Comparison of Predictive Ability of Epidemiological Factors, Inflammatory Biomarkers, and CT Severity Score for Mortality in COVID-19

Introduction: COVID-19 patients are categorized as per their clinical severity and their level of care is decided based on the clinical severity. Apart from clinical severity of patients, a need for robust predictors was also felt for early categorization and accurate prediction of final fatal outcome in hospitalized patients.

Material and method: In this retrospective observational cohort study all the adult patients admitted during November month were included. Available data for epidemiological factors, inflammatory biomarkers and CT severity score were collected and analyzed by univariate and multivariate logistic regression analysis to know predictive ability of each variable. A Receiver operating characteristic analysis was done to compare the predictive ability of each factor for final outcome of death.

Results: We analyzed records of 735 total patients. Most of them were male (72.38%), have a median (IQR) age of 60 years (50-69). Diabetes (42.85%), and hypertension (39.86%) were the most common co-morbidities. After univariate and multivariate regression analysis we could find that CRP, D-Dimer and CT severity score levels only can predict final outcome of death. During multivariate regression and receiver operative characteristic (ROC) analysis also, age and Charlson's co-morbidity index failed to predict in hospital mortality. CRP and D-Dimer on admission positively predicts final outcome of in hospital mortality with AUROC of 0.749(p=0.007, CI 0.61-0.88), and 0.864(p= 0.000, CI 0.74-0.99) respectively. Whereas, CT severity score had AUROC 0.73 (p= 0.014, CI 0.575-0.83). Cut off for CRP was 45 mg/L (Sn 0.8, Sp 0.56), D-dimer was 1000 µg/L (Sn:0.8, Sp: 0.9), and CT severity score was 15 (Sn 0.8, Sp 0.58).

Conclusion: CRP level of 45 mg/l, D-dimer level of 1000 µg/L and CT severity level of >15 at the time of admission can be added to conventional clinical severity algorithm to more accurately predicting final outcome and stratifying the level of care offered at the time of admission, and hence may improve odds off survival.