The mortality predicting ability of chest X-ray severity scoring systems in COVID-19 pneumonia

Introduction: Although several chest X-ray (CXR) severity scoring systems are in use to assess COVID-19 pneumonia (CP), inhomogeneity has been observed among the assessment methodologies. Objectives: To describe and validate severity scoring system based on different features to identify the most suitable scoring system to predict CP severity and outcome. Method: This retrospective study examined CXRs (n=147) of CP patients (n=85) to calculate severity scores using three scoring systems based on area infiltrated and the density patterns: A, A&D, and New. The best scoring system to predict the mortality was identified using the area under the curve (AUC) and linear regression analysis. Results: Regardless of the scoring system used, CXR severity has shown a good correlation to clinical CP severity (A: chi(2) =6.745; p=0.034; A&D: chi(2) =12.404; p=0.002; New: chi(2) =10.219; p =0.006). The mortality predictability of all scoring systems were satisfactory with high AUC ("A": AUC=0.685, sensitivity:67.4%, specificity: 54.5% at a cut-off point of 5/8; positive predictive value (PPV): 40.3%, negative predictive value (NPV): 78.6%"; A&D": AUC=0.748, sensitivity: 69.6%, specificity: 61.4% at a cutoff point of 7/16, PPV: 45.1%, NPV: 81.6%; "New": AUC=0.727; p <= 0.001, sensitivity:67.4%, specificity:68.3% at a cut-off point of 18/48, PPV: 49.2%, NPV: 82.1%). Additionally, the mortality predicting ability of the "New" scoring system was higher than the other two systems. Conclusion: COVID-19 pneumonia severity assessed with the CXR severity scoring systems correlated significantly with clinical severity and outcome. Overall, the "New" CXR severity scoring system is comparatively better at predicting the mortality of COVID-19 pneumonia.