Closing in on Outcomes: Radiology Attempts to Close the Loop

For many years, the pursuit of outcomes in radiology was not a common practice for several reasons.  A marked logistical divide separated patient outcomes from imaging exams, in large part because of a lack of communications channels between radiology and referring clinicians as well as other providers. The absence of metrics measurements, as well as the fact that radiologists did not “own” their patients, only complicated matters.

More recently, a considerable shift has occurred in this arena, with individual institutions and hospital systems emphasizing radiology’s role in patient outcomes and engaging in radiology outcomes research initiatives. Supported by grants and sponsorships, research activity also is occurring on a national level.

A significant portion of this change stems from developments in healthcare, such as the widespread adoption of electronic health records (EHRs) and, in turn, the increased ease with which multiple practitioners involved in different aspects of patients’ care can access information about how that care is progressing. Payors—including, but not limited to the nation’s biggest payor, the Centers For Medicare and Medicaid Services (CMS) —have changed their priorities, with a heightened, seemingly singular focus on value.

“The tie-in between value and reimbursement has really been a catalyst here,” according to Ruth C. Carlos, MD, MS, Professor of Radiology, Division of Abdominal Radiology, University of Michigan Health System, Ann Arbor, Mich. Other catalysts have also been at work, she adds, noting that the definition of outcomes has expanded beyond clinical parameters. Carlos cites as examples the quality of life and decisions patients must make about their care.

C. Craig Blackmore, MD, MPH, director, Center for Health Care Improvement Science, Virginia Mason Medical Center, Seattle, corroborates Carlos’ comments. “It’s still difficult to make a connection between imaging and whether a patient lives or dies, but access to metrics and other data do make it easier to determine whether certain imaging studies have the potential to improve patient care and outcomes, or not,” he asserts. “This is important to multiple audiences—patients, providers and payors.”

A growing emphasis on patient-centered care also is pushing the envelope. “One of the cornerstones of patient-centered care is involving patients in making decisions about their care,” Rebecca Smith-Bindman, MD, professor in residence, radiology, epidemiology and biostatistics, obstetrics, gynecology, and reproductive medicine, University of California San Francisco (UCSF), states.

When less is more

Smith-Bindman, who also serves as director of UCS’s Radiology Outcomes Research Laboratory, explains that such patient involvement has increasingly come to mean utilizing the results of imaging studies to determine whether, for example, to opt for surgery or some other mode of treatment, and to weigh the benefits of different approaches to care against each other. “There’s a higher level of expectation now to show that what we’re doing helps patients,” Smith-Bindman asserts. “Payors and patients want to see that, although there’s often value in doing more, sometimes there’s also value in doing less. It’s important to know when using new, sophisticated, expensive-to-use technology is the better course of action, and when less expensive no-radiation technology will do the job.”

She cites as an example a three-year study conducted at UCSF and 14 other institutions, using a $9 million grant from the U.S. Department of Health and Human Services Agency For Healthcare Research and Quality. Investigators, Smith-Bindman explains, looked at whether CT scans were any better than “less-often-used” ultrasound procedures to diagnose kidney stones in hospital emergency rooms.

Emergency room patients whose pain was suspected to be due to kidney stones were randomly assigned to one of three groups. Patients in one group received an ultrasound exam performed by an emergency room physician on site. Those in a second group underwent ultrasonography performed by a radiologist; those in the third group, an abdominal CT scan also conducted by a radiologist.

With six months of patient follow-up, the study revealed that health outcomes for 2,759 patients were “equally good with ultrasound as with CT,” Smith-Bindman notes, and that patients fared no worse when ultrasound exams were performed by emergency physicians rather than radiologists. Additionally, serious adverse events, pain, return trips to the emergency department, and hospitalizations did not differ significantly among groups.

Smith-Bindman says the study results did not and do not suggest that ultrasound alone be used to assess patients with suspected kidney stones, but rather, that ultrasonography should be used as the initial diagnostic tool, with further, more involved imaging studies performed should the physician’s clinical

judgment warrant. “What this does show, though, is that there can be value in doing less,” she says, characterizing studies of this type, which are intended to help differentiate between when imaging is necessary and when it is not, to be low-hanging fruit for institutions and groups of institutions that want to get started with imaging outcomes research.

At Virginia Mason, outcomes research has entailed assessing the value of decision support systems to ascertain whether imaging is even necessary. “Evidence-based use of imaging ties into outcomes, and clinical decision support is, as we see it, potentially an ideal means of improving it,” Blackmore says. “Clinical decision support systems are not only consistent; they are educational, transparent and efficient.”

Blackmore describes a retrospective cohort study conducted at his institution, which looked at the “staged” implementation of evidence-based clinical decision support built into ordering systems for selected high-volume imaging procedures: lumbar MRI, brain MRI, and sinus CT. Imaging utilization rates and overall imaging utilization before and after the intervention were determined. The study revealed that rates of imaging after intervention were 23.4% lower for low back pain lumbar MRI, 23.2% lower for headache head MRI, and 26.8% lower for sinusitis sinus CT.

Cracking the code

Among other institutions undertaking radiology outcomes research endeavors is the Hospital of the University of Pennsylvania in Philadelphia, where a project known as Code Abdomen got underway in July 2013. Hanna M. Zafar, MD, MHS, assistant professor of radiology at the hospital’s Perelman School of Medicine, was one of the physicians spearheading the project, along with co-investigator Tessa Cook, MD, PhD. Other key members of the team were programmers Darco Lalevic and Christopher Pizzurro and department chair Mitchell Schnall, MD. PhD.

In conceiving Code Abdomen, Zafar and her colleagues acknowledged that while radiology has mastered the first two levels of efficacy in diagnostic imaging—technical and diagnostic accuracy—fulfilling the mission to improve the value of care as dictated by healthcare reform would entail going beyond these levels. “Diagnostic efficacy, treatment efficacy and societal efficacy are where the interest lies when it comes to improving the value of care delivered through radiology,” she explains.

Other forces were at work as well. There was a desire to better correlate imaging with outcomes, and to standardize variables in reports across radiologists, modalities and hospitals in the University of Pennsylvania Health System. The frequency with which imaging studies uncover incidental findings that may represent cancer (which Zafar says are seen in up to 18% of patients with no known cancer and 31% of patients with known cancer), and a perceived need for radiology reports to clearly communicate the malignant risk of these findings while also sharing specific follow-up recommendations, figured equally into the equation.

The group was inspired by the success of the ACR’s Breast Imaging Reporting and Data System (BI-RADS®), which has improved communication among radiologists and clinicians through standardized report content, consistent lesion classification and actionable language. Consequently, Code Abdomen’s design follows the BI-RADS model.

The system initially focused on the liver, adrenal glands, pancreas and kidneys because of the frequency and clinical importance of focal masses in these organs. It uses nine codes—0 through 7, plus 99—to assign lesions to five categories: benign, indeterminate, suspicious, known cancer, and non-diagnostic.

Codes are applied only to fully visualized organs, with Code 99 assigned to organs only when technical factors make it impossible to exclude focal masses. Unlike other pre-existing standardized coding schemes, Zafar observes, Code Abdomen was envisioned and configured as a global coding scheme under whose umbrella a uniform code is consistently applied to each of the four organs on CT, MRI and visualized organs on ultrasound examinations, no matter why the study was ordered and where the patient was imaged.

Code Abdomen aligns with such coding schemes as the Bosniak classification system for cystic renal masses and the Liver Imaging Reporting and Data System for patients at risk of hepatocellular carcinoma. Where coding schemes do not exist for specific organs and clinical indications, radiologists subjectively assign the malignant likelihood of any visualized masses, helping to integrate Code Abdomen into radiologist workflow. In organs with multiple findings, radiologists must assign the code that reflects the most suspicious lesion in that organ or, in the case of the kidneys, pair of organs.

Radiologist compliance is monitored to ensure that relevant codes are included within all abdominal radiology reports. Subsequently, an automated system mines radiology reports daily to identify patients with codes corresponding to masses that are indeterminate or suspicious for cancer. These patients are moved to an ‘active’ follow up queue.

Closing the loop

The system automatically updates a patient’s status if follow-up does occur and a subsequent image study addresses an organ for which follow-up is recommended. It has been tweaked several times, once to allow for special considerations of lesions in the indeterminate category, whereby radiologists were required to provide guidance to ordering clinicians on both a modality for follow-up and a timeline within which the follow-up should occur.

“After the follow-up piece was added, we were able to begin reviewing the data as a whole and start moving towards really measuring outcomes,” Zafar says. Manual chart review of patients who received imaging through December 2013, established that 9% of patients with findings indeterminate or suspicious for malignancy had relevant pathology follow-up.

The system debuted at the University of Pennsylvania Health System main hospital and was subsequently implemented at two other system hospitals, with plans for a system-wide rollout and further refinements.

Eventually, the Code Abdomen will issue automated notifications to referring clinicians caring for patients with masses indeterminate or suspicious for malignancy with no follow-up with the system, Zafar reports. The goal will be to discover if follow-up was not clinically indicated; the patient received another form of follow-up; or follow-up inadvertently was not completed. “This way we can include that data and close the loop on those patients,” Zafar notes.

Zafar and her team are working to gather long-term data on rates of malignancy by code and by organ. “Ultimately we want to tailor reports so that we can say, ‘Your patient has a mass assigned a code ‘x’, and within our system, x% of these masses turn out to be cancerous,” Zafar explains. “It’s difficult to pull that likelihood from free text reports. Through the uniform use of codes within radiology reports, we can mine reports to detect patients with specific codes, and we can correlate those codes with data in the electronic medical record to examine outcomes. Although the system is currently only used in the abdomen, it is applicable to other organ systems, including pulmonary nodules.

Meanwhile, on a national level, both Smith-Bindman (in conjunction with the Radiology Outcomes Research Laboratory) and Carlos are focusing broadly on investigating both the negative and positive impacts of imaging. Smith-Bindman is the principal investigator in a five-year, $10.9 million multi-site study under the aegis of the National Institutes of Health (NIH) that seeks to evaluate patterns of medical imaging and associated radiation exposure in pregnant women and pediatric patients; as well as to determine the risk of pediatric cancers from exposure to radiation in utero and during childhood.

A second, $14.5 million study in which Smith-Bindman is involved is being funded by the Patient-Centered Outcomes Research Institute (PCORI) and led by Michael K. Gould, MD, MS, associate professor of medicine (pulmonary and critical care), Stanford University, Stanford, Calif. The multi-site study seeks to identify optimum surveillance strategies (imaging) to maximize early diagnosis for individuals with cancerous lung nodules without subjecting patients to unnecessary, excessive surveillance procedures.

For her part, Carlos is participating in a multi-center study of MRI and multi-parameter gene expression assay in ductal carcinoma in situ (DCIS) being conducted through ECOG-ACRIN. The goals of the trial are to determine the proportion of patients undergoing mastectomy after MRI has been integrated into the management of DCIS patients who would otherwise be candidates for wide local excision based on standard mammographic imaging, physical examination and a diagnostic core biopsy demonstrating DCIS (without invasion or micro-invasion); to correlate findings with the DCIS score; to evaluate patient-related outcomes when managed with MRI in addition to standard care; and to estimate five- and 10-year ipsilateral breast event rates in patients treated with wide local excision for DCIS after MRI, with low DCIS score treated without radiation and intermediate-high DCIS score treated with radiation.

Information obtained from the trial will serve as the foundation for a randomized Phase III trial evaluating the role of MRI in DCIS, similar to an ACRIN trial that is ongoing in patients with invasive breast cancer. Constance Lehman, MD, PhD, director of breast imaging, Massachusetts General Hospital, Boston, is the lead investigator.

Challenges and demands

Not surprisingly given the newness of radiology outcomes research, every initiative has had its challenges—challenges from which other institutions can learn. At the University of Pennsylvania Hospital and its sister institutions, a manual assessment conducted three months after Code Abdomen was introduced revealed that radiologists’ initial lack of familiarity with the codes was causing them to characterize too many patients as needing follow-up. Lesions that were too small to characterize were being coded as “O” or “3” because radiologists did not know how else to classify them.

“Eventually, we started discussing the issue with the radiologists, and they realized that most of this classification wasn’t appropriate,” Zafar says. “We knew that that was the situation based on the literature—except in instances where a patient had a known cancer.”

Radiologist buy-in was also an obstacle at first. Emails were used to remind clinicians to leverage the coding system, and the fact that using it was “the right thing to do” for the sake of enhanced patient care (and hence, enhanced value) was emphasized, Zafar says. Presenting the coding system as an antidote to future problems—such as malpractice suits—proved helpful as well.

Blackmore points to a tendency among some constituents to view radiology outcomes as existing in a separate silo from outcomes in general. The key, he believes, is recognizing that “it’s all part of a multi-disciplinary process, because patients don’t go to the hospital for good radiology outcomes; they go for good outcomes.”  To overcome the challenge, radiologists must increase their willingness to work with other clinicians to arrive at a consensus on imaging appropriateness, and clinicians must do the same.

While the type of buy-in described by Zafar and Blackmore is critical to the success of radiology outcomes research and initiatives, certain support and infrastructural elements must also be in place in order for these endeavors to succeed and move forward. On the support side, Zafar says, a chairperson or a champion is essential. Accordingly, she makes herself available to radiologists working within the University of Pennsylvania Health System to answer questions about coding and help to handle any problems pertaining to tracking follow-up of patients whose condition warrants vigilance.

“Protected time” for hospital clinicians to conduct imaging outcomes research, along with researcher resources, is also a must, Smith-Bindman has discovered. “It’s not something that can be done in fits and spurts; institutions need to understand that,” she observes.

As for infrastructure, Smith-Bindman points to the need to involve in individuals who possess considerable experience in epidemiological, biostatistics, and/or health services research—and preferably, all three. This, she contends, is the best way to ensure that data is collected in the appropriate manner, and that it is unbiased. It also increases the potential for obtaining meaningful feedback. “The skillset here goes beyond radiology,” she states.

The ability to robustly track data through EHRs and to combine data across different EHR systems to assess imaging outcomes is equally critical, Carlos claims. Under the umbrella of the ECOG-ACRIN study, researchers are looking for means of “pulling claims data from one (institution’s) system to another, so we can see, for example, whether Institution A is doing a better job than Institution B in managing downstream testing,” and to effect change accordingly.

High radiological stakes

Despite the fact that imaging outcomes research is, as sources put it, still in its infancy, it holds high stakes for radiology.  According to Blackmore, the more information on imaging outcomes that comes to light, the greater the specialty’s potential to retain its relevance.

“Radiologists must be involved in decision-making in regards to imaging use, and to ensure that imaging—and the right imaging studies—are used when there is a strong likelihood of benefit to the patient,” he asserts.  “Radiologists, with their unique training, have a depth of understanding that should be applied to assessing the appropriateness of imaging. Research and data on outcomes [lend credibility].”

Both Carlos and Smith-Bindman concur.  Carlos states that better outcomes mean better value, and the ability for radiology to provide better value in turn supports the ability to deliver the right care, for the right disease, to the right patient, at the right time.

Concludes Smith-Bindman: “Many times, there have been surveys asking physicians and patients what they know about the potential harms of imaging, be it radiation risk or false positives or over-diagnosis, and it’s been revealed that they know very little about these things. It’s time to improve that, and to apply it. Outcomes research—again, whether results are positive or negative—is the way, and as such, it will be key to radiology’s continued role in patient care.”

Editor’s Note: Further coverage of the University of Pennsylvania’s Code Abdomen project will appear in the December issue of Radiology Business Journal.