Chest CT for triage during COVID-19 on the emergency department: myth or truth?

Abstract

Purpose: We aimed to investigate the diagnostic performance of chest CT compared with first RT-PCR results in adult patients suspected of COVID-19 infection in an ED setting. We also constructed a predictive machine learning model based on chest CT and additional data to improve the diagnostic accuracy of chest CT.

Methods: This study's cohort consisted of 319 patients who underwent chest CT and RT-PCR testing at the ED. Patient characteristics, demographics, symptoms, vital signs, laboratory tests, and chest CT results (CO-RADS) were collected. With first RT-PCR as reference standard, the diagnostic performance of chest CT using the CO-RADS score was assessed. Additionally, a predictive machine learning model was constructed using logistic regression.

Results: Chest CT, with first RT-PCR as a reference, had a sensitivity, specificity, PPV, and NPV of 90.2%, 88.2%, 84.5%, and 92.7%, respectively. The prediction model with CO-RADS, ferritin, leucocyte count, CK, days of complaints, and diarrhea as predictors had a sensitivity, specificity, PPV, and NPV of 89.3%, 93.4%, 90.8%, and 92.3%, respectively.

Conclusion: Chest CT, using the CO-RADS scoring system, is a sensitive and specific method that can aid in the diagnosis of COVID-19, especially if RT-PCR tests are scarce during an outbreak. Combining a predictive machine learning model could further improve the accuracy of diagnostic chest CT for COVID-19. Further candidate predictors should be analyzed to improve our model. However, RT-PCR should remain the primary standard of testing as up to 9% of RT-PCR positive patients are not diagnosed by chest CT or our machine learning model.

Keywords: CO-RADS classification; COVID-19; Chest computed tomography; Emergency Department; Machine learning; Prediction model; Real-time reverse transcription polymerase chain reaction (RT-PCR).

Fig. 1

a CO-RADS 1: A few fibrotic bands in the lower lobes. No evidence of infection. RT-PCR−. b CO-RADS 2: Bronchial wall thickening, small centrilobular nodules, and tree in bud abnormalities in the left upper lobe. Consistent with bronchiolitis. RT-PCR−. c CO-RADS 3: Consolidation with surrounding ground glass opacity in right upper lobe. RT-PCR−. d CO-RADS 4: Bilateral areas of patchy ground glass opacity with associated small peribronchovascular consolidations. Predominantly central distribution. RT-PCR+. e CO-RADS 5: Bilateral peripheral ground glass abnormalities with areas of associated consolidation. RT-PCR+

Fig. 2

Flowchart of study with included and excluded patients

Fig. 3

ROC curve of CO-RADS score for COVID-19 diagnosis, taking the first RT-PCR result as a reference. AUC 0.914 (0.879–0.949)

Fig. 4

PCR results for each separate CO-RADS category

Fig. 5

ROC curves of CO-RADS alone (green line) and the prediction model (yellow line). Accuracy 10-fold cross validation is 0.91 ± 0.10. AUC for CO-RADS alone is 0.920 and 0.953 for the prediction model