To keep or not to keep? The question of whether or not to retain the markings produced by mammography computer-aided detection (CAD) software to highlight suspicious findings could have important medicolegal implications, according to Dr. Eliot Siegel.
If CAD results are discarded, it may be difficult to defend in court the fact that information used to help make a diagnosis is no longer available. And it may also not be sound medical practice to discard CAD markings, Siegel said during a Monday talk on medicolegal concerns at the New York Medical Imaging Informatics Symposium (NYMIIS) in New York City. Part 1 of our two-part series on his talk is available here.
Dr. Eliot Siegel from the University of Maryland.Siegel, who holds appointments at the University of Maryland and the Baltimore Veterans Affairs (VA) Medical Center, collaborated on his NYMIIS 2015 presentation with Dr. Jonathan Mezrich, JD, also of the University of Maryland.
How do radiologists view CAD?
Mammography CAD is by far the most utilized CAD application. But what do radiologists really think of it? Mezrich and Siegel conducted an online survey in 2013 in collaboration with the Society of Breast Imaging (SBI) and DiagnosticImaging.com to gather radiologist opinions on legal questions such as the following:
- To what extent has CAD become the standard of care in the subspecialty?
- If CAD is performed, is one then obligated to follow or biopsy CAD findings one finds questionable or would have otherwise ignored?
- Will a questionable finding not mentioned or dismissed by the radiologist but marked by CAD that ultimately developed into a malignancy be grounds for malpractice?
- To what extent do clinicians archive CAD markings, and if not, is there a worry that future, better versions of CAD might be used in the courtroom to show that findings were evident on CAD?
- If CAD markings are discarded, is this not a case of "spoliation" that should be determined in favor of an injured plaintiff?
Of the 47 respondents to the online survey, 89% said they always use CAD when reading screening mammograms; only 4% indicated they rarely or never used CAD, Siegel said.
Paradoxical use
Despite its widespread use, clinicians differed significantly on their perceptions of the value of CAD; only 2% said they always rely on CAD to provide an accurate diagnosis. Meanwhile, 49% reported they sometimes rely on CAD to make an accurate diagnosis, while 49% said they rarely or never do.
"It's an interesting paradox, at least superficially; you have 90% of mammographers who are using CAD, but 50% of mammographers essentially say that they rarely or never rely on CAD," he said. "So it's likely that the mismatch may be related to the reimbursement -- in many places it averages about $12 per study or about $1,000 per day if you're reading approximately 83 cases, or approximately $2,400 per day if you're reading 200 screening mammograms or so."
Most of the survey respondents had not changed an interpretation based on CAD results; only 2% indicated they had altered their opinion. As for the rest, 36% said they sometimes changed their interpretation based on CAD, and 61.7% said they rarely or never changed their interpretation based on CAD.
One of the reasons for the mismatch between the use of CAD and the lack of reliance on it by radiologists is its current approved use by the U.S. Food and Drug Administration (FDA) as a second-opinion tool, Siegel said.
"It would be really helpful if the computer would indicate its level of suspicion and would allow you to potentially be able to understand why a lesion was marked by CAD," he said. "Most mammographers say that the microcalcification finder is helpful but that the lesion finder is significantly less helpful."
In other survey findings, 15% of respondents said CAD was often helpful, 49% considered it to be sometimes helpful, and 36% said it was rarely or never helpful. However, when asked if they felt reviewing CAD analysis in conjunction with their own analysis represents the standard of care in breast imaging, 80% of mammographers said yes and only 20% disagreed.
"Yet when asked if a radiologist could not use CAD and still meet the standard of care, 87% of clinicians said yes, they could, and 13% disagreed with that," Siegel said. "This may reflect some ambiguity about the legal meaning of the standard of care and ... the common meaning of standard of care."
Archiving CAD marks
When clinicians were asked in the survey if they generally archive CAD output in their PACS along with the study, 23% said they do, while 72% said they rarely or never do. Four percent said they sometimes archive CAD output. After being asked if they worry that archived CAD results may lead to more lawsuits or increased liability, 81% of mammographers answered yes and 19% said no.
CAD is also still evolving and improving. The abnormalities identified by subsequent generations of CAD software in two years may far surpass the ability of today's algorithms, Siegel said.
"So if you go to court in the future on a case from today, it's possible that one might have a different opinion from the CAD software of the future," he said. "So one recommendation has been, if you are using CAD and you are saving the markings -- which we believe you should -- it's probably important to note what level of the software you used at that time or have that [information] be saved associated with the CAD markings."
A contrarian opinion
In response to an article published online on these topics by Mezrich and Siegel in the Journal of the American College of Radiology (June 2015, Vol. 12:6, pp. 572-274), Dr. Leonard Berlin, a well-respected radiologist and frequent author on legal topics in radiology, shared his views on the legal issues with CAD.
While acknowledging that it's true that many and perhaps even a majority of breast imagers use CAD when interpreting mammograms, Berlin said there is "no evidence to support their suggestion that the use of CAD has become -- or will become in the foreseeable future -- the legal or 'de-facto' standard of care."
Berlin went on to say in his JACR letter that it has never been established that CAD markings constitute a medical record, and there is no indication they ever will.
"The mammography report itself is, of course, a medical record; however, an informal consultation with a colleague-radiologist, or a CAD marking, that assists the radiologist in making his or her interpretation is not, and therefore neither is subject to the spoliation law," he said.
Because the legal definition of standard of care covers reasonable conduct made under the same or similar circumstances, it's inconceivable that any judge would allow an attorney or radiology expert's use of a future version of CAD software to rediagnose a mammogram that had been interpreted several years earlier, according to Berlin.
He said in the letter that he does not think all radiologists should use CAD, and he does not believe the use of CAD should become the standard of care. Furthermore, he suggested that CAD markings be discarded. The American College of Radiology (ACR) and SBI haven't issued any recommendations on this matter, he pointed out.
"It is not for Drs. Mezrich, Siegel, or Berlin to predict what the jury verdict in the above scenarios would be or to instruct radiologists as to whether they should archive or discard CAD markers," Berlin said. "The readers of this [letter] must draw their own conclusions and act accordingly."
Sound medical practice?
In their own reply to Berlin's response, Mezrich and Siegel noted that his position on archiving CAD markings was similar to that of many nonradiologist specialties: Why save image data that could come back to hurt you in court if your society hasn't recommended that the information be stored?
"But we believe actually that it would be difficult to defend Dr. Berlin's approach of not retaining or storing the CAD markings, and we think it's probably also not sound medical practice to do that," Siegel said. "The definition of the electronic medical record is rapidly evolving and is going to soon encompass not just radiology reports and images, but also other types of medical images such as those obtained by dermatologists, pathologists, cardiologists, etc."
Importantly, lawyers at his hospital and those consulted across the U.S. believe that outside images and any other data used to arrive at a diagnosis should be retained, Siegel said.
"If you used it in your clinical decision-making, then it's part of the electronic medical record," he said. "That's what we're increasingly hearing."
Mezrich and Siegel said they believe it would be erroneous and legally difficult to equate CAD markings with an informal consultation with a colleague; radiologists do not bill for informal consultations -- but they do bill for CAD. Using CAD is more akin to a formal consultation, with a bill sent to a patient for the services, Siegel added.
"Despite the fact that there is no specific recommendation from the ACR or Society of Breast Imaging [about these formal billable second opinions], we would expect that the radiologist obtaining a consultation formally would feel uncomfortable discarding the written consultation and that it probably wouldn't serve the patient's best interests," Siegel said.
Optimal patient care
Retaining CAD markings could also contribute to optimal patient care. When reviewing follow-up studies in the future, radiologists could, for example, review the CAD markings on the study and re-examine areas selected by the CAD software as suspicious.
"Also, retention of CAD markings would allow future technological development of CAD software that could use previous markings as a baseline for assessment of change, which would make the CAD [software] a more usable and powerful tool," Siegel said.
Mezrich and Siegel also recommended that ACR consider developing a guideline about whether or not mammography CAD markings should be saved and if it's considered best practice.
"But we're not convinced at this point that there really is compelling evidence to support not retaining and keeping CAD markings," Siegel said.
