do,” she begins. “What they forget is that the images have to be delivered to the workstation or maintained by a system that is somewhat transparent to the end-user. How images get acquired from the imaging modalities, how they are archived, how they can be retrieved, getting relevant prior examinations to the workstation, how images are hung at the workstation — what we call hanging protocols — those kinds of things now have to be done automatically by the computer. There are definitely systems out there that have fine and adequate display stations, the user interface and all of the different tools that everybody has, but they don’t have the infrastructure or what we call the folder manager, the workflow manager piece, and that — to me — is the critical piece.

“It is like having a fancy speaker on the end of a stereo system,” she adds. “If you have a real expensive speaker but you have a terrible tuner, it is not going to sound good.”

Andriole, professor in radiology and bioengineering at University of California San Francisco (UCSF), also is that site’s PACS clinical coordinator. When she came on board the university in 1992, it was designing and building its own PACS, which it did exclusively until seeking a vendor partner in 1997. Its current PACS, predominantly an Agfa system (Agfa-Gevaert Group, Mortael, Belgium), handles approximately 250,000 studies a year for the university hospital; Mt. Zion, a San Francisco community hospital; San Francisco General, a county hospital; and an affiliated VA hospital.

Don’t risk it without RIS
Sites that are mindful of workflow also understand the important role a RIS (radiology information system) or HIS (hospital information system) plays in achieving a smooth working environment, Andriole says.

“One basic thing I always tell people is, if you are considering getting into a PACS, and you do not have a RIS or a HIS — don’t do it! Do the RIS first because a lot of the functionality that one gets from a PACS is the information that a PACS gets from a RIS. If you do not have [a RIS], you are not going to see all the functionality that you should get in a PACS, and you will probably end up being disappointed.”

Andriole offers hope for those lacking a RIS, however.

Such a network can be added after-the-fact, but she advises that hospitals avoid proprietary interfaces and instead choose a RIS or a HIS that incorporates DICOM, HL7 and TCPIP network capabilities to ensure that their separate systems interface seamlessly. PACS vendors can assist in that selection by demonstrating the kinds of interfacing they have done and with whose products, she offers.

David Avrin, M.D., Ph.D., Andriole’s colleague at UCSF, wholeheartedly agrees with her PACS-RIS assessment.

Avrin, professor of clinical radiology and vice chairman of clinical services, trained as an interventional radiologist and CT imager. His doctorate is in electrical engineering and computer science.

“Poor PACS-RIS integration leads to poor PACS,” he declares. “The problem is, in general, the PACS knows of studies that are digital and the RIS does not know how a study was produced. It gets the report for the study, but it does not necessarily know if it was produced on conventional film or PACS. So each environment is blind to the other.

“I would say, for good PACS implementation versus bad, a side note of that would be good RIS-HIS integration with, most important, a good master patient index and automated order entry.”

Like Andriole, Avrin insists that core components — workflow concepts, PACS database, archiving system, for example — meet DICOM and HTL standards. But Avrin also recognizes DICOM’s limitations in that the current standard does not define a database interface. As a result, vendors implement proprietary client-server applications to tell the system who the patients are, what studies they have had and so on.

And therein lies the problem: It becomes extremely difficult to pair one vendor’s workstation with another’s PACS and have it work appropriately because even though it has the DICOM connectivity to receive images, it does not have the client-server capability of interrogating the system for patients and studies.

“DICOM has been incredibly successful on the modality sides,” Avrin states. “You can put any vendor’s equipment on any DICOM-compliant PACS today and get images into it. That is virtually a true statement. So on the input side, DICOM really solves the problem.

“The problem is on the workstation side with the workflow issues,” he points out. “There is a missing true database level in DICOM that has been left basically to the individual vendors. There are exceptions,” he allows, preferring not to mention vendor names. The question is: Is it a technical or a business reason? The vendor’s will tell you it is a technical reason, but it becomes a business reason.”

Avrin’s concerns about connectivity standards have fueled his involvement in the Integrating the Healthcare Enterprise (IHE) initiative, the joint project of the Radiological Society of North America (RSNA of Chicago) and Health and Information Management Systems Society (HIMSS of Chicago) which has been demonstrated at the annual meetings of both organizations.

The same can be said of Siegel.

“The IHE effort is essentially a recognition of the fact that, although we have standards such as DICOM and HL7, there is enough ambiguity or flexibility — depending on whether you see the negative or positive in those standards — that makes it so you cannot easily plug-and-play hospital information systems with PACS, with radiology imaging modalities,” he comments. “IHE is a real exciting effort, at least from my workflow perspective, to allow places like community hospitals that buy PACS that do not have the expertise, or do not have the personnel or the funding to reinvent custom interfaces between their equipment and their information systems, to say ‘When I buy equipment, I want it to be IHE-compliant; when I buy a RIS [I want it] IHE-compliant.’ Once that word gets out, even the smaller hospitals will not have to pay inordinate sums of money to vendors and wait significant periods of time that may delay their implementation and their being able to achieve the workflow that some of the bigger places have.”

Making sure it is secure
But with the capability for flowing information and images back and forth — especially with a successful PACS implementation extending throughout the hospital enterprise into emergency departments, operating rooms and critical-care units and beyond, into group practices off-site or private physicians’ offices in the community — comes a complementing concern for security.

“There is a security issue going over the Internet that I have a problem with,” voices Janice Honeyman, Ph.D. “I just do not think anybody’s solved that yet.”

As director of informatics for radiology and associate professor in radiology at the University of Florida (Gainesville) for the last 11 years, Honeyman administers an enterprise-wide PACS that takes in a main hospital, Shands at University of Florida; other, smaller Shands in outlying areas; a sports medicine clinic; separate imaging centers; and an ambulatory care center. Those facilities rack up 200,000 studies a year.

Andriole counters that UCSF has set up a system of protective firewalls — passwords and data compression, for example.

And Siegel, who is co-editor of the book, Security Issues in the Digital Medical Enterprise, indicates that biometrics — fingerprint recognition — is proving to be reliable and inexpensive at approximately $100 per workstation.

Another up-and-coming technology comes in the form of a radio frequency transmitter that is worn on a lapel or carried in a pocket or purse. When the wearer comes within a certain proximity of the workstation, it recognizes the user and logs him or her on automatically; when the user leaves, the transmitter logs him or her off.

But, according to Siegel, devices address only the technical aspects of security. Another, possibly more pervasive, threat to security is the medical culture itself, he says.

“What we found in a study that we did looking at PACS implementation at the time, in 1997, was that, in general, PACS sites were very lax in security. It was very common to have Post-It notes up on the workstations, that users would not sign off or log off when they were finished and so the next user would come on and essentially use the previous person’s sign on,” he relates. “There was a lot of sharing of sign-ons and access codes, and the bottom line was that the culture within the medical centers was not one to promote security; it was one to maximize ease of access. We are going to have to find a really good compromise between easy access to images which everybody wants and being able to protect a patient’s confidentiality and the security of the data.”

Size does matter
Aah … the data. Because when all is said and done, successful PACS implementation involves images, viewed as data, and information, also considered data.

Honeyman, a PACS consultant, says she has seen the good, the bad and the ugly of PACS implementations. Mistakes, miscues and missteps range from the minor — expectations of bandwidth gone sour, for example — to the major — undersizing the system and networks, and underestimating archiving needs.

While some users may not have been sophisticated enough to know what they were or were not buying, some vendors employ engineers whose software solutions cannot work in the “real world” of radiology because those engineers have never bothered to find out how radiology uses a PACS, she asserts.

“I have known some sites where the workstations that the radiologists are supposed to be reading images from really never did what they were supposed to do, so the radiologists refuse to use them,” she says. “Or you get an archive that is not big enough, so the next thing you know you have got to have some sort of shelf management where you take tape or disk out of the archive, put it on the shelf, and you have to call somebody if you want to get something off of that. That takes too long and people get frustrated. And things get lost that way. So I would say undersizing it, undersizing the networks, not having a clear understanding of how the information flows. That’s the kind of mistakes I mainly see.”

“I think that people need to plan for what they think they need, then plan for more than what they think they need,” echoes Andriole. “I think people do not anticipate image size; the size of imaging studies is growing substantially. We have some new technologies now, like spiral CT, where there are many more images than there used to be. Also, putting more networks in because things like the cost of laying networks initially is the main cost, hooking up additional lines is a lesser cost, but if you have to go back and lay more networking, that is a huge cost. So lay lines you do not think you need now, because you probably will need them in the future.”

And another thing… and another…
So workflow… security… system and network size are among the basics of a successful PACS implementation. But the list of considerations is seemingly endless. Here’s a few more:

• Forge a multidisciplinary planning team. Include cardiologists, orthopedic surgeons, emergency department personnel, critical-care personnel and other physicians; staff members in computer science or informatics, information services or information technology, maintenance, the business office. The more points of view, the better.

• Visit vendor installations. “The best test, if you are looking at a number of vendors for a PACS, is to go to a clinically functioning site of that vendor,” advises Andriole. “If the vendor cannot demonstrate to you a clinically functioning site with their equipment and they would rather take you to their laboratory in some warehouse, I would be a little reluctant to go with that system because things work very well in the laboratory. Then you get them in the clinical arena and there are problems you cannot anticipate in a laboratory. Speak with the technologists and the business manager [at a site] and ask them about that system and all the things that go with that system. When dealing with new vendors that have yet to establish a clinical site, you need to say things like, ‘Can you let me try this system on site for a while? Can I have my clinicians sit down and drive this system? How is the workflow handled?’ “

• Buy COTS (components off the shelf). You find COTS in stores and computer catalogs. “In this day and age when you have 100 megabit Ethernet available for $14.95 per network card and 1 gigabite Ethernet coming available at not very expensive cost, you are crazy to use any proprietary network or storage technology,” says Avrin.

• Go in stages. It gives physicians a chance to get used to working with workstations and helps diffuse resistance. Some hospitals start with CT and MR; others start with the emergency department. It also may blunt the financial impact on your institution.

• Evaluate the options. Accessories like voice recognition and 3D imaging are nice, but not necessary from the get-go.

Easier said than done
“Over the last 10 years, there have been more bad [PACS implementations] than good ones,” Honeyman sums up. “Well, we were learning. It all looked easy on paper when we started out.” end.gif (810 bytes)