Predictive model for coronavirus disease 2019 severity based on blood biomarkers: a retrospective study

Abstract

Objective: To develop and validate a clinical prediction model for assessing the severity of coronavirus disease 2019 (COVID-19) using blood biomarkers, aiming to support clinical decision-making and treatment guidance. Methods: A retrospective cohort study was conducted at Beijing Shijitan Hospital on January 5, 2023, including SARS-CoV-2 positive patients with initial chest CT-detected from outpatient and emergency departments. Data on demographics, symptoms, and blood biomarkers were collected. Patients were categorized into non-severe (mild and moderate) and severe (severe and critical) groups based on clinical symptoms and disease progression. Outpatient data served as the training set for modeling and validation using logistic regression and 10-fold cross validation. Emergency department data functioned as an independent external validation set to test the model’s generalizability. Results: The study included 1,007 patients, with 778 in the training set and 229 in the validation set. The C-reactive protein (CRP), neutrophil count (NE), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR) were significantly higher in the severe COVID-19 group, while lymphocyte count (LY) and eosinophil count (EO) were significantly lower in the non-severe group (p < 0.001). The predictive model integrating these factors exhibited high discriminative power, achieving an AUC of 0.85, accuracy of 0.80, sensitivity of 0.73, and specificity of 0.81 in 10-fold cross validation, and an AUC of 0.86, accuracy of 0.82, sensitivity of 0.60, and specificity of 0.90 in the validation set. Conclusion: The predictive model, informed by blood biomarkers, successfully discriminates against COVID-19 patients at higher risk for severe outcomes, offering a valuable tool for clinical management and resource optimization.