Standardization: An Answer to Three of Radiology’s Vexing Problems

In healthcare, the bar for quality is being set ever higher. One key approach to attaining quality is standardization: Just as other industries have embraced Lean, Six Sigma, and the Toyota Production System, each of which leverages standardization as a tool to achieve quality improvement, imaging service providers are establishing standards for multiple aspects of imaging services delivery in the quest to enhance patient care as well as safety.

Consider the example of the University of Florida College of Medicine in Gainesville, whose primary teaching hospital, Shands at the University of Florida (UF Shands), now utilizes standardized names for each imaging procedure, an initiative referred to as Standardized Nomenclature for Imaging Procedures (SNIPs), along with standardized reports. SNIPs serve to define consistent, predictable protocols, as well as to facilitate the assessment and comparison of clinical outcomes. They indicate modalities and anatomy; contain specific indications for imaging protocols; and are applied in ordering, requesting authorization for, scheduling, performing, interpreting, and billing studies. In ordering studies, clinicians choose a particular SNIP, which links to the appropriate exam and report template.

Anthony A. Mancuso, MD, professor and radiology department chair, says that the absence of such protocols generally leads to discrepancies between the clinical indicators for performing a given study and the caliber/relevance of the information it yields, in turn compromising quality as well as patient safety. “The construct of the protocol must be targeted to the specific clinical indications, or scenario,” Mancuso asserts. “Otherwise, there is no assurance that the exact answers being sought in an exam will show up in the report, and there is the risk of missing important findings (as well as of doing harm to the patient) because of the mismatch.”

Careful consideration

Mancuso emphasizes that properly formulating the SNIPS necessitated a meticulous approach by a team of colleagues. Notably, the team conducted a careful assessment of whether individual clinical indications should be classified under a single protocol, or under multiple protocols. “It’s not like there is a SNIP for every indication, and some indications do fit under one protocol,” Mancuso says. “However, the complexity was such that we did end up, for example, with more than 20 head and neck SNIPS, 30 SNIPS for the brain, and 10 to 15 SNIPS for the spine.”

Meanwhile, the standardization of reports—called structured reports at UF Shands—is an ongoing cooperative endeavor involving clinicians, radiologists, and, in some instances, specialists from several other departments, among them cardiology, surgery, neurosurgery, and otolarangology. To ensure that the purpose of standardization—specifically, quality attainment—is indeed fulfilled, clinicians have been and continue to be asked for specific feedback “about what they want to see in and know from individual reports,” Mancuso states. For instance, some clinicians were queried as to details to be incorporated in a structured report of a pancreatic cancer CT protocol.

Pushing the quality and safety envelope a few steps further, different series of questions, some developed based on input from radiologists alone and others, in tandem with additional suggestions from specialists outside the discipline, are also a component of each structured report. Dubbed “forced-choice observations,” these questions, which radiologists must answer, mesh tightly with individual protocols. Case in point: Some questions posed in a report of a children’s seizure protocol pertain to the normalcy of hippocampal formation and current hippocampal volume, as well to whether there is evidence of improper/abnormal brain development and/or long-term effects of seizures on the patient’s brain.

“Without this type of structured information, it is difficult for the clinician to determine if the radiologist who completed the report possesses knowledge in that domain and whether the data” are of real value, Mancuso states. “That all ties in to patient care.”

Yet another element of the standardized reports is a report acuity scale on which radiologists must rate the urgency with which referring physicians should review the findings. Ratings range from one to five, with a rating of one indicative of normal findings and a rating of two, indicative of the presence of a finding that merits review by the clinician, but may not necessarily warrant other action. A rating of three indicates the presence of actionable and non-actionable findings that should be reviewed within 24 hours if at all possible, while a rating of four or five means, respectively, that the findings must  be addressed within an eight-hour window from the time the report was received or immediately.

Mancuso concedes that there has been some resistance among UF Shands’ radiologists to the standardization of imaging protocols and reports alike; most or all of these physicians, he says, have pet imaging protocols and their own, personalized reporting style. On the imaging protocols side, the majority of complaints received concern the additional time and effort needed to ensure adherence.

But no exceptions to the rules are permitted. “Quite simply,” Mancuso notes, “we tell them this is a non-negotiable, because we owe it to our patients. They understand the logic in that argument.” To bolster efficiencies and soften radiologists’ resistant edge, the protocoling tool resides inside UF Health Shands’ RIS system. Toward the same end, each SNIP has an identifying number within the EHR, which in turn is linked to a CPT code; the latter is linked to a procedure code.

Meanwhile, radiologists’ objections to the reporting protocols center on a reluctance to deviate from preferred and, in many cases, long-ingrained approaches to preparation. Such objections are countered with the argument that reports generated in accordance with protocols are of higher and more consistent quality and hence, improve safety and caliber of patient care.

“However, it goes beyond this,” asserts Mancuso. “We continue to emphasize to our radiologists not only that standardization in reporting is what our clinicians want, but also that at some time, radiology will be held accountable for providing value —which comes partially from reporting consistency—and we need to be ready. Not to mention, again, that it’s just the right thing to do for our patients.”

To date, UF Shands has not attempted to quantify improvements resulting from having standardized either its imaging protocols or its reports. Nonetheless, Mancuso says, anecdotal evidence points to positive change. “Clinicians tell us they are getting what they want in terms of information,” he observes. “And where we previously received complaints” about reports pretty much daily, “we have not heard one in the past six months. Most importantly, we are doing what is right for our patients. ”

Not so incidental

Radiologists’ approach to the reporting and handling of incidental findings, known colloquially as incidentalomas, varies almost as much, if not as much, as their approach to the reporting of clinical findings in general. Again, quality implications come into play.

“Show a set of images with an indeterminate finding to several radiologists, and be prepared to hear multiple recommended courses of action,” states James A. Brink, MD, radiologist-in-chief at Massachusetts General Hospital (MGH) in Boston and the Juan M. Taveres Professor of Radiology at Harvard Medical School. “One might suggest follow-up scans at six-month intervals; another, more frequent follow-up scans. A third might advocate leaving things entirely alone, and another will say more invasive procedures need to be performed immediately. Under-recommend, and you may be denying the patient the care he or she really needs.”

On the other hand, over-recommend, and you risk subjecting the patient to stress and possible complications for a procedure that wasn’t needed in the first place, Brink says. “There also can be misunderstandings about the verbiage in reports.  And even if there are written guidelines, it isn’t always reasonable or feasible for radiologists to stop to find them or look them up. All of this can affect [the caliber of patient care].”

In 2009, Brink says, a study conducted at MGH revealed a 300% increase in the number of patients for whom follow-up imaging examinations had been recommended over the previous 15 years. A second internal study conducted last year showed that 50% of recommendations for patients in whom lung nodules had been detected via abdominal CT included errors in either under- or overutilization of follow-up CT procedures.

Algorithms for incidentalomas

To address such concerns and trends, a decision was made to devise standard algorithms for application by MGH’s radiologists in recommending a course of action for patients with pulmonary and adrenal nodules, and to incorporate these algorithms into a proprietary decision-support tool. Additional algorithms for handling and reporting on renal and craniovascular incidentalomas, among others, are currently being formulated. (So, too, are standardized diagnostic algorithms; for instance, a grading scale for lacerations of the spleen).

The decision-support tool runs on each of the radiology department’s workstations and is integrated with its speech-recognition software for ease of access and use. Radiologists locate the algorithms by clicking on the appropriate corresponding icon (pulmonary nodule or adrenal nodule). From there, they input into the system information about the nodule’s history, morphology, size, and growth pattern. Recommendations for follow-up are then generated; these, along with the historic, morphologic, and size data, are automatically entered into the radiology report.

Brink says an MGH team chose to introduce algorithms for addressing pulmonary and adrenal nodules, rather than for handling other types of incidentalomas, based on the belief that doing so would be the “most impactful, most do-able move.”  Both algorithms are drawn in part from content found within the American College of Radiology’s series of white papers on incidental findings; the pulmonary nodule algorithm also incorporates Fleischner Society recommendations for the management of small pulmonary nodules.

Future algorithms will take into account ACR principles, but will be derived from other sources where applicable. As an example of the latter, Brink cites an algorithm for determining follow-up protocols for liver lesions identified during MRI examinations of the breast; “the ACR’s principles apply to liver lesions found during dedicated imaging” of that particular organ, and a broader perspective is needed, he explains.

Several strategic steps have helped to foster stakeholder acceptance of the decision-support tool and algorithms. Notably, the latter were designed not only to be consensus-based and modifiable, but also  to reflect input from MGH’s radiology subspecialty chiefs, practicing radiologists, and referring physicians. “If you don’t build a standardized model that everyone trusts and feels comfortable with, you haven’t gained much in the way of value,” Brink says.

Although it is still in the “very early days,” he continues, the standardization initiative already appears to be a success. Adherence to radiology department guidelines for recommendations of follow-up imaging for pulmonary nodules now stands at 65%, up from 50% before the algorithms and decision-support tool made their appearance. In a recent internal study, concordance with the department’s guidelines increased by 96% in the 40% of cases in which the combination was used. Efforts to incent radiologists to use the two existing algorithms, and to leverage new ones as they emerge, are now underway.

Across a multi-site system

Imaging service providers like MGH and UF Shands clearly have much to gain from standardizing image acquisition and reporting protocols. The patient care benefits of radiation dose optimization also are significant and recognized beyond the walls of the radiology department, with additional advantages that bubble to the surface when dose metrics are standardized across health care providers’ systems rather than within individual hospitals alone.

Such is one of the major premises behind the University of California Dose Optimization and Standardization Endeavor (UC DOSE), undertaken by the University of California Health System with funding from the University of California’s Center for Healthcare Quality and Innovation.

Led by Rebecca Smith-Bindman, MD, professor of radiology and biomedical imaging, University of California, San Francisco (UC-SF), the project is aimed at standardizing and optimizing CT protocols across all five of the system’s academic medical centers: UC Davis Medical Center, UC Irvine Medical Center, UCLA Medical Center (Los Angeles), UC San Diego Medical Center, and UC San Francisco Medical Center.

Compliance with California State Bill 1237, which took effect on July 1, 2012 and requires the reporting of CT radiation doses in patients’ radiology reports, ranked among the health system’s rationales for moving UC DOSE forward. The law, however, was not the sole impetus for doing so, asserts J. Anthony Seibert, PhD, professor of radiology and assistant vice chair of radiology informatics, University of California, Davis (UC-Davis), who serves as that insitution’s co-principal investigator for the project at UC-Davis.

Says Seibert: “Sensationalized, high-profile incidents involving excessive radiation exposure—like the one at Cedars-Sinai Medical Center in Los Angeles in 2009—brought the issue of safety to the forefront. But it was also clear to us that standardization and optimization could break down barriers between our institutions. CT doses by protocol and indication are highly variable, both within and across institutions. When protocols are standardized, multiple hospitals within a system can be on the same page in imaging. Cross-system imaging services—for example, late-night reads for one institution by another—can be handled consistently, and care can be delivered at a lower cost.”

Under the project umbrella, a collaborative working group of radiologists, physicists, and CT technologists from all hospitals within the system authored a guidance document on dose reporting, and dose reporting procedures were automated via the implementation of a packaged software solution. Infrastructure (dose reporting software and database) and methods of capturing CT dose metric information—CTDIvol, dose-length product (DLP) and other dose-related indicators—for each type of CT examination across all UC medical centers were standardized and entered into a master database.

Additionally, CT acquisition protocols were optimized to achieve lower dose metric values while still providing acceptable image information and image quality. The latter was achieved by calling, for example, for CT tube current modulation, the application of iterative reconstruction algorithms, the elimination of unnecessary image acquisitions (like pre- and post contrast as opposed to one or the other) or the use of multi-phasic studies with one or more fewer phases.

“From the database, the exam-specific CT dose metric distributions provide average and percentile values, which are used to set local diagnostic reference levels and to identify outlier exams,” Seibert explains. He points out that such cross-hospital standardization paves the way for inter-institutional comparisons. These comparisons, in turn, bring to light opportunities for improvement that involve modifying and optimizing CT acquisition protocols to achieve low radiation dose without sacrificing the image quality deemed necessary for proper diagnosis. Deviations from the standards are reviewed, with areas of improvement sought in cases where they are deemed unjustifiable.

Getting buy-in

Concerted efforts were made to get radiologists on board with the project, as well as to ease implementation and encourage compliance. “Soliciting their input at every stage was and is crucial,” Seibert says. “Medical physicists understand what is being administered to patients, but radiologists’ involvement is critical because they are the ones who will be reading the exams. They need to see that their perspective is important.”

A system-wide meeting held last year offered educational sessions on dose reduction, along with opportunities for attendees to upload their protocols and benchmark them against those shared by others. Earlier this year, Smith-Bindman organized a two-day retreat to bring together radiology decision-makers from all UC medical centers. The goal of the meeting was to arrive at a consensus concerning the most common ways to image chest, abdomen/pelvis, and head, and to discuss low-dose protocols for lung and colon screening.

To spark discussions and encourage participation, radiologists from the various sites were asked to share CT images from their current clinical practice. Overall goals were discussed as a group, with additional breakout sessions focusing on neurologic, chest, and body imaging as needed to reach consensus. Both events helped to foster trust by allowing participants to see the data showing variation in radiation doses and then discuss low-dose protocols.

“Sharing is a powerful motivator (for compliance),” states John Boone, PhD, FAAPM, FSBI, FACR, professor of radiology and biomedical imaging, vice chair of radiology research, and co-principal UC-DOSE project investigator, UC-Davis. “It’s an entirely different scenario when you find out that your protocol has several times the dosage of someone else’s. You think that maybe your personal preference isn’t optimal, after all. What we’ve found is that you can discuss metrics and the whole dosing issue on the phone until the cows come home, but unless it is also done face-to-face, things don’t stick as well as they should.”

Moreover, a meeting of various radiology section chiefs from within the University of California Health System and principal investigators from each of its five sites to discuss dose metrics and the results of the project to date was held this past April. A follow-up meeting is scheduled for October of 2014. A virtual symposium, “The UCSF Symposium on Radiation Safety and Computed Tomography,” is available to radiologists, as well as to physicists, CT technicians, and the general public, at http://rorl.radiology.ucsf.edu/symposium.

Next up for UC DOSE participants is the standardization of CT scanning protocols, a task Siebert and Boone deem challenging because of the variety of scanners deployed across the system as well as because of variances in procedure nomenclature. “What is called a ‘head scan’ at UC-Davis may be called a ‘routine brain scan’ at one of our other hospitals,” Seibert notes. “There are different doses for different nomenclatures. Standardizing here—and matching nomenclatures with the intent of a protocol—will result in much better comparisons of procedures, too.”

Anecdotal evidence points squarely to system-wide radiation dose reduction, without “any untoward loss of image quality” and with a higher standard of safety for patients, Siebert and Boone say. Feedback from radiologists indicates that the caliber of CT images produced is more than appropriate for their needs. Identification of trends—for example, variations in DLP levels—has helped to reduce dosing discrepancies.

“We’ve sliced into the pie,” Seibert concludes. “Maybe it’s a small slice—but it is a very important one for the safety and quality of imaging going forward.”