Soetomo COVID-19 Prognostic Score: A Multi-Parametric Model for Early Prediction of Disease Severity of COVID-19 in Tertiery-Resource Hospital

Abstract

Objective: Coronavirus disease 2019 (COVID-19) became a global pandemic, with high mortality in severely ill patients. This study aimed to develop a novel scoring system to prognosticate disease severity in COVID-19 patients that is effective and widely available in tertiary medical resource settings. Material and Methods: Laboratory-confirmed COVID-19 patients were enrolled in this retrospective cohort, divided into severe and non-severe groups. We randomly assigned 70% of the subjects to establish a novel scoring system, while the remaining 30% was used for internal validation. The model was constructed by multivariate logistic regression using the first clinical, laboratory, and radiological finding of statistically analysis of group patients. receiver operating characteristic (ROC) and cross-tabulation were used to evaluate the performance of our score and compare it with other models. Results: A total of 599 patients were included. The Soetomo COVID-19 prognostic score predictors included age, fever, specific comorbidities (diabetes, hypertension, cardiac disease, lung tuberculosis), respiratory rate, heart rate, SF ratio, whole blood cell (WBC) count, neutrophil lympocyte ratio (NLR), blood urea nitrogen (BUN), and a RALE score. The area under the ROC of the model indicated an excellent discriminatory ability (training datasets 0.715 [95% CI 0.664-0.767, p-value<0.001]; testing datasets 0.720 [95% CI 0.638-0.802, p-value<0.001]). Our scoring system was superior to both qSOFA and MEWS regarding predictive value. The sensitivity and specificity were 60.6% and 82.5%, respectively. Conclusion: The developed scoring system accurately predicted a significant proportion of severe disease in COVID-19 patients.