Development and validation of a predicting nomogram for in-hospital mortality of COVID-19 Omicron variant: A cohort study of 1324 cases in Beijing Anzhen Hospital

Abstract

Coronavirus disease 2019 (COVID-19) is continuously posing high global public health concerns due to its high morbidity and mortality. This study aimed to construct a convenient risk model for predicting in-hospital mortality of COVID-19 Omicron variant. A total of 1324 hospitalized patients with Omicron variant were enrolled from Beijing Anzhen Hospital. During hospitalization, the Omicron variant mortality rate was found to be 24.4%. Using the datasets of clinical demographics and laboratory tests, three machine learning algorithms, including best subset selection, stepwise selection, and least absolute shrinkage and selection operator regression analyses were employed to identify the potential predictors of in-hospital mortality. The results found that a panel of twenty-four clinical variables (including age, hyperlipemia, stroke, tumor, and several cardiovascular markers) identified by stepwise selection model exhibited significant performances in predicting the in-hospital mortality of COVID-19. The resultant nomogram showed good discrimination, highlighted by the areas under the curve values of 0.88 for 10 days, 0.81 for 20 days, and 0.82 for 30 days, respectively. Furthermore, decision curve analysis showed a significant reliability and precision for the established stepwise selection model. Collectively, this study developed an accurate and convenience risk model for predicting the in-hospital mortality of COVID-19 Omicron.