Lung ultrasound based prediction of CT-scan Severity Score in COVID-19

Abstract

Background: CT-scan Severity Score (CT-SS) is the gold standard for the quantification of COVID-19 pneumonia, however CT-scan is not always available. Aims and objectives: Lung ultrasound (LU) is able to identify lung abnormalities, hence we hypothesize that can be used to predict CT-SS. Objectives are to determine whether it is possible to predict CT-SS from the LU score, and whether the change in LU score associates with a change in CT-SS during hospitalization. Methods: This is a retrospective observational study. Hospitalized patients with COVID-19 pneumonia who performed LU within 6 hours from CT-scan were included. Two LU scores, the LU-Mean, calculated by diving the sum of scores of explored chest areas for the total number of areas, and the LU-Sum, calculated as the sum of chest areas with a score ≥ 2, were derived and used to predict CT-SS using linear regression models. The agreement between fitted values and CT-SS was assessed using Bland-Altman plot. The correlation between the change in CT-SS and LU scores was reported using the Pearson correlation index. Results: The median CT-SS was 11 (IQR:6). LU-Mean and LU-Sum were linearly correlated with CT-SS (r_LU-Mean=0.78 and r_LU-Sum=0.79), with a Beta of 7.34 (P-value<0.001) and 0.94 (P-value<0.001), respectively. Two predictive models, based on LU scores and type of respiratory support, were developed, with an adjusted R-squared of 0.64 and 0.67, respectively. The correlation between the change of CT-SS and LU scores was 0.86 (P-value<0.001) for LU-Mean and 0.87 (P-value<0.001) for LU-Sum. Conclusions: CT-SS can be predicted from LU scores, and its change correlates with that of LU score. LU score can be used to predict CT-SS when CT-scan is not available.