As one of three presenters during a refresher course on “Communication in Cancer,” Fergus Coakley, MD, professor and chair of radiology, Oregon Health and Science University (OHSU), Portland, made a strong value case for deep subspecialization in radiology. Not incidentally, the model described came with a proven business case.
He began with his description of a subspecialist, acknowledging that not all subspecialists will possess all characteristics: fellowship trained; generally working in an academic medical center because those are the departments that are big enough to support deep subspecialization; and with a few years of post-fellowship experience to acquire the clinical experience to understand how their reports and findings correlate with the clinical management of a patient.
Another important ingredient that also is acquired over time in the job is the local knowledge to understand, for instance, how a surgeon approaches his or her cases and what they want in their reports. “That is the point where you really start to add value to patient management,” Coakley said. “This is a key feature of health care as we look to the future.”
Exhibit A in Coakley’s case was a male patient with a biliary stent to relieve jaundice and a diagnosis of pancreatic cancer who was referred by a pancreatic surgeon in preparation for a Whipple procedure. Noting no appreciable mass in the pancreatic head and looking more closely, the radiologist recognized the typical halo sign—the encapsulated appearance of the pancreas—that is typical of autoimmune pancreatitis. The call was made, the patient was managed with steroids, and the Whipple averted.
“These are the kind of cases—they don’t come every day—but these are where you are really adding value as a radiologist. Prevent a big, morbid surgery, the system has been saved the cost of that surgery, which is enormous, and this is where you are starting to be helpful as a subspecialist.”
Coakley referenced several studies from different sites showing consistently between 5% and 10% of patients going to Whipple for pancreatic cancer will end up with a benign diagnosis. The commonest diagnosis among those patients is simple, chronic pancreatitis and Coakley acknowledged that surgeons would argue that those patients are still benefitting from the Whipple because they have an inflammatory stricture in the pancreatic head, causing jaundice.
The second most common cause of a Whipple with no cancer found in the specimen is autoimmune pancreatitis, and the radiologist is often the key person to make or suggest that diagnosis. “I was reading a follow-up scan, and I went back to look at this baseline scan,” Coakley recalled. “It was striking to me that for somebody who had pancreatic head cancer there was really little to no pancreatic focal dilatation, little to no biliary dilatation. The mass itself had some odd features, you see some tubes and some circles and cysty-looking things but none at all typical of true pancreatic adenocarcinoma.”
He asked the pathologist to go back and look at the specimen again, and they reinterpreted the specimen as benign. “This is the only patient I’ve ever ‘cured’ of pancreatic cancer,” Coakley noted, “but it just shows that pathology is not always the gold standard. I think we often accept pathology as the ultimate and it really isn’t. They have the same issues of inter-observer variability that we do, and often there isn’t anyone checking them in the background.”
Coakley provided further evidence with the case of a 68-year-old woman who had unergone two cycles of chemotherapy for hepatic metastases that were apparently from occult primary pancreatic carcinoma, diagnosis based on an outside PET scan that showed a hot spot in the pancreas. Coakley was unable to get a copy of the original study, but the pancreas looked normal when the patient was re-imaged.
“What was most striking was that the liver, over the initial cycles of the chemotherapy, had developed appearance of pseudo-cirrhosis, which in my experience is nearly exclusively seen in metastatic breast cancer not pancreatic cancer,” he said. He also thought he saw an abnormality in the breast.
The somewhat exasperated medical oncologist ordered a mammogram on Coakley’s advice; upon biopsy, the patient was found to have metastatic breast cancer. “These are cases where the radiologist, using your radiological knowledge, can really impact on the diagnosis and change patient outcome, prognosis, and management,” he said.
Data suggest that pancreatic cancer misdiagnoses, in particular, may be relatively common. Researchers have looked at patients who are still alive five years out from a pancreatic cancer diagnosis, and about 10% are found to not have cancer, he said. This may account for the discrepancy between the American Cancer Society’s five-year survival rates for pancreatic cancer (6%) and Memorial Sloan Kettering Cancer Center’s (MSKCC) 1.8% five-year-survival rate for pancreatic cancer, which Coakley said probably represents a more accurate picture of survival rates for pancreatic cancer patients.
“These are probably patients with neuroendocrine cancers and other malignancies and inflammatory disease that has been misdiagnosed,” Coakley said, referring to the higher survival rates from the ASC. “You could argue that we should require imaging and pathology concordancy before making a diagnosis of pancreatic cancer much as we do in mammography.
Coakley went on to describe a patient referred for presumed ovarian cancer that turned out to be metastatic breast cancer; a 54-year old woman thought to have primary ovarian cancer who had metastatic colon cancer.
“When I look at a case of suspected ovarian cancer, one of the questions I ask myself is this: Is it really primary ovarian cancer?” Coakley said. “Like the literature on the benign diagnosis of post-Whipple patients, studies have pretty consistently shown that of patients who go to surgery for primary ovarian cancer, somewhere on the order of about 5% or so will have a diagnosis of a other, non-gynecological malignancy with metastatic disease to the ovaries.”
A Further Word on Value
With value defined as outcome or quality divided by cost, it is incumbent on radiology to justify the cost pf imaging, Coakley said. “You need to be doing something above the line to improve the quality, or outcome, of care to the patient, because that is where you start to add value and justify the cost of imaging, which is, of course, substantial.”
Issuing a report that is merely descriptive is not enough, Coakley said. “In order to add value, you do need to integrate the imaging findings with both the clinical and the scientific data that you as a radiologist and subspecialist hopefully know,” he said. “It is that clinical knowledge and that radiological knowledge—knowing the patient’s prior history, prior imaging—and it’s only when you integrate that together that you truly are going to add value, because otherwise you are just a monkey describing the report.”
Measuring value in a large, complex health system is very difficult to do, but a number of studies have attempted to measure the value of subspecialist reads both in oncologic imaging and in other radiologic domains, he said.
- A study out of MSKCC looked at the detection of extracapsular extension in prostate cancer patients and compared subspecialist reads to outside reads, using pathology as the gold standard and the subspecialists performed significantly better.
- A similar study out of MSKCC looked at subspecialty reads on 469 gynecological malignancy exams and the subspecialist opinion changed treatment in about 20% of cases, and in those cases in which there was some kind of proof of the opinion, they were correct 83% of the time.
- Another MSKCC study looked at pediatric studies and found major disagreements between subspecialists and generalists about 20% of the time; subspecialists were correct in about 90% of the cases where there was some standard of proof.
- A study from researchers at Johns Hopkins found major discrepancies between subspecialists and outside readers almost 10% of the time, with subspecialists correct 84% of the time.
Coakley pointed to a consistent theme of major changes in patient management 10% to 20% of the time when subspecialists provide second reads, with the subspecialists correct about 80% to 90% of the time. “Nobody’s perfect, obviously, and sometimes the subspecialists are wrong,” he said, “but consistently you can expect an actionable change in about 10% to 20% of those reinterpretations.”
While many major medical centers offer second opinion services, few have forbidden curbside reads as Johns Hopkins has done. “This was actually fairly draconian,” Coakley said. “They started this in their neuroradiology section, they actually said they would do no more curbside reads, and if you wanted a second opinion, you had to do that as a formal reinterpretation.”
Initial buy-in from their major clinical referrers was obtained before the service was rolled out. It grew to 40 or 50 studies per day, the reimbursement experience was favorable, and the program was expanded department wide. “They are reading about 100 studies a day, which is probably about two FTEs, a significant impact to the department,” he noted.
About a year-and-a-half ago, OHSU promoted second reads an as optional additional service, although it didn’t prohibit curbsides. “We mainly did this at our multidisciplinary tumor board, and just made referring clinicians aware that we were willing to look at and reinterpret outside studies,” he said.
OHSU’s collection rate on these studies has been about 40%, close to what the practice collects on baseline reads. It has grown over time, with body imaging accounting for the greatest number of reinterpretations, followed by neuroradiology—which Coakley expected would have been the leading section—MSK, and mammography.
In an analysis of CMS claims, focusing on studies with the 26 and 77 second-read modifiers, Coakley found that second reads accounted for just .1% of all claims between 1999 to 2012, and the denial rate is not much different that for primary reads. While the trend is upward, they still account for less than 1% of all CT reads. “I guess CMS has recognized, based on the data I showed you, the value of these reads,” he said.
Concerns about second reads include exposure to litigation, but Coakley has not come across a malpractice claim involving either a reinterpretation or a curbside read. There also are IT issues, such as generating an accession number, and managing expectations.
“Of course inevitably what happens is when you provide a service during the day time, then people expect it to be 24/7,” Coakley observed. “Our trauma service expects us to read far more studies from outside institutions overnight, and that is not something we are set up to do, at least for the formal reinterpretations. They can get the informal curbside opinion, of course, as they always could.”
A bigger, ethical concern is that of adding cost to the system. “If you are changing patient management 10% to 20% of the time, then you are potentially driving patients’ more cost-effective treatments,” Coakley said. “You don’t have to spot too many cases of auto-immune pancreatitis to pay for a lot of second reads and reinterpretations.”
In conclusion, Coakley said: “From a purely practical perspective, providing a formal second opinion service not only codifies or documents that clinical impact, but it can also provide reasonable compensation for your services and recognize the value that you are hopefully providing.”