Strategic Enterprise Imaging: An Impossible Dream Takes Shape

Vendor neutral archives (VNA) represent the core foundation of enterprise imaging, according to imaging informatics experts at three leading academic medical centers. Development of a comprehensive implementation strategy represents its brain. No matter how excellent are both elements, implementation must proceed in a step-by-step process. The complexities of people, politics, equipment, and healthcare processes must be anticipated and addressed individually, as necessary.

Radiology Business Journal interviewed representatives from the radiology departments of three large, complex healthcare organizations for insight into how they are addressing the challenges of enterprise image management. Christopher Roth, MD, CIIP, director of imaging informatics Strategy, and vice chair of radiology, Duke University Health System discussed the Duke experience; Jon Baird, Denver Health’s medical imaging manager; and Steven Horii, MD, professor of radiology and member of the department’s IT committee discussed the experience at Penn.

The issue of diversity—the diverse needs of different medical specialties—

challenges the concept of an integrated PACS for acquisition, management, distribution, and storage of DICOM and non-DICOM images. Roth knows this well. His organization implemented an enterprise-wide electronic health record (EHR) in June 2013 and purchased a VNA in November 2012. More than 100 clinical applications needed to be integrated in a single platform for use by the medical professionals and staff at its three hospitals in Durham, N.C., and Duke’s many outpatient clinics, physician offices, and other treatment centers.

The ongoing migration

Images from the enterprise radiology PACS were migrated into the VNA over the course of 15 months, although there were stops and starts during this time, for system upgrades and quality checking. From the time of this implementation, the objective has been to incorporate every clinically relevant image into the EHR, from all the specialties at every entity of Duke.

This process is ongoing today and is expected to continue for several more years, due to mergers and investments in nearby practices—and as technical and contractual limitations with existing archives are discovered. “You should plan for at least a terabyte migrated per day, on top of normal hardware activity,” Roth advises.  

Expect delays along the way, Roth advises. “In our case, a migration of images out of one large cardiology PACS right now has stalled because our vendor cannot systematically and reliably move the images,” he says. “When you get to smaller, potentially less well supported archives in smaller subspecialties, you never know what you are going to find.”

Pragmatism was a key part of the Duke strategy. Roth and his colleagues initially focused on identifying the most important imaging problems that needed to be addressed from an enterprise imaging perspective. They sought to identify locations that had standalone PACS not integrated into the EHR, as well as specific workflows to consolidate. They also sought to identify problems associated with inconsistent or absent image storage leading to patient safety risks, compliance concerns, and incomplete charge capture.

“We tried to identify high yield areas that would provide the most bang for the buck,” he says. “Specifically, we wanted to identify how we could solve the image storage issues of many physical locations performing similar imaging procedures using a single consolidated workflow build that many sites can use.”

After early pilots, Roth and team prioritized clinically high yield images at high volume imaging sites, with a focus on where charge capture can be improved the most. “When it is possible to show potential significant revenue gain, this adds visibility to the value of enterprise imaging for the entire organization and can make additional resources easier to justify,” he explains. “To date, an enterprise imaging approach has already won us revenue gains through better charge capture. It doesn’t at all cover the cost of the hardware, people, and time, but the net effect is positive because a patient’s record is more complete.”

User endorsement—specifically physicians—is the most critical factor of implementing an enterprise imaging strategy, Roth asserts. When he approaches a medical specialty to begin an image integration initiative, identifying a physician within the specialty as the key contact is mandatory.

“You have to have an engaged person on the other end,” he says. “If there isn’t a physician ‘cheerleader’ and if the other practice physicians aren’t engaged, why invest resources on a project whose end result may not be used? I also require their department chair to sign off that all groups within that specialty performing the exam will move to [the new] workflow. All the departments we have completed or are working on today are sites that came to us because they knew we could solve a problem for them; we haven’t had to go find them.”

Ultrasound: An example

Adding ultrasound imaging into the enterprise strategy proved to be a high priority. Different physical locations throughout Duke Health System perform ultrasound of various body parts and bill for them—or do not. Of the groups that engaged the image management team for integration, they were prioritized by volume of exams, potential charge-capture wins, and clinical importance of the captured images.

Initially, the physical locations where ultrasound exams were being performed needed to be identified, as did the people who were performing the exams in inpatient and outpatient environments. “Endocrinology approached us for help with storing thyroid ultrasounds,” Roth begins. “After we dug into thyroid imaging, we learned it wasn’t just endocrinology we would need to accommodate, but endocrine surgery, otolaryngology, radiation oncology, and pathology as well, across all of our ambulatory and inpatient settings. The workflows and storage mechanisms varied. We didn’t want to build something for endocrinology that wouldn’t scale to other groups; so we shook the trees, got all the right people in the room from many specialties, and we are working through a single integration for thyroid ultrasound now.”

For many of the stakeholders, permanent archiving of thyroid ultrasound images was highly inefficient: Images were printed and sent to health information management to be scanned into the patient record; burned onto DICOM CDs; or stored on thumb drives. Reports were being integrated with the EHR, but images were not.

Prior to inviting stakeholders to a meeting, Roth’s team asked them to complete an intake form that asked how many thyroid ultrasound exams were performed in a year, where they were performed, and the number and type of clinical staff who performed them, for instance. “We asked questions about ultrasound devices, specifically the make and model of the device, if it has DICOM modality worklist capabilities, storage capabilities, if it has network capabilities, et cetera,” Roth adds.

At the meeting itself, Roth’s team told those gathered that it had a system to provide the capability for order-based storage. “We explained that when an order was placed in the EHR, the order would automatically flow to the modality to be used, whether an ultrasound machine, a scope camera, or fluoroscopy,” he explains. “They wouldn’t have to type in name, medical record number, date, other salient metadata. The person performing the exam would choose the order at the modality. When images were acquired, they could be associated with the order—just as the radiology department does—and send these images to the patient’s medical records. People were very receptive to this workflow.”

The ultrasound initiative began with perioperative anesthesiology, a group that was enthusiastic, and had a small footprint and low volumes, ideal for a pilot study. Other groups followed.

Beyond enthusiasm

Enthusiasm about participating in enterprise imaging is not enough. Roth explains that the urology department was very interested in capturing both the prostate ultrasounds and images from their scope cameras. Moving images from scope cameras into the VNA, however, would have required the purchase of new middleware.

“The scopes and associated hardware were bought with what was necessary to do scope endoscopy at the time, years ago,” Roth shares. “No one thought about hooking them up to an EHR and doing systematic image storage back then.” Based on a cost analysis, the team decided to work toward the urology ultrasound storage only for now and focus scope camera endeavors on sites that were already technologically prepared.

One of the most complex questions to answer in an enterprise imaging initiative is how to handle visible light photography. There are many sources of visible light images, smartphone, scope camera, and high-definition dermatology camera among them. It is possible to take any of these image types, turn them into DICOM, or archive them as non-DICOM. HIMSS and SIIM both have workgroups currently tackling the question of optimal visible-light image workflow. The workgroups are identifying the gaps, so that technology can be developed to address and resolve them. 

To date, Duke University Health System has migrated all images from all hospitals performing radiology. It also is in the process of integrating all visible light and ultrasound images from cardiology, anesthesia, orthopedics, thyroid ultrasound, urology, and emergency medicine.

Duke is composing a strategy for enterprise imaging that will lay the framework for the future migrations. The job is never done, because any time that the health system absorbs a practice, it is necessary to standardize their terminology and workflows, as well as absorb and integrate their images.

The process is complex and laborious. In spite of an EHR, a VNA, an enthusiastic IT team, engaged clinical staff, and a well thought-out, realistic strategic plan, implementing enterprise imaging takes time, and, for a large enterprise, will take years. “It’s an excellent form of job security,” Roth noted wryly.

Enterprise imaging mindset

Denver Health consists of a 525-bed acute care hospital, eight community health clinics, and 16 school-based health centers. This city-funded healthcare system considers itself to be Colorado’s primary safety net institution.

About 25% percent of Denver residents uses its services, as do many in surrounding cities of the Greater Front Range Metropolitan Denver area. Denver Health operates the only academic level 1 adult trauma center in Colorado, and as such, provides services to trauma patients who reside throughout the state, as well as in other areas bordering Colorado in the Rocky Mountain region.

The organization’s radiology department implemented its first hospital-wide PACS in 2006. This included a VNA, one of the first in the United States to be deployed. It presently runs two parallel VNA systems, one as backup. Denver Health operates a federation of siloed PACS, overseen by Jon Baird, Denver Health’s medical imaging manager. Each specialty PACS has its own database, its own cached workflow, and low- and high-tier image storage.

The reason for these different data sources—specialty PACS silos—is that medical record numbering differs among systems, Baird explains. Some of the systems, including the radiology department PACS, do not have medical record numbers that contain a leading zero or multiple leading zeros. The challenge of reconciling the different databases would require a major investment of IT technologist time and coordination among different IT specialty-specific teams. Baird, for example, heads a group that includes the radiology, dermatology, and dental departments, but not cardiology.

All of this will change in April 2016, when the hospital launches a new enterprise-wide EHR. With this EHR, all of the imaging platforms will have a unified data management structure, unlocking the next step of Denver Health’s enterprise imaging goal.

Baird explains: “The EHR vendor did the radiology IT staff a huge favor, because its system is absorbing all medical record numbers with leading zeros. This means that I don’t have to try to integrate a system that is not mine. Rather, I need to modify my data to match the data of the EHR, a much simpler and easier process.”

The new EHR is also responsible for an initiative now underway to consolidate clinical specialty IT staff. Baird said that to achieve the enterprise imaging goals of Denver Health, groups managing the siloed PACS applications have started to consolidate. As part of this initiative, Baird’s team recently acquired obstetrics/gynecology responsibilities. 

Living with gaps

As with Duke Health, not all departments are currently storing images in the VNA. Part of this has to do with established workflows within specialty departments, and part has to do with medical equipment lacking connectivity.

Cardiac ultrasound is one example. Vascular ultrasound images natively store to the VNA, but the cardiovascular sonographer does not receive orders sent to the ultrasound modality he uses. Baird explains that this individual inputs data manually. He or she also may perform ultrasound at several sites on a patient’s body. Because this system does not structure data at the outset, images may capture in a single exam, and subsequently separated into multiple exams by body part. This process does not provide for efficient workflow.

Dental and ophthalmology imaging systems also are lagging with respect to technology. “A portable digital radiography system has intelligence to it, but few of the systems being used in the ophthalmology and dental departments are sophisticated enough natively to build DICOM,” he explains. “It has been necessary to integrate a software product that can convert them or wrap them in DICOM. As the conversions take place, images are transmitted to their respective PACS and to the VNA.”

Denver Health staff input smartphone images using hospital-supplied smartphones. These are equipped with an app that does not retain images on the phone’s memory after they are sent to the VNA through an import station. This functionality is useful not only to record images of wounds, but also the physical condition of patients who have been abused.

The hospital retains physical abuse photos for use by police, social service agencies, and/or attorneys, but these have additional protection. If a user tries to access them, they are given an electronic “glass door” warning not to access unless they have the authority to do so. Images that are not of clinical relevance can be deleted at will by the individuals responsible for them, and do not need to adhere to the seven-year retention policy for clinical images.

Baird said that efforts to standardize protocol and terminology have just begun as part of the enterprise focus of the new EHR. The radiology department is leading this endeavor, with the cardiology department expected to follow.

“We expect that 2016 will kick off several years of consolidation programs,” Baird predicts. “This will include downsizing some of the storage of our various medical specialty PACS, but our enterprise imaging vision includes the federated operation of the largest ones.”

Baird acknowledges that each specialty has unique requirements. Universal viewers don’t meet specialty needs, and then there is the issue of specialized workflow. “I don’t believe that a commercial system exists yet that seamlessly resolves these issues,” he says.

Hospital of the University of Pennsylvania

The Hospital of the University of Pennsylvania (HUP) in Philadelphia would like to have an integrated PACS as part of an enterprise imaging strategy. In essence, this integrated PACS would consist of a sophisticated image management system interfaced with the hospital’s VNA and run by its EHR.

Records of any image—whether DICOM or non-DICOM—from any medical entity or department would be easily identifiable, accessible without measurable delay, and displayed by a universal viewer that any user could understand, in a format for efficient use, and containing the exact image manipulation and processing tools needed by a particular user.

That’s the Penn vision. Unfortunately, an off-the-shelf system with these capabilities doesn’t exist yet, according to Steven C. Horii, MD, professor of radiology and member of the department’s IT committee. The radiology department is exploring replacement of its PACS in 2016-2017, and has been evaluating new systems, including those that are represented as integrated, all-image-type image management systems. These systems tend to have a selection of image viewers that meet the requirements of various specialties within their systems, but as yet do not offer workflow solutions that work well for all types of clinicians.

The persistent issue is diversity. Not only do clinical specialties need specialized viewers, many also require specialized workflows, creating a magnitude of complexities.

Forcing diversified workflows and diversified image viewing requirements into one system for the sake of having a truly integrated PACS is not recommended. Instead, Horii believes that the strategy of federated PACS makes more sense: Different departments operating their own PACS with as much local storage as is needed. The PACS have their own databases, which are interfaced with a central manager of data and knowledge of the location of all images within a central archive that contains them.

“Today’s ‘integrated PACS’ would require a lot of fine tuning,” he says. “Radiologists are not the only group of physicians who are idiosyncratic. We want images displayed the way we want them and the tools that we need to use them, and so do the other specialties.

“The advantage of a federated PACS strategy—think of this as a collection of PACS all connected together—is that every department has a system that is basically tailored to its needs, but shares a common archive and manager that knows where to rapidly find archived images.”

Managing inefficiencies

HUP’s existing radiology PACS will manage an image from any clinical department and from the Penn Medicine enterprise of hospitals and more than 50 outpatient clinics and physician practices. It must, however, have an order to generate a valid accession number and a report. Creating an order in the radiology PACS requires extra work for emergency physicians who perform bedside ultrasound, but when they do, these images are accessible from the patient’s medical record in the EHR.

The hospital’s vascular lab uses the same ultrasound equipment that the radiology department does, equipment that has the ability to create DICOM structured reports, but the lab staff doesn’t use this function. It also is closed on weekends. Problems arise when the physician of an inpatient who had a venous thrombosis study on a Friday afternoon calls the radiology department to learn the findings because she’d like to discharge the patient. In such a situation, it is sometimes necessary for the radiology department to repeat the study because the images cannot be accessed.

Radiologists have the option of viewing images from the cardiology department either by using a cardiology workstation or by accessing images through the EHR. There also is a PACS workstation provided by the Children’s Hospital of Pennsylvania  for use by radiologists who interpret some of the diagnostic imaging studies performed there, such as the fetal imaging studies that Horii interprets. However, data is not routinely shared with this physically adjacent hospital, and network connections are occasionally severed if one or the other hospital firewalls get rebooted and the workstation’s IP address is accidentally omitted.

Radiologists also are not currently able to access images online from the obstetrics department, although they have access to the reports. Although the ultrasound technology used is capable of creating DICOM structured reports, the department uses a commercial reporting package that requires a technologist to enter all of the measurements he or she tracks and complete a checklist of their observations, Horii explains. This creates a report that enters the EHR. Images are stored on digital media. This is a limitation of the commercial software packages the obstetrics department uses with ultrasound machines that could create DICOM structured reports.

While an integrated PACS may be an impossible dream, HUP radiologists currently are working to standardize terminology among the radiology departments of affiliated hospital entities using the RadLex terminology.

Exam protocols also are being standardized across the system, and work has begun on developing enterprise-wide standard radiology reports and templates. In an effort to simplify and be more cost effective, an enterprise-wide evaluation of medical equipment and modalities is also underway.

While technology may be lagging the ambitions of enterprise image management, there clearly is much work to be done to prepare for the day when technology can deliver a single image experience across specialties and sites—and beyond.