The CIO Perspective: Issues in Image Management
It is taxing enough for radiology and IT decision makers to contend with the image-management consequences of multidetector CT, high–field-strength MRI, 3D reconstructions, and various recent other trends responsible for a rising tide of diagnostic data. Added to this are the burdens created by other disciplines across the enterprise when they use imaging systems of their own and send the output to a common archive in an environment where the goal is universal access, not just from one corner of the organization to the other, but bidirectionally from multiple points of contact. It’s not surprising, then, that the departments of radiology and IT at a growing number of institutions are working together more cohesively these days. The challenges that they face are held in common, rising from a customer base that is larger and broader than ever. The Medical University of South Carolina (MUSC) in Charleston has an exceptionally big imaging environment (with at least 89 separate buildings on a 76-acre parcel) and a growing number of disciplines across the enterprise (including cardiology, oncology, pathology, rheumatology, and orthopedics) that want to get in on imaging. To address this expanding customer base, the first concern of Frank C. Clark, PhD, MUSC’s vice president of IT and CIO, is ensuring that the enterprise has adequate archival capacity. “We have a SAN storage solution built up now to almost 500 terabytes,” he says. “Because of the explosion of imaging here, we’re adding to that storage capacity at the rate of about 15% annually.” Frank Clark, PhD, CIO Even at that, MUSC has found it necessary to tier the archive to facilitate efficient retrieval of images and help hold the line on costs. “Basically, the newest or most important images are placed on fiber channel, older or less important ones go to a SATA archive, and the oldest or least important are archived further down the performance chain,” Clark says. The catch is that image users are permitted to define the tier where they want their images to be stored. Since the natural impulse is to want image retrieval to be nearly instantaneous, most users automatically load everything onto the fiber channel. “That, of course, is unrealistic,” Clark acknowledges, “so we try to help users make better choices by providing education about which types of images should be placed where, ideally, among the tiers.” The help that Clark describes is delivered through a governance process that, among other things, encourages each clinical user to articulate his or her image needs. “We use a suite of automated tools that constantly monitors the different storage tiers to measure frequency of use and also storage growth within specific applications and segments over a protracted period of time. These not only help us identify and track the largest archive consumers, but also help us home in on what it is each user is actually trying to accomplish, so that he or she doesn’t make tier-selection choices in a vacuum,” Clark says. “We believe this is key to being able to effectuate significant savings.” Related to image-retrieval speed is bandwidth, and MUSC has been working to improve that as well. “Our radiologist workstations operate on at least a gigabyte of bandwidth; we’re attempting to provide as close to that same amount as possible out at all the desktop-based viewers used by the clinicians across the enterprise,” Clark explains, adding that it helps to have a thin-client PACS viewing application. “This is allowing us to extrapolate PACS very economically across the enterprise,” he says. Still, the shift to thin-client PACS is not without a price. “We have 10,000 or so desktops to maintain and upgrade, rather than a handful of high-end workstations confined just to the radiology reading rooms,” Clark says. “The good news is that most of our viewers are managed using a thin-client model, which allows us to push out software upgrades remotely from a central point. This saves us from having to go out and physically touch each desktop in order to implement the changes. We still must send out IT personnel, however, any time hardware is due for replacement, or if we want to expand the device’s storage, so there still are logistical issues to be dealt with here.” Images acquired from modalities throughout MUSC are ultimately funneled into a centralized clinical data repository, where they then are aggregated with each patient’s laboratory results, chart, and other data into a longitudinal electronic medical record. It is not just key images, but all images that are accessible by means of this record. “From within the repository’s clinical results viewer, one can link back to our PACS to access images fully,” Clark explains. Resource Optimization BJC Healthcare, St Louis, Missouri, has image-management concerns of its own. On any given day, no fewer than 2,000 users concurrently access BJC’s data repository to view clinical results, and they do so by means of a homegrown portal called Clinical Desktop (ClinDesk) that features an embedded viewer of nearly diagnostic quality. David A. Weiss, senior vice president and CIO, says. “ClinDesk is an electronic health record (EHR) strategy, image-viewing functionality being one important component of this system. We don’t believe you can have an effective EHR solution without a strong image-storage, -distribution, and -viewing solution component.” The ClinDesk system was conceptualized in 1993 and made partly operational in late 1996, with full activation achieved by 1998. Today, it contains information on approximately 4.5 million patients. “We foresaw the emergence of integrated health delivery networks, which led us to begin thinking about how information could be managed, distributed, and accessed across such a network,” Weiss says. ClinDesk was pieced together in cooperation with selected vendors; none of it featured off-the-shelf technology. Weiss continues, “PACS is really not a part of ClinDesk, but represents another data source, similar to a laboratory system or a nursing-documentation system. Radiology PACS, cardiology PACS, ultrasound, and so forth all have their diagnostic-quality image-interpretation systems integrated within their PACS solutions. These images are automatically fed to a long-term image-archive system, with the patient context tied to the enterprise’s master patient index and the subsequent radiology report.” He adds, “Within ClinDesk, these images are available for viewing through a separate image viewer (in order to deliver images to all physicians in a review mode, although the quality of these images is of nearly diagnostic quality). Via ClinDesk, these images are integrated with the other key clinical results for a patient to deliver a holistic view of the patient’s medical condition and to outline aspects of care planning, including clinical events intended to happen outside the traditional inpatient setting: follow-up outpatient care, home care, and so forth.” In the main, though, ClinDesk is (and will continue to be) a homegrown product. “Hopping aboard a different train while the one you’re on is rolling along the track at speed is very difficult to do,” Clark says. “Not only that, but we haven’t yet seen anything on the market that is compelling enough, in terms of greater features and functionality, to make us want to attempt such a changeover.” David Weiss, CIO As an enterprise, BJC consists of 13 community hospitals (including its anchor institutions, Barnes–Jewish Hospital and St Louis Children’s Hospital) stretching from Western Missouri to Southeastern Illinois. “Our goal is to distribute clinical information and images cost efficiently to the user, wherever he or she is, so that faster and better medical decisions can be made,” Weiss says. One way that the costs linked to the fulfillment of that mission are being held in check is the optimization of existing resources and assets. That optimization includes virtualizing as many servers as practical. “At present, nearly 40% of our servers have been virtualized,” Weiss reports. He cannot yet say, however, whether the virtualization is living up to its fullest potential. “If you ask me if we’re receiving all of the benefits virtualization promises, I can’t answer. True, after two years of virtualizing, we have fewer servers on the floor, and that does translate into reduced maintenance requirements, but it is unclear whether we’re maximizing the compute cycles of the servers that remain.” Weiss explains his concern this way: “Let’s say my unvirtualized server is operating at 10% of processor capacity and reaching peaks of 30%. Adjacent to it is another unvirtualized server running at about 7% of processor capacity and also peaking at 30%. These would appear to be ideal candidates for virtualization on a single server, so I go ahead and virtualize, only to end up with even less processor throughput and utilization. Is that plausible? Yes, if I yield to the temptation to replace one of the virtualized servers physically with something even larger,” he says. Weiss continues, “Application vendors sometimes encourage this very thing because they’re concerned that, in a virtualized world, their products will not perform as contractually promised and certified. What they fear is that their product will end up sharing a virtualized server with another product, and the result will be a bottleneck or constraint on the hardware resources their product needs to perform as contractually promised. For them, the easy way to get around that risk is to push you to overbuy on the hardware side (the capacity side).” Cloudy Outlook Another trend with the potential to lower medical IT costs is cloud computing. “The idea behind cloud computing is that you avoid capital expenditures on hardware and software by renting access to servers in a third party’s infrastructure,” Weiss says. “You access that rented infrastructure over the Internet, with a browser. You pay for access either on a per-click basis or by subscription fees.” While cloud computing does promise initial savings and a way to escape the overbuying of capacity/hardware, Weiss suspects that it might prove more costly in the long run, when the rent comes due. “Organizations with limited operating-budget flexibility may find that the cloud-computing model makes little fiscal sense if they turn out to be large consumers of the service,” he says. Weiss also expresses doubts about the viability of cloud computing for health-care organizations because of security issues. “You’re sending private, patient-sensitive medical information into a cloud that you neither own nor control,” he says. “I think there is rightly some nervousness about all of this.” Cloud computing has not yet captured much interest at MUSC. “We’re waiting to see if this is one of those faddish things that burns brightly for six months and then goes out of fashion,” Clark notes. “We’ve certainly given it thought, but haven’t done anything meaningful with it.” Caution appears to be the order of the day, no matter which enterprise image-management strategy is under consideration. That’s why Weiss recommends that radiology and IT join in lockstep to grapple with these issues. “When it comes to supporting the image-management needs of various departments that make up an enterprise, I see IT’s role being to ensure integration of all clinical-care components of a patient’s coordination of care,” he says. “This should be viewed in a broad perspective and, as such, would include all key elements of medical imaging,” Weiss adds. “Here at BJC, we have been fortunate to have a healthy working relationship between radiology and IT. The result has been that information strategies and products have been designed with the patients’ best interests uppermost in mind.” Clark has cultivated the relationship between IT and radiology by assembling a team of 50 cross-trained IT specialists, available at a moment’s notice to help users of imaging modalities and services resolve problems in the field. Typical calls for help involve matters such as images that fail to move into the work queue, radiologists’ notes that do not become attached to the images, Web access to PACS that is oddly limited, voice-recognition functions that perform poorly, and systems that become sluggish. “An advantage of having this team is that it all but ensures a higher level of interaction with the clinicians who generate or use images,” Clark explains. “This interaction equips us to understand the needs of our customers better. We want to be as attuned as possible to the business of care delivery, to the workflow, and to the processes that our customers live by; this also helps us bring more to the table with our radiology partners. Our attunement to clinicians’ needs gives us the ability to assist radiology’s decision makers in focusing on process redesign and workflow improvement, rather than just looking at technology as something they can layer on top of existing workflow and processes.” Additional Reading - HIE on the Horizon Rich Smith is a contributing writer for Radiology Business Journal.