ABSTRACT
ABSTRACT Background: COVID-19 is a new disease and the most common complication is pneumonia. The Radiological Society of North America (RSNA) proposed an expert consensus for imaging classification for COVID-19 pneumonia. Objective: To evaluate sensitivity, specificity, accuracy, and reproducibility of chest CT standards in the beginning of the Brazilian COVID-19 outbreak. Methods: Cross-sectional study performed from March 1st to April 14th, 2020. Patients with suspected COVID-19 pneumonia submitted to RT-PCR test and chest computed tomography (CT) were included. Two thoracic radiologists blinded for RT-PCR and clinical and laboratory results classified every patient scan according to the RSNA expert consensus. A third thoracic radiologist also evaluated in case of discordance, and consensus was reached among the three radiologists. A typical appearance was considered a positive chest CT for COVID-19 pneumonia. Sensitivity, specificity, positive and negative predictive values were calculated. Cohen's kappa coefficient was used to evaluate intra- and inter-rater agreements. Results: A total of 159 patients were included (mean age 57.9 ± 18.0 years; 88 [55.3%] males): 86 (54.1%) COVID-19 and 73 (45.9%) non-COVID-19 patients. Eighty (50.3%) patients had a positive CT for COVID-19 pneumonia. Sensitivity and specificity of typical appearance were 88.3% (95%CI, 79.9-93.5) and 94.5% (95%CI, 86.7-97.8), respectively. Intra- and inter-rater agreement were assessed (Cohen's kappa = 0.924, P= 0.06; Cohen's kappa=0.772, P= 0.05, respectively). Conclusion: Chest CT categorical classification of COVID-19 findings is reproducible and demonstrates high level of agreement with clinical and RT-PCR diagnosis of COVID-19. In RT-PCR scarcity scenarios or in equivocal cases, it may be useful for attending physicians in the evaluation of suspected COVID-19 pneumonia patients attended at the emergency unit.