A nomogram for predicting the severity of COVID-19 infections among patients in Chengdu, China

Abstract

Introduction: Although most COVID-19 infections are currently mild, poor prognoses and fatalities continue to occur, which remain a threat to the safety of people in China. The goal of this study was to create an efficient model that combines the clinical characteristics with computed tomography (CT) scores at the time of admission to predict the severity of COVID-19. Methodology: A total of 346 COVID-19 patients in the current study, of whom 46 had severe infections and 300 had non-severe infections according to the clinal outcomes. Clinical, laboratory, CT findings, and CT scores at admission were collected. To identify the independent risk factors, univariable and multivariable logistic regression analyses were performed. A nomogram model was built with the extracted risk factors. The calibration curve and decision curve (DCA) operated to validate model performance. Results: The receiver operating characteristic curve indicated that the severity CT score had an area under the curve of 0.933 (95% CI, 0.901-0.965) and a cut-off value of 6.5 (sensitivity, 95.70%; specificity, 78%). The CT score, age, lactic dehydrogenase and hydroxybutyrate dehydrogenase levels, and hypertension were exacted for the nomogram. The nomogram had good calibration (P = 0.539) and excellent clinical value based on the DCA. Conclusions: The nomogram presented herein could be a valuable model to predict severe COVID-19 among patients in Chengdu, China.