ABSTRACT
Introduction Accurate prognosis is important either after acute infection or during long-term follow-up of patients infected by SARS-CoV-2. This study aims to predict COVID-19 severity based on clinical and biological indicators, and to identify biomarkers for prognostic assessment. Methods We included 261 Vietnamese COVID-19 patients, who were classified into moderate and severe groups. Disease severity prediction based on biomarkers and clinical parameters was performed by applying machine learning and statistical methods using the combination of clinical and biological data. Results The random forest model could predict with 97% accuracy the likelihood of COVID-19 patients who subsequently worsened to severe condition. The most important indicators were IL-6, Ferritin, and D-dimer. The model could still predict with 92% accuracy after removing IL-6 from analysis to generalize applicability of the model to hospitals with limited capacity for IL-6 testing. The five most effective indicators were CRP, D-dimer, IL-6, Ferritin, and dyspnoea. Two different sets of biomarkers (D-dimer, IL-6, and Ferritin;and CRP, D-dimer, and IL-6) are applicable for the assessment of disease severity and prognosis. The two biomarker sets were further tested through the machine learning algorithms random forest;and relatively validated on two Danish COVID-19 patient groups (n=32;and n=100). The results indicated various biomarker sets combined with clinical data can be used for detection of potential develop severe conditions. Conclusion This study provided a simple and reliable model using two different sets of biomarkers to assess disease severity and predict clinical outcomes in COVID-19 patients in Vietnam.