The informatics team at MD Anderson Cancer Center, Houston, from left, Stephen Thompson, MS; Charles T. Suitor, MS, Director, Diagnostic Imaging Informatics; Stan Hildebrand, Manager, Diagnostic Imaging Informatics; S. Jeff Shepard, MS, senior medical physicist; Raimond Polman, systems analyst II; and Kevin W. McEnery, MD, Associate Division Head for Informatics, Division of Diagnostic Imaging.

The division of diagnostic imaging at the University of Texas MD Anderson Cancer Center (MDACC) was like a lot of radiology departments in the late 1990s. Administrators at the Houston facility knew that sooner rather than later the department would have to make the switch from reading off film to reading electronically. There were too many clinical and financial incentives not to. But as a hospital devoted exclusively to oncology, MDACC was caught up in a situation that mandated against a quick switch to soft-copy reading. Everyone, from radiologists to clinicians, was dependent on prior images. Priors are important in any hospital context, but they are constantly in demand when monitoring the growth or remission of tumors, a necessity at MDACC.

To make a smooth transition to soft-copy reading, MDACC radiologists estimated they would need a backlog of electronic priors, as many as 2 years’ worth, online before they stopped reading from film. To diagnostically read hard and soft copy at the same time would have been too cumbersome to orchestrate efficiently. If a PACS (picture archiving and communications system) was to be up and running for 2 years while primary reads were still done from film, that raised the question of how to make best use of the soft images during that 2-year span.

While MDACC radiologists were considering this question, a second technological wind was blowing across the department. Radiologists wanted access to the clinical data that was available electronically on a number of systems in the various departments in the hospital. Pathology, for instance, had been keeping electronic records for two decades, but that data could be accessed only from within the pathology department. A similar situation pertained with numerous pools of electronic text data that were accessible within the domains where they were stored. Radiologists wanted quick access to this data to help them understand what they were seeing on film. The data, they reasoned, would help them dictate more illuminating and useful radiology reports on the diseases that they were encountering as they looked at images in hard copy.

Kevin W. McEnery, MD, was one of a handful of computer-adept specialists in MDACC’s diagnostic imaging division who set about creating a way for radiologists to call up the data they wanted to see on their PC screens as they were preparing their reports. “There was a perceived need within the division to get improved access to information for the radiologists to use when they were interpreting images,” he recalls. “We focused on getting clinical informationpatients’ operative notes, their progress notes, their lab values, their pathology reports. This information was in a variety of different systems.”

Providing clinicians access to information was the driving force behind the MD Anderson Cancer Center PACS.

McEnery, who is an associate professor at the UT MDACC, is Associate Division Head for Informatics in MDACC’s Diagnostic Imaging Division. He is also a full-time musculoskeletal radiologist specializing in soft tissue tumors and sarcomas. He is a self-trained computer expert who learned much of what he knows, he says, while working on an informatics project at the National Institutes of Health while completing his medical degree at Georgetown University. “I am a bridge between the technical and the medical,” he explains. “I answer to the division head, and we are meeting the business needs of the division,” he says. “This is not for fun; what we do has an operations focus. The projects we work on get deployed, whereas in some places, informatics projects are never deployed.” In other words, for McEnery, getting data to radiologists was serious business. Better reports would lead ultimately to better patient care. In the beginning, getting information to radiologists was the sole focus of the data convergence project. That this effort to bring together data for radiologists would quickly expand into a much larger effort to dispense this data to all of MDACCin effect creating an EMR (electronic medical record) for each of MDACC’s patientswas in the beginning neither apparent nor planned, says McEnery.


One of the key players McEnery enlisted in the data convergence effort was Charles T. Suitor, MS, who received his degree in systems management from the University of Southern California. Suitor is now MDACC’s Director of Diagnostic Imaging Informatics in the Division of Diagnostic Imaging. McEnery and Suitor devised a clever strategy for accessing the existing pools of legacy data at MDACC that radiologists wanted to see. They wrote web service-based software interfaces that allowed the legacy data to be tapped and transmitted to web servers, where it could be accessed by radiologists. Web servers are computers that share information with other computers, designed to share information over the Internet, but also useful in an intranet environment. At MDACC, the shared data flows only on campus over MDACC’s private intranet. According to Suitor and McEnery, this intranet is protected from outside access to ensure privacy. Doctors who want to access the data from their home computers must be given special passwords and hardware to link them in a VPN (virtual private network) so they can view the patient-sensitive material.

The University of Texas MD Anderson Cancer Center is one of the world’s best-known cancer centers. It is big. It is licensed for roughly 500 beds and has on staff about 700 physicians, according to Kevin McEnery, MD, associate division head for informatics in the division of diagnostic imaging. He says the facility has about 50 radiologists on staff. Total faculty and staff for the facility number about 12,000.
According to MDACC’s annual report for 2000/2001, the physical plant in addition to the hospital includes 27 outpatient treatment centers and a clinical research building that opened in 1998.
In 2001 MDACC had gross patient revenues of roughly $1.5 billion. That same year it admitted more than 18,000 hospital patients and conducted nearly 500,000 outpatient visits, treatments, and procedures. It conducted roughly 300,000 imaging examinations in 2001.

Software interfaces are essentially electronic codes that allow one data system to send and/or receive data from another one. Suitor and McEnery and their colleagues used the codes Extensible Markup Language (XML) and Simple Object Access Protocol (SOAP) and several other software applications to allow data from the legacy pools to be transferred and viewed by radiologists as they were making their reports. According to Suitor, this essentially meant adding the radiologists’ web servers to the list of bona fide “users” for each of the collections of legacy data. The “clients,” the radiologists’ PCs, could then access the legacy data via the web service and the radiologists could look at it as they were doing their reports. It is an important distinction that using this software interface method did not mean that all the legacy pools of data would have to be replicated in order for the radiologists to view the information. The radiologists would instead tap into the data like any other users of that data system, and they would view the data from the site at which it had been stored.

While this project sounds daunting to the technologically unprepared, Suitor says it turned out to be “surprisingly easy.” Furthermore, after one interface had been written, writing the next one, he says, was less difficult. The data pools that the radiologists could access began to snowball. “In its first incarnation,” says Suitor, “it would automatically display the patient’s previous radiology reports, and lab and pathology reports. Then we added transcribed clinic notes, including recent physicals, discharge summaries, surgical notes, telephone notes, letters, and consultations. Over time, we continued adding data sources.”

This computer application that collected data for the radiologists, McEnery,? Suitor, and colleagues called RadStation. Suitor estimates the cost for the servers to run it at about $200,000. It first went into use in 1999. Word of RadStation’s capability to present diverse data spread quickly. “And then,” says McEnery, “a curious thing happened.” The various cancer specialists at MDACC wanted to use RadStation too. They wanted the same instant access to the legacy data that the radiologists had by simply making a series of clicks on their computers. So McEnery and his group, working with MDACC’s Management Information Systems division, set about creating an altered version of RadStation for use by the clinicians. This version they called ClinicStation. It was different from RadStation in that it focused on the needs of clinicians.

“In radiology, the work flow is basically about reading studies,” says McEnery. “For clinicians, the work flow is about treating patients. In this context, we created an application for the clinicians’ work flow based on the patients they were seeing in the hospital and the clinics. It’s actually the same application except that RadStation comes from the left so to speak and ClinicStation comes from the right in terms of work flow. The information being displayed is the same in both cases.”


The radiology department had known all along that electronic imaging was on the way. Modalities like CT and MRI had been producing digital images since their inception. To begin storing and delivering electronic images, MDACC signed contracts with an imaging ASP (applications service provider) vendor that used a web-based, streaming technology to send images from the modalities to high-end PCs located in clinicians’ offices. Images could also be sent to the wards and departments throughout the hospital and to the outpatient clinics. The web-based imaging system that was installed was chosen because it was a good fit for MDACC. It melded with the web technology used by RadStation and ClinicStation. By adding digital images to ClinicStation so that clinicians could access them with a few clicks on their computers, MDACC had taken another huge step toward making ClinicStation a viable EMR.

No Prefetching-A 24×7, Online Archives
The installation of a web-based PACS at the University of Texas MD Anderson Cancer Center in Houston has eliminated the need to prefetch prior studies on film for clinicians expecting patients. The images are available to doctors electronically on PCs instead. Moreover, there is no minutes-long waiting for deep-archive searches for relevant priors electronically, because with the PACS technology used by MDACC, all images are stored online all the time on magnetic disk boxes called redundant array of inexpensive disks (RAID). The RAID units are located at different points on the campus in a network attached storage configuration.
According to Charles Suitor, MS, director of diagnostic imaging informatics in MDACC’s division of diagnostic imaging, the modalities send the images to the PACS archive, which consists of the RAIDs. There is dual site storage on separate RAIDs on campus, and there is a disaster-recovery copy of all images stored on tape at a location in California, Suitor says.
“For each set of images, the examination is on two different RAIDs,” Suitor says. “If they are both up, we get improved performance because the software will retrieve from both at the same time. If only one system is up, you as a user would not see any difference, but with both systems up, it is a few milliseconds faster.” Suitor says MDACC’s electronic image archive begins in November 2000 and will always run from that date to the present. He says the plan for now is to store everything forever on RAIDs. When one RAID fills up, another one is simply hooked into the system, he says, so the capacity is potentially infinite. “At this point we haven’t set a date for any purging that may take place. We want to be able to access all historical exams. We don’t want to have to prefetch or fetch off some older, slower storage.”
-George Wiley

According to Suitor, in November 2000, the first images went live over the PC network. Cancer specialists did not have to make the trek down to radiology to look at film any longer; they could call up the images on their PCs. By making the images available to the rest of the institution before PACS diagnostic reading stations had been installed for radiologists to read from, the radiology department had essentially, as McEnery notes, “put the cart before the horse.” But there was solid logic behind the move: While a backlog of electronic prior studies was being compiled, clinicians could be weaned from using film. It was backward from the usual scenario, but it made its own sense. At MDACC, clinicians would be leading radiologists into the soft-copy landscape.

“One of the holy grails of PACS has been getting clinicians to rely on electronic images,” McEnery explains. “We thought that if we could integrate the imagesnot as an application clinicians went to separately, but actually integrating the system into their work flowwe would be more successful long term in pushing the institution toward an entirely digital environment. It’s actually a very sophisticated way of approaching information management in the sense that the images need to be integrated into the enterprise in the patient context, not in the context of the PACS system. It’s not about radiology, it’s not about the PACS system, it’s about the enterprise. It’s about the enterprise taking care of patientsand if there is anything that ClinicStation does, it places the focus on the patient.”


What McEnery and everyone else did not know for sure was whether the clinicians would actually use the digital images and come to prefer them to film. The short answer wasyes, they would. According to tracking data on usage that was compiled between January and June of 2002, there was a 250% increase in the use of ClinicStation during that time. In June 2002 alone, more than 1,000,000 documents were reviewed electronically. In that same month 86,238 imaging studies were reviewed by clinicians. McEnery and his colleagues very conservatively calculated that these electronic reviews saved 90 seconds per case of clinicians’ time over using film. Assigning a value of $100 per hour to that saved time, they calculated that ClinicStation was saving MDACC more than $215,000 per month, or more than $2.5 million per year. Using more recent figures, McEnery calculates even greater efficiency savings. In October 2002, he says, 125,000 studies were accessed. During that same time the radiology department created only 18,500 studies, which meant that clinicians were looking at high volumes of electronic priors stored on the PACS archive. Giving a more generous time valuation of 3 minutes for efficiency gained per study over using film, McEnery estimates that having the images on ClinicStation is saving MDACC roughly $7.5 million per year in clinicians’ time, a huge savings for a hardware investment in web servers of $200,000, as Suitor estimated. In the same month, October 2002, more than 500,000 patient queries were made to ClinicStation, says McEnery, which means ClinicStation is being used to access a lot more patient data than just the imaging component.


For an investment of $200,000, ClinicStation is saving MDACC $7.5 million per year in clinicians’ time.

Money savings and use by clinicians are not the end of the story, however. A change in information systems such as occurred with ClinicStation often results in organizational changes. There was concern in the beginning that the deployment of ClinicStation with images would reduce the value of radiologists’ reports for clinicians, because clinicians would be able to quickly look at electronic images and confirm their own diagnoses. This apparently has not happened. Clinicians are looking at more images, but they are looking at even more radiology reports. According to McEnery, in October 2002, when 125,000 studies were accessed, more than 300,000 radiology reports were also called up by clinicians on their computer screens. “That means the clinicians are still finding value in the reports that we issue,” says McEnery. “That means that we, as radiologists, are valuable, because if they didn’t look at my report, I would no longer be valuable.”

Radiologists have remained valuable, but what might be called the imaging culture at MDACC has been profoundly altered, McEnery says.

“One of the nice things about film is that you basically have your clinicians as a captive audience,” he says. “Now with the enterprise distribution of images, the clinicians don’t need you to distribute the film. They just click on the screen and they have it.” What has happened, says McEnery, is that radiologists have become more like consultants, and he says that is a good thing. “We are seeing a lot more consults on the telephone. The clinicians will look at the image and we’ll look at the image, and then between us we’ll come to an understanding of what the pathology is.”

The ultimate beneficiaries of this increased consultation, McEnery adds, are the patients, because a more collaborative effort between clinicians and radiologists means better patient care. That is why he emphasizes that one of the major benefits of ClinicStation is that it has taken the PACS out of the radiology department and turned it into an enterprise-wide patient care tool. “Some places might be threatened by having their PACS be available to the enterprise,” he says, “but we think it’s absolutely crucial to have the images as part of the EMR. If the radiologists are the only ones getting value out of the PACS, then the enterprise is not getting a good return on its PACS investment.”


The successful development and deployment of ClinicStation as a source of the EMR is catching the eye of others, McEnery says. He says the University is looking into patenting and licensing the process that was used to interface with the numerous legacy data systems. Venture capitalists are also taking a look, he adds, and so is Microsoft Corporation, whose software products figured into the creation of ClinicStation. “We have caught the eye of Microsoft,” he says, “because what we have done is integrate legacy systems to meet the clinical needs of patients.”

Now that MDACC has 2 years of prior imaging on its PACS, it is proceeding with the transition to soft-copy reading by radiologists. “We turned off our last analog chest unit in December 2001,” McEnery says. “We have chest x-rays online for 1 year and we have 2 years of CT and MR online. The archive that’s necessary for the radiologists reliably to have all the comparisons that they need electronically is now available. Our hope is to not be printing film by this time next year.”

Filmlessnessand having an institutional EMR up and running as wellis no small accomplishment. It is even more of a feat when you consider that it was initiated and shepherded through by a handful of experts in a radiology department.

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