UPMC’s Rasu Shrestha, MD, MBA: Improving the Value Proposition of Imaging Informatics
The University of Pittsburgh Medical Center (UPMC), with 20 affiliated hospitals and 30 imaging centers in western Pennsylvania, could be seen as ground zero in the effort to digitize medicine. From its innovative financial and development partnership with IBM to a recently announced pact with Google to develop a personal electronic health record (EHR), UPMC is blazing multiple trails for electronic communications in medicine. Radinformatics recently discussed the value of imaging informatics in medicine with Rasu Shrestha, MD, MBA, medical director of digital imaging informatics at UPMC, radiologist, and the steward of imaging informatics across disciplines, as well as the enterprise. Radinformatics: As a radiologist and the medical director of informatics in one of the nation’s most wired hospitals, does radiology get better than average support, or is it fighting just as hard for resources as everyone else? Describe radiology’s relationship with information services. Shrestha: We believe in the policy—or the reality—that there is one bucket. You might take out resources from one place and put them in another, but at the end of the day, it does go back to that one bucket. What is definitely true is that bettering health care starts with first having the right infrastructure and the right tools in place. Overall, whether in the clinical departments or information services division, we are a driven bunch of folks. The relationship between radiology (and any of the other -ologies) and information services is extremely complementary, and it is structured for efficiency and innovation. I am charged with strategizing for all of digital imaging informatics, including radiology, cardiology, pathology, ophthalmology, and any other -ology that entails imaging, which could mean otorhinolaryngology, gastroenterology, and wound care. It’s fairly unique and focused, as opposed to having fiefdoms across the board. The idea here is to have more of a unified strategy for digital imaging. With radiology charging ahead as PACS matures/evolves, we bring along the lessons so feverishly learned over the past dozen-plus years in radiology to the rest of the enterprise, whether they are successes that can be replicated or things absolutely not to repeat as more of the -ologies embrace the digital era. This confers more of a synergistic strategy for imaging and how it meshes with the rest of the electronic medical record. That is where we are at UPMC. Radinformatics: Does the radiology department at UPMC have its own IT department? Shrestha: There is a larger enterprise imaging team in radiology—consisting of support personnel, system analysts, and developers—just as there are other teams within other departments. These teams do fall under the information services division, except that they are all driven by the same motivation as the clinical departments—to better patient care, constantly improving the coordinated delivery of care across the UPMC enterprise. There are teams under these divisions, and I oversee these teams across the -ologies. We work hand in hand with each of the clinical departments to serve that master goal. I work extremely closely with information services. I wear many hats, and this is essential, especially in a complex environment like ours here at UPMC (and, arguably, in other institutions as well).
At the same time, when we either make demands or experience pain points in the clinical environment, I relay those. I have worked very closely with the CIOs across all of our hospitals, directly under the CMIO, and, at the same time, with our physicians on the ground. Radinformatics: With hospitals focusing on efficiency and cost, how much of your section’s resources are going toward serving those masters? What are some of the initiatives you’ve launched lately? Shrestha: There is no doubt that these masters need to be served. We strive to have what is in the best interest of efficient and effective patient care drive our decisions and the actions that might result. It is not just the bottom line that drives all of these decisions. We try to replicate successes across our hospitals, whether those are standardized protocols or even applications, while at the same time catering to some of the different flavors that define core academic hospitals (compared with some of our community hospitals). That helps with the efficiency and cost equations. There have been multiple initiatives. One fairly large initiative would be one that I call imaging interoperability. The goal here isn’t just to address the master questions of efficiency and cost, but essentially and primarily, it is to improve patient care. For instance, we built an application called SingleView with the goal of essentially federating our patients’ imaging studies and reports across the UPMC enterprise, across all of our 20 hospitals, with unified access to all images and reports across all of the different -ologies, not just within the many different PACS we have across the hospitals.
I function not just as the person charged with working with the various teams and leadership to come up with the strategies for enterprise imaging across the board and how it meshes with the EHR world, but also, at a basic level, as a translator. I am the clinician who can relay any downstream effects of IT decisions that are made—not just outside the reading rooms and in clinical corridors, but to the hospital.
—Rasu Shrestha, MD, MBA
We went live with the SingleView application primarily so we are not blindly ordering studies. For instance, if we are looking at this chest radiograph with a big mass in the left lower lobe, and are inclined to order a CT angiogram (CTA) for this patient, we now have instantaneous and accurate access to all of the patient’s prior imaging studies and reports from disparate archives and systems. Effectively, this access also is improving patient safety and addressing some of the issues around radiation dose. Overall, what that is leading to is cost savings. We do serve the masters of improving efficiency and decreasing costs, but it starts off, first, with the goal of improving patient safety. Radinformatics: How important will informatics be in addressing and supporting quality and safety initiatives in radiology? For instance, how and where should radiation dose be measured? Shrestha: I strongly believe that informatics will play a pivotal role in enabling us to pursue quality and safety initiatives. Even as we become more saturated with higher-end, multislice imaging modalities that spew out ever–thinner-cut data (at lower radiation doses and with quicker scan times), the challenge will be in figuring out how best to manage the tsunami of datasets. The standard of care in the CT world isn’t really the single-slice or four-slice CT scanners: it’s the 64s, with a fairly rapid progression towards the 256s and higher. How do we make sense of these huge datasets that are being thrown at us, and at the same time, make sure that we are conforming to quality and safety standards, primarily around the question of radiation dose? With some of these newer modalities (for example, the 320 as opposed to the 64), there is actually more control in terms of the amount of radiation dose for the patient, primarily because these scanners are more efficient and the scan time is quicker. Having said that, there is no doubt that we are scanning more these days than ever before. The trend has been toward a more rampant ordering of these so-called high-tech studies, left, right, and center. It poses many challenges for referring physicians, radiologists, technologists, physicists, and even the vendors. Ideally, we would be collecting this information to know how much exposure there is for each of our patients. We are starting to do that; for instance, the ACR® has begun a couple of initiatives to start collecting this information.
It allows for enhanced quality-assurance, quality-improvement, and continuous–quality-improvement initiatives, but at the same time, it saves time, improves clinical efficiency, and increases patient safety. We also started to notice a dramatic decrease in our National Imaging Associates denials.
—Rasu Shrestha, MD, MBA
If we have this information intelligently embedded into the computerized provider order entry (CPOE) system—so at the ordering end, the physician who is about to order yet another CTA for this patient gets a vivid idea that this patient may actually be glowing—then the test won’t be ordered if it’s not needed. Essentially, this meshes with the logic behind the decision-support rules that are behind the CPOE system. That would be an ideal scenario and something I’d like to strive toward. I am not sure there is a product or a service that does this, but wouldn’t it be nice to have a score that follows the patient wherever he or she is seen, not just across the hospital or UPMC, but across the state or the country? No matter where the patient goes, there would be a score that is essentially the cumulative dose of radiation that he or she has acquired in a lifetime. These are the questions that come to mind, and the answers obviously have a very strong flavor of informatics. Radinformatics: How might informatics enhance the radiologist’s ability to get more information from a given study? Shrestha: That goes back to imaging interoperability. I like to tell the story of how a particular patient doesn’t really come into the hospital to get a CT study; he comes in because he has a headache, or because he bumped his head, and he is rushed into the emergency department. He is treated by the primary-care practitioner or emergency-department physician, who may refer the patient to the ophthalmology department for a series of tests, possibly including a retinal scan. The ophthalmologist looks at the image and may generate a report. Perhaps the ophthalmologist determines that this patient has a growth at the back of his eye, and feels that he needs to determine the extent of the growth by taking a CT scan. It’s only at that point that the patient ends up in the radiology department. Maybe, by that time, he has had a series of laboratory tests, and he may have a pathology report. Maybe there is further searchable history; further actionable data hat is critical to the story behind the images. Traditionally, in radiology, all we’ve been doing—and this is both before we had PACS and after we embraced PACS—is treating just a series of images at a time. We obviously are doing it more efficiently today than we ever have before, but if all we are doing is treating a series of images at a time, then we have to ask ourselves, “Are we doing the right thing?” Are we really doing what is right for the patient? What we ideally should be doing is treating the patient as a whole, not just treating a series of images at a time.
Wouldn’t it be nice to have some of this information available upstream at the ordering level? Think about it: When I, as a radiologist, sit behind a workstation and open up a study, it’s too late. The study has been ordered. It could be an inappropriate study that was ordered, but the procedure has been performed; the patient has gotten a good zap of radiation and has left the hospital and moved somewhere else in the enterprise.
—Rasu Shrestha, MD, MBA
That is what we are striving to do at UPMC. What that means is having a more holistic view of the patient and having more information that is readily available specific to that patient. It is more of a patient-centric approach than an image-centric approach. If we are able to extract the relevant information from the plethora of information sources that often define what the EHR universe is for that patient, and we have that available to that radiologist at the point of care, then suddenly, we are talking about a new paradigm in the way we are able to practice digital imaging. Radinformatics: You coauthored a paper, a few years ago, that described an interface for viewing images that incorporated information from across the biological continuum, from viscera to genes. Is this the holy grail? Shrestha: It was an attempt to point at what could be the holy grail. What we were trying to do, with that paper, was to point out the fact that a patient doesn’t just live in a silo. A patient is more of a story, and there are contributing factors to what could be the pathology you are looking at or the region of interest that you are trying to home in on in a chest radiograph or a CT study. I would think that the optimal, next-generation clinical-information system would be patient centric, evidence based, and data aware (with aspects not just of decision support, but of intelligent decision support). With it, we would have knowledge representation and information-extraction and structuring; it’s not just about information silos anymore. Radinformatics: What are the implications of this, both for providers and for reading groups? Shrestha: One is that we get an opportunity to start treating the patient as a whole, as opposed to treating a series of images at a time. That’s huge, and this in itself will have various implications for providers across the board. There is, however, an increasing opportunity for the emergence of a newer business model where we have integrated diagnostic centers that might become the reality in a couple of years.
Let’s move away from this culture of just treating a series of images—a habit somewhat rightly reinforced not just by the reimbursement models that we have, but also by the existing IT systems that we use and by our clinical work habits. If we could start treating the patient as a whole, that would be so much better, not just for the health system, but also for the patient.
—Rasu Shrestha, MD, MBA
Scanning, image transmission, storage, viewing, as well as everything I talked about earlier in terms of imaging interoperability can all come together and potentially lend themselves to the benefits that referring physicians and patients need. In the area of women’s health, when we are looking at mammograms, for example, today, we very often look at pathology images and reports to try to correlate our findings with the pathology. As pathology enters and embraces the digital era, I see the emergence of these integrated diagnostic centers. Radinformatics: Describe the radiology department of the future. Shrestha: There is no doubt that radiology will play an increased (and critical) role as the central nervous system of the hospital. The fears of some that radiology has become commoditized will—and should—be combated by specific initiatives around two main goals. The larger goal is to bring the very value of radiology back to the clinical community at large, and the other is to focus on quality and not just quantity. This can be extrapolated to all of the other imaging-centric -ologies as well, but if the value of a radiology department is condensed to that one report that we generate and fax out to the referring physicians or send back to the clinician base through an HL7 interface of some sort, let’s dramatically enhance that radiology report by making it richer, more meaningful, and hence more useful to the ordering physician. Essentially, let’s use this output, the imaging report, to bring the value of radiology back to the clinical community at large. When you increase the value by incorporating meaningful data, embedded dynamic images that typify the findings, and more, you directly negate some of the fears that radiology is becoming commoditized. There is also a definite and purposeful focus on quality rather than quantity. You can very easily have radiologists who are sitting in a foreign country or a different state provide mass readings for your reports, possibly at a lower cost, but we’re moving away from just looking at quantity to looking at quality. We’re realizing how much more valuable an interpretation would be if I actually had a subspecialist radiologist, sitting somewhere within my hospital environment, who had a better idea of what this patient has been through. How much more valuable would that report be to me, and what does that mean to patient care—and even to the bottom line? I think there also will be a focus away from what has been a flourishing of applications used in image interpretation that remain disparate and siloed. Is it a RIS-driven or PACS driven workflow? Do I now move from my PACS to the voice-recognition system to dictate a case, then over to a different system to do the 3D postprocessing—but wait, there’s the computer-aided detection system, and then back to the dictation system. The focus needs to be less on specific applications and more on a seamless workflow around the patient. The other big focus will be on bringing communications back to the practice of imaging in general. A side effect of being successful with PACS deployment was taking away (and, to a large extent, negating) the communications factor. Today, we are talking less with our surgeons and our clinical counterparts in the other departments.
In addition to the development that has already happened in radiology, there is also a fair amount of work going on in digital pathology. What that spells is more of a paradigm shift in the very practice of health care. There are many diagnostic-interpretation synergies between radiology and pathology, and this does lend itself to the creation of a broader and more synergistic diagnostic template.
—Rasu Shrestha, MD, MBA
Traditionally, radiologists are seen as providing their interpretations—a clinical road map for surgeons and clinicians. I think we are moving away from that printed-roadmap concept to more of a real-time GPS concept, with enhanced communications, true imaging, and data interoperability for more of a bidirectional dialog between the radiologist and the referring physician. In this paradigm, as things mature or change (or if there’s a significant clinical event that necessitates a sudden U-turn to be made—for example, interoperatively), there’s more direct input from the radiologist than there was in the past. The imaging department of the future will have to deal a lot more intelligently with data in general. We are at the brink of data and information overload. As we embrace the digital-imaging realm across our health-care enterprise, we will continue the transition from terabytes to petabytes of data. How do we tame the beast so that we are talking more about actionable data and knowledge and less about the silos of information systems that we have? Cheryl Proval is editorial director of Radinformatics and is vice president, publishing, for imagingBiz, Tustin, California.
We are being relegated to the dark corners of the reading rooms in hospitals to read that study and give the results in the shortest possible time. If we could leverage the technologies that we have around us today with unified communications or other such tools, we could bring the art of communication back, so that we could reengage the radiologists in the clinical processes where we have surgeons and radiologists talking to each other. Indeed, we have an opportunity to enhance this paradigm and have a much more meaningful dialogue.
—Rasu Shrestha, MD, MBA