With the feds brandishing more than $20 billion dollars to encourage health IT adoption and usher in a new era of streamlined healthcare, interest in integration stands at an all-time high.  The ability to integrate, share and route data among disparate systems is essential. And the rationale extends beyond dollars. “This isn’t just another way to deliver information. It’s a way to stay connected to the communities where you do business, so patients can get the best possible care,” says Susan Hollabaugh, project manager, clinical information at RadNet, Inc., in Los Angeles.

The road to interoperability is riddled with obstacles. For starters, there is the sheer universality of integration; it is an ongoing enterprise and cross-enterprise project that comprises hundreds of imaging and IT systems including the EMR, PACS and much more. Current priorities range from image access in the EMR to the deployment of a vendor-neutral archive. It’s no surprise that stakeholders’ understanding of integration is uneven; many may not have a clear picture of their needs. In addition to appreciating the enormity of the undertaking, facility teams should anticipate logistical, leadership and communication challenges. On the plus side, vendors’ openness to interoperability has improved in the last few years.

This month, Health Imaging & IT visits a few pioneers to learn more about the process of integrating data among disparate IT systems.

Customers first

RadNet is a massive outpatient imaging network and integration powerhouse. The 188 imaging center practice currently relies on two dozen separate outbound interfaces to share data among disparate EMRs and health information exchanges (HIEs) and aims to complete close to 100 by the end of 2011.

Current RadNet integration solutions run the gamut from radiology reports and image exchange to a pilot order entry project. About 90 percent of the company’s integration requests center on electronic results distribution. But every provider has unique needs, and their needs evolve as they comprehend the potential of a mere interface. Some providers want a link to the PACS viewer in the report; others ask for key images embedded in the report. Finally, image-intense specialists, like radiation oncologists, want DICOM images routed directly into the treatment planning system.

RadNet’s Hollabaugh preaches provider-centered flexibility. “We want to be the easiest people to do business with. We are open to trying anything that might meet our customers’ needs.”

In each case, RadNet aims to determine the provider’s specific goals by asking physicians how they plan to use the data and when and where in the care stream they need access. The best approach, says Hollabaugh, begins with meeting providers to help them become more informed. “Many times they don’t know what questions to ask; the provider may not understand implications of a workflow.”

Asking providers and vendors probing questions often produces interesting answers, says Hollabaugh, and RadNet finds the pat answer isn’t always right. For example, HL7 isn’t necessarily the best way to share reports because the text-based standard does not generate a crisp presentation in the EMR. Sometimes, a PDF or mime-encoded file better meets users’ needs, she explains.

In other cases, users haven’t fully grasped their needs. For example, although RadNet developed multiple options to give many radiation oncologists what they thought they wanted—DICOM images—none of the initial solutions quite met the clinicians’ needs. Ultimately, after RadNet staff met with radiation oncologists, they realized the physicians wanted to log into RadNet sites and pick images to load into treatment planning systems. A pilot project that securely transmits DICOM images into treatment planning systems is “very promising.”

After meeting with the provider, Hollabaugh and her team work with the vendor to craft a solution. Hollabaugh tosses a fair amount of vendor finesse into the process. Recognizing that vendors are busy with meeting meaningful use requirements, recertifying systems and handling an ever-growing numbers of requests for interfaces, Hollabaugh tries to begin the interface process by partnering with a practice that has a very good relationship with the vendor. When the target interface works, it can be transferred to other customers.

The universal patient

Cleveland Clinic epitomizes the connected enterprise. Its core EHR provides physicians at any of its 11 hospitals or 17 family health centers access to patient data and images, and its DrConnect application gives physicians outside of the enterprise access to a secure portal where they can view all of their patients’ Cleveland Clinic activities.

One of the clinic’s primary integration projects is focused on improved image access across the enterprise. The new model will replace a point-to-point, first generation model that connected PACS to the EMR. Physicians read the radiology report in the EMR and clicked on an image icon to view images. Other image viewing processes like EKG and digital photography relied on separate point-to-point solutions.

The Top 5 Integration Pitfalls

The road to interoperability is paved with landmines. The enterprise sets out with the best of intentions: streamlined access to a holistic patient view rich with images and information. It is easier said than done, and all too often, integration falls prey to politics, budgets or mediocre leadership. Interoperability expert Paul Chang, MD, professor and vice-chairman of radiology informatics at the University of Chicago School of Medicine, identifies the top five no-nos and provides a few tips for bypassing them. 1. Radiology is not the center of the universe, and a radiology-centric bias courts disaster. Radiology is not the only specialty that needs or creates image data. This means the EMR web viewer needs to launch cardiology, endoscopy and dermatology images, in addition to x-ray and CT images. Work with IT and the ‘ologies to devise an interoperability model that will scale throughout the enterprise, advises Chang.   2. Don’t overlook context. At the second level of integration, the focus is on getting multi-ology images into PACS. “Getting the images into PACS is easy; trusting the image is hard because there is no RIS to provide context like an accession number,” warns Chang. There are two options. Co-opt the RIS to schedule ‘ology studies, or create a pseudo RIS to provide accession numbers and key data for the ‘ologies. 3. Don’t proliferate mini archives. Every archive requires HIPAA-compliance, which is expensive and unwieldy. The glitch? Every specialty workstation requires an archive to support its functionality. Chang recommends facilities deploy peripheral limited persistent caches that hold six months of imaging data to support workstation functionality. Leverage the enterprise PACS for the true archive. “It saves money and lets users have their cake and eat it, too,” he explains. 4. Don’t fall into the one-size-fits-all trap. Imaging consumers are appropriately idiosyncratic, and software should support their workflow rather than force them to adapt to it. Tap into IT to modify software to meet the diversity of clinical needs. For example, at the University of Chicago, essential data like pathology results are available for radiologists in the RIS. There is no need to log into a separate system to complete the regular task of checking pathology results on a liver biopsy. 5. Remember that the EMR is not a singular system; it is an ecosystem. The systems need to work together to orchestrate the same data in a variety of ways to meet the particular needs of specialists, which means users across the enterprise need to communicate and collaborate to optimize the EMR.

Cleveland Clinic is evolving away from this point-to-point imaging strategy to an enterprise image distribution model that encompasses all departments, says C. Martin Harris, MD, MBA, chief information officer. Each image-generating department will use a departmental system to acquire and store images locally. The local archive supports the immediate work of the department, while the enterprise archive houses the permanent image and connects images to the EMR. Each departmental system will feed into the enterprise distribution system, sending copies of all images to the enterprise system. Images will be purged from the local archive after 90 days with the permanent copy housed on the enterprise archive.

The project started earlier in 2010; by the end of the year all enterprise radiology images will be transferred to the enterprise system. The next steps are the image-generating ‘ologies: digestive medicine, women’s health, ophthalmology and pulmonary medicine. “The process will be ongoing because every specialty is developing an imaging capability,” shares Harris. 

An enterprise steering committee comprised of representatives from radiology, IT, orthopedics, digestive medicine, neurosciences and more developed the new model. Involvement across the enterprise is critical because each stakeholder group needs to lay out its requirements and build them into a series of use cases: ambulatory office, procedure room, inpatient and OR. “Use cases are very important because image use varies by location. We needed to understand how images are acquired in each setting, how they would be archived in the enterprise system and indexed to the EHR,” explains Harris.

The vendor-neutral archive

Three years ago, Marshfield Clinic in Marshfield, Wis., embarked on a project to deploy a vendor-neutral archive to consolidate and manage images from internal disparate RIS/PACS, external image management systems, non-DICOM datasets and scanned images. “We’ll always need separate PACS to support various reading environments,” says Director of Ancillary Applications Lane Solverud, “but the vendor-neutral archive is the long-term enterprise archiving solution.”

Imaging data, including primary radiology images, digital photographs and scanned documents, are available in the EMR; and an embedded DICOM viewer provides the ability to manipulate radiology datasets. Thirteen terabytes of non-DICOM images also are slated for the vendor-neutral archive to meet image management needs within the enterprise.

The clinic, however, has encountered a few glitches. For example, different departmental and organizational workflows present an ongoing conundrum. Initially, stakeholders had agreed to standardized workflows; however, in practice the clinic has entered the “accommodation business,” that allows departments and organizations to use existing workflows, admits Project Manager Joyce Heintz. “At some point, we have to decide if we want to spend development funds on workaround workflows [or insist on standardized workflows],” states Heintz.

The workflow accommodation issue is fairly significant. In the simple, single-site radiology image management era, it was fairly easy to distinguish a raw image from a reviewed image, saving only the reviewed image in the PACS. Fast forward to the central archive linked to multiple organizations. The clinic may need to return both the raw image and golden image to some outside organizations, says Solverud. The process requires a different workflow.

Another ongoing challenge is the buy versus develop debate. Some site-specific features are essential; however, vendors may be less than eager to develop features that don’t provide widespread utility into their products. “When our interests start to diverge from the vendor, we have to decide if we should pay them for an enhancement or if we should develop it on our own,” explains Solverud. For example, the clinic internally developed a DICOM gateway to provide advanced routing and tag manipulation so image data could be moved among various systems rather than asking its vendor to develop a feature specifically for Marshfield Clinic.

Integration Help Desk

Experts who have a head start on the road to interoperability share a few pointers with their colleagues. Approach hospitals with a win-win attitude; a clear explanation of why integration benefits both parties ups the odds of a successful project, says Thomas. Work on documentation throughout the process—document what’s being done, what works, what doesn’t work and intricacies of the project, says Hollabaugh. Double your time estimate, recommends Thomas. It takes time to build relationships and trust with partners. Plus, system testing devours massive amounts of time. Think about the long term to minimize impacts of changes made down the road and avoid degradations in services from future upgrades or changes, recommends Hollabaugh. Don’t neglect outside partners, says Harris. “Think outside the organization and how images might be used outside of the organization.”

The unified worklist

Ten years ago, Austin Radiological Association in Texas, read the writing on the wall and invested in a high-end image management system and enterprise SAN. “Our strategy was: build it and they will come,” explains CIO R. Todd Thomas. The strategy worked.

In the last decade, the practice parlayed its initial investment into an ASP service for most of the 27 organizations that contract with it for reading services. The ASP model makes for smooth integration; however, two hospitals use PACS from other vendors, necessitating a hybrid model. 

Austin Radiological developed a specialized application to read data from disparate PACS and present a unified worklist that launches the appropriate PACS viewer. “It’s not completely integrated and interoperable by definition,” admits Thomas. However, it does get the job done.

Politics, rather than technology, has hampered the practice’s integration efforts. “Some hospitals are very protective of their data. It isn’t easy to get past firewall restrictions to get the real-time data needed to feed into our system.” One security-conscious hospital holdout, for example, requires the practice to use its workstations for image review.

The next steps

As interoperability gains steam, processes and practices should become better defined. Know-how will be disseminated among vendors and providers. And business needs may sway reluctant organizations. For example, an upcoming expansion of Austin Radiological Association’s voice recognition system may persuade protective partners to relax firewall restrictions because the practice needs real-time HL7 feeds to get data into the voice recognition system. In addition to attending to “small” details, organizations need to think big. “The future is whatever is captured as part of the medical exam,” says Marshfield’s Solverud. Think voice clips, movies and more. These datasets need to be housed in the vendor-neutral archive and shared via the EMR, requiring new twists on integration. It should be an interesting ride.