Exploring and optimizing models to predict the risk of death during hospitalization (within 30 days) in elderly patients with COVID-19

Abstract

Abstract To explore new models on the basis of the CURB-65 score for predicting the risk of death in elderly patients with COVID-19 during their hospitalization and estimate their prediction effectiveness. A retrospective analysis of the clinical data of 200 patients aged ≥ 60 years who received inpatient treatment for COVID-19 at the First Hospital of Qinhuangdao between December 13, 2022 and January 16, 2023 was done. The study population was divided into two groups, death (n = 55) and survival (n = 145), and the clinical data of patients in both groups were compared. The new risk predictors of death and their respective β values were obtained by Logistic univariate regression analysis and multivariate analysis. The score weights of the variables contained in the new models were determined according to the β values. New models were constructed, and the AUC of the ROC curve was applied to evaluate the predictive performance of each model. Logistic multivariate regression analysis revealed CRP/albumin ratio (β = 1.517, P < 0.001), AST (β = 0.879, P = 0.021), and D-dimer (β = 1.025, P = 0.008) as independent risk predictors of death in elderly inpatients with COVID-19. The construction of models to predict the risk of in-hospital mortality in elderly patients with COVID-19 showed that the AUC for Model 1 (CURB-65 score) to predict the risk of death during hospitalization in elderly patients with COVID-19 was 0.777 (95%CI: 0.700-0.854), the AUC for Model 2 (CURB-65 + CRP/albumin ratio score) was 0.829 (95%CI: 0.762–0.896), the AUC for Model 3 (CURB-65 + CRP/albumin ratio + AST score) was 0.846 (95%CI: 0.782–0.910), and the AUC for Model 4 (CURB-65 + CRP/albumin ratio + D-dimer score) was 0.836 (95%CI: 0.771–0.901), Model 5 (CRP/albumin ratio + AST + D-dimer score) had an AUC of 0.782 (95%CI: 0.704–0.860), Model 6 (CURB-65 + CRP/albumin ratio + AST + D-dimer score) had an AUC of 0.849 (95%CI: 0.785–0.914), Model 7 (CURB-65 + AST score) had an AUC of 0.817 (95%CI: 0.747–0.886), Model 8 (CURB-65 + D-dimer score) had an AUC of 0.796 (95%CI: 0.723–0.869), and Model 9 (CURB-65 + AST + D-dimer score) had an AUC of 0.824 (95%CI: 0.754–0.895). Model 3 and Model 6 showed statistically significant differences (P < 0.05) compared to AUC with CURB-65 scores. The difference in AUC between Model 3 and Model 6 is not statistically significant, but the sensitivity of Model 3 is higher at 0.898. Model 3 and Model 6 have better predictive power compared to the CURB-65 score for the risk of death during hospitalization for elderly patients with COVID-19, and Model 3 is more concise.