Abstract
Background: This study aimed to evaluate the diagnostic performance of the coronavirus disease 2019 (COVID-19) imaging reporting and data system (CO-RADS) in admitted patients with suspected COVID-19 infection.
Methods: This retrospective study included all patients admitted to our hospital with COVID-19 pneumonia suspicion within March 20-May 15, 2020, who were examined by both computed tomography (CT) and real-time reverse transcription polymerase chain reaction (rRT-PCR) at initial presentation. Four radiologists, who were blinded to the rRT-PCR results and medical history, assessed all images independently and classified the CT findings according to the CO-RADS previously defined. Diagnostic value of the scoring system and interobserver agreement in rRT-PCR positive-negative groups and for CO-RADS 1-5 were evaluated.
Results: In this study, 274 (153 men and 121 women; 48.8±17.3 years) rRT-PCR positive and 437 (208 men and 229 women; 49.0±19.5 years) rRT-PCR negative individuals were included. It was found that CO-RADS had a good diagnostic performance with area under the receiver operating characteristic roc curve of 0.857. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were obtained at 81.9%, 89.4%, 75.7%, 92.5%, and 84.8%, respectively. The interobserver agreement of four radiologists in CO-RADS 1 and 5 was substantial to almost perfect according to the kappa values. Other CO-RADS scores showed a fair to moderate agreement. The interrater agreement was slightly higher in the PCR (-) patient group than in the positive one.
Conclusion: In conclusion, CO-RADS was a successful scoring system for distinguishing highly suspicious cases in terms of COVID-19 infection lung involvement, showing high interobserver agreement.