Chest radiograph classification and severity of suspected COVID-19 by different radiologist groups and attending clinicians: multi-reader, multi-case study

• DOI: 10.1007/s00330-022-09172-w
• Publication Date: 2022-10-25T06:03:05Z
• Authors: Arjun Nair; Alexander Procter; Steve Halligan; Thomas Parry; Asia Ahmed; Mark Duncan; Magali Taylor; Manil Chouhan; Trevor Gaunt; James Roberts; Niels van Vucht; Alan Campbell; Laura May Davis; Joseph Jacob; Rachel Hubbard; Shankar Kumar; Ammaarah Said; Xinhui Chan; Tim Cutfield; Akish Luintel; Michael Marks; Neil Stone; Sue Mallet
• Keywords: Chest radiograph; Neuroradiology; Interventional radiology; Guideline; 2019-20 coronavirus outbreak; Indeterminate

Abstract Objectives To quantify reader agreement for the British Society of Thoracic Imaging (BSTI) diagnostic and severity classification COVID-19 on chest radiographs (CXR), in particular an indeterminate CXR that could instigate CT imaging, from single paired images. Methods Twenty readers (four groups five individuals)—consultant (CCR), general consultant (GCR), specialist registrar (RSR) radiologists, infectious diseases clinicians (IDR)—assigned BSTI categories addition to modified Covid-Radiographic Assessment Lung Edema Score (Covid-RALES), 305 CXRs (129 paired; 2 time points) 176 guideline-defined patients. Percentage with a consensus two radiologists was calculated (1) categorisation those needing (indeterminate) versus did not (classic/probable, non-COVID-19); (2) severity; (3) change using scoring systems. Results Agreement category low across all (28–37%). other highest classic/probable three (66–76%) compared GCR (49%). normal similar (54–61%) but lower IDR (31%). severe (65%), (84–95%). For groups, changes modest. Conclusion is low, generally moderate change, suggesting test, rather than readers, limited utility both deciding disposition serial monitoring. Key Points • Across different varies widely. degree may render alone ineffective triage, especially cases. moderate, limiting guiding management.