Interobserver Agreement in the Analysis of Different Radiological Classifications of COVID-19 on Computed Tomography

Abstract

Introduction: Computed tomography (CT) has approximately 98% sensitivity for Coronavirus disease-2019 (COVID-19). Various algorithms were designed using CT images. However, the interobserver agreement of different radiological classifications of COVID-19 is not yet known. Thus, this study aimed to investigate the interobserver agreement of different radiological classifications of COVID-19. Materials and Methods: This study included 212 patients who were positive on the polymerase chain reaction test and eligible for CT. Four radiologists examined all CT images simultaneously. They reached a consensus that CT images can provide definite findings of COVID-19. The Radiological Society of North America (RSNA) consensus statement, the British Society of Thoracic Imaging (BSTI) structured reporting statement, and COVID-19 Reporting and Data System (CO-RADS) were used. Fleiss’ Kappa was used to detect interobserver agreement. Kappa values of 0.00-0.20 were considered as slight, 0.21-0.40 as fair, 0.41-0.60 as moderate, 0.61-0.80 as substantial, and 0.81-1.00 as near-perfect agreement, and p<0.05 was accepted as significant. Results: A total of 137 patients did not have any pathological CT findings. The most prevalent radiological findings were ground-glass opacities and consolidations. The agreements on all classifications were at near-perfect levels: RSNA, 0.86 (0.82-0.90); BSTI, 0.83 (0.79-0.87), and CO-RADS, 0.82 (0.79-0.86). The RSNA classification has the highest consistency rate, followed by BSTI and CO-RADS. However, substantial and moderate agreements were found in the subcategories of each classification. Conclusion: In this study, some subcategories had a lower agreement, despite the high consistency rates for COVID-19 radiological classification systems in the literature. Therefore, improving the items without consensus can lead to the development of better radiological diagnostic approaches.

Keywords

• covid-19
• diagnosis
• tomography
• interobserver variability