MARCH 20, 2020 — The coronavirus pandemic will strain many healthcare sectors, including radiology. Radiologists have been conscripted to the front line because COVID-19 has signatures on chest CT befitting of viruses that damage lungs. But CT can be normal in early illness, and after each potentially infected patient is scanned, the machine must be completely disinfected. Therefore, CT isn’t recommended to screen for COVID-19.
Of course, CT will still be used in patients with acute respiratory symptoms, some of whom may have coronavirus infection. How should radiologists report findings suggestive of COVID-19 in patients imaged for other conditions? The answer isn’t straightforward and needs careful thought.
When present, the findings of COVID-19 on CT — notably peripheral ground-glass opacities — are sensitive but not specific for coronavirus; other pneumonias resemble COVID-19, particularly viral and Pneumocystis jirovecii pneumonia, cryptogenic organizing pneumonia, and acute lung injury from drug toxicity, hypersensitivity, and autoimmune diseases, to name a few pathologies. This means that false-positive errors don’t occur so much from falsely labeling healthy people with COVID-19 infection but rather from falsely attributing COVID-19 in ill patients with other acute respiratory pathologies — ie, misattribution.
Radiologists thus face a familiar dilemma, choosing between overcalling or undercalling, and both errors are costly. If radiologists omit COVID-19 infection in their reports when they see suggestive findings, and patients are actually infected, they won’t be appropriately isolated and could infect others. If radiologists call COVID-19 infection when they see suggestive findings, and patients aren’t infected, wrong protocols will be activated and they may not be treated for the condition they actually have, not to mention that the CT scanner will be unnecessarily nonoperational until decontaminated.
Furthermore, with constrained resources, attention on patients who don’t have coronavirus will divert attention from those who do.
One approach is for radiologists to report what they see and let clinicians decide how to use that information. The problem is that radiologists’ comments will influence how clinicians think and act; similarly, how clinicians think and what they might do with the information affects what radiologists say.
Image interpretation is not an island; it’s a complex archipelago.
Many radiologists throw responsibility back to clinicians with disclaimers such as “clinically correlate” or “pneumonia can’t be excluded,” a universally unhelpful practice that should unequivocally be abandoned during this pandemic.
A few steps may reduce undercalling and overcalling. First, radiologists should be familiar with the spectrum of findings of COVID-19 on chest CT and also recognize its most characteristic findings. Second, COVID-19 should only be mentioned after radiologists speak with clinicians and all agree that coronavirus infection is possible. Radiologists should express their confidence of COVID-19 infection on chest CT and grade their confidence as low, intermediate, or high. Clinicians should express the pretest probability of COVID-19 infection. Combining information from radiologists and clinicians will improve CT’s accuracy.
A joint effort prevents the burden of diagnosis falling on one side. The diagnosis of coronavirus should be confirmed with RT-PCR, though precautions should commence, including disinfecting the CT.
Some patients at higher risk for mortality from COVID-19 also are likely to have acute respiratory pathology that resembles COVID-19 on chest CT, such as those with chronic heart or lung disease, the elderly, oncology patients, posttransplant patients, and others with immunosuppression. Diagnosis is hard in these groups because misattribution in either direction is harmful.
We recommend a triangulation approach here. A second radiologist should look at the images and answer two questions: How characteristic are the CT findings of non-coronavirus respiratory pathology? How likely are the clinical features of non-coronavirus respiratory pathology?
Radiologists are better at answering “Could it be COVID-19?” than “Is it COVID-19?” A consensus, multidisciplinary approach will give us an idea of the pretest probability of COVID-19, conditional probability of CT, and posttest probability of COVID-19; such numbers aren’t reliably available for a new disease with unknown and changing prevalence.
We discourage using CT to rule out multiple pathologies at once, such as pulmonary embolus (PE), dissection, and COVID-19 infection. Such quests may increase false negatives and false positives because radiologists, when looking for multiple pathologies without knowing which is more likely, can miss what’s important and amplify what isn’t.
Furthermore, PE CTs are often obtained at the end of tidal breathing, which increases lung density that mimics diffuse ground glass opacities that can be mistaken for COVID-19 infection.
CT is often used to exclude a second acute pathology in patients with known acute respiratory disease, such as superimposed pneumonia in patients with acute pulmonary edema. We strongly discourage using chest CT to exclude “superimposed COVID-19” in patients with other acute respiratory pathologies such as pulmonary edema, because of their resemblance. The quest to exclude “superimposed COVID-19” is forlorn because the answer will always be, “Yes, superimposed COVID-19 infection is possible.”
Positive COVID-19 CT cases should be collected in a central databank to develop algorithms, using machine learning, to improve CT’s specificity and to reduce misattribution.
Although CT should be used judiciously, its use may increase if diagnostic uncertainty increases during the pandemic. The demand on resources from CT isn’t just the scan itself but downstream follow-up of incidental findings (incidentalomas) such as thyroid nodules, which are overwhelmingly likely to be harmless. In times of coronavirus, pursuit of incidentalomas could divert resources from higher-impact endeavors.
To reduce the incidentaloma burden, radiologists should limit what’s seen and see only what’s clinically most relevant. In the COVID-19 chest CT protocol, the field-of-view should be restricted to avoid the thyroid and adrenals, and thick slices should be created to avoid small nodules. The changes will reduce both the number of images radiologists view — typically 1000 per study — and the radiation exposure.
Our culture of seeking abnormalities with all our visual might, summed up by the ethos to “find, measure, and document anything and everything, no matter how small or clinically significant,” should temporarily change. This will require a deliberate effort because our search pattern has evolved to seeking trees rather than seeing forests. Amnesty from litigation during the pandemic for missed incidentalomas that cause future harm will encourage radiologists to focus on what will harm the patient in the next 20 hours rather than what might harm the patient in 20 years. Professional organizations should say in no uncertain terms that the search for, and documentation of, incidentalomas is ill advised during the pandemic.
Other habits that should be suspended include ordering daily portable chest radiographs in the intensive care unit in stable intubated patients, in-patient malignancy workups, and frequent surveillance of aneurysms and cancer. Some would argue that these practices should be discontinued permanently, but that diminishes the significance of the moment, so we ask that they be suspended only until the pandemic ends.
Thrift can boost capacity. It will take a village to achieve thrift.
Saurabh Jha, MBBS, MRCS, Associate Professor of Radiology, University of Pennsylvania; Staff Physician, Department of Radiology, Hospital of the University of Pennsylvania, Philadelphia, Pennsylvania
Scott A. Simpson, DO, MS, Assistant Professor of Clinical Radiology; Associate Program Director, Department of Radiology, University of Pennsylvania, Philadelphia
©2020 WebMD, LLC. All Rights Reserved.