MON-300 Enhancing the CURB-65 Score for In-Hospital Mortality Prediction in COVID-19: A Glucocentric Modeling Approach

Abstract

Abstract F. Mohamed: None. Introduction: A bidirectional relationship between COVID-19 and diabetes mellitus (DM) has been described. While the CURB-65 score has traditionally been utilized to predict mortality in cases of community-acquired pneumonia (CAP), its performance in predicting mortality in COVID-19 CAP is suboptimal. This analysis aims to enhance the predictive accuracy for in-hospital mortality among COVID-19 patients by augmenting the CURB-65 score with biochemical variables. Methods: A single centre retrospective observational analysis assessed the effectiveness of the CURB-65 score in predicting in-hospital mortality among adult patients hospitalized with moderate to severe COVID-19 from March to September 2020. The performance of CURB-65 for in-hospital mortality was assessed through Receiver Operator Characteristic Curve (ROC) analysis. A binary logistic regression model was proposed to identify the driving factors in the low-risk CURB-65 group and two expanded CURB-65 scores were proposed. Results: Among the 517 patients admitted with moderate or severe COVID-19, 117 (22.6%) died. Patients with moderate or severe SARS-CoV-2 CAP, requiring admission, were risk stratified based on CURB-65, and a higher mortality rate was seen in each category. A 13%, 47%, and 64% mortality rate occurred within the low, medium, and high-risk groups respectively. However, 45% of the overall patients who died were classified as low-risk. 37 patients were diagnosed with new onset dysglycemia, with 22 (59.5%) of them dying, and 26% had DM. The chi-square test of association and univariate analysis indicated a significant correlation between new-onset dysglycemia and in-hospital mortality (p-value < 0.001). The CURB-65 score demonstrated a modest area under the ROC curve (AUC) of 0.75 (95% CI, 0.70 to 0.81) for in-hospital mortality in COVID-19 CAP. The cost-effective expanded CURB-65 score, incorporating an admission fasting plasma glucose (FPG) and neutrophil:lymphocyte (N:L) ratio, showed improved prognostic performance with an AUC of 0.80 (95% CI, 0.76 to 0.85). The second model included FPG, N:L ratio, lactate on arterial blood gas, and lactate dehydrogenase, which showed an AUC of 0.82 (95% CI, 0.78 -0.86). The integrated discrimination index (IDI) showed an 11% and 24 % higher discrimination slope respectively. Conclusion: The expanded CURB-65 score suggests that the addition of commonly used clinical and biochemical parameters, including an admission fasting glucose, enhances the predictive performance of in-hospital mortality among patients hospitalized with moderate or severe COVID-19 CAP. The study suggests that COVID-19 mortality prediction models could be broadly applied, given the similarities in the cytokine storm with other viruses, and will offer valuable insights in developing treatment policies on glucose optimization in the acute setting. Monday, June 3, 2024