Development and validation a nomogram for predicting the risk of severe COVID-19: A multi-center study in Sichuan, China

• Published in: PloS one 2020-05, Vol.15 (5), p.e0233328-e0233328
• Main Authors: Zhou, Yiwu, He, Yanqi, Yang, Huan, Yu, He, Wang, Ting, Chen, Zhu, Yao, Rong, Liang, Zongan
• Format: Article
• Language: English
• Subjects: Adolescent; Adult; Aged; Betacoronavirus; Biology and Life Sciences; Body temperature; Calibration; Cardiovascular diseases; China; Confidence intervals; Coronavirus Infections - diagnosis; Coronaviruses; Cough; COVID-19; Decision analysis; Dyspnea; Emergency medical care; Female; Fever; Health aspects; Health risks; Humans; Hypertension; Kidney diseases; Liver diseases; Logistic Models; Male; Medicine and Health Sciences; Middle Aged; Mortality; Nomograms; Nomographs; Nomography (Mathematics); Pandemics; Pneumonia, Viral - diagnosis; Prediction models; Regression analysis; Regression models; Respiration; Retrospective Studies; Risk; Risk Factors; SARS-CoV-2; Severity of Illness Index; Statistical analysis; Statistics; Viral diseases; Young Adult
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0233328

Since December 2019, coronavirus disease 2019 (COVID-19) emerged in Wuhan and spread across the globe. The objective of this study is to build and validate a practical nomogram for estimating the risk of severe COVID-19. A cohort of 366 patients with laboratory-confirmed COVID-19 was used to develop a prediction model using data collected from 47 locations in Sichuan province from January 2020 to February 2020. The primary outcome was the development of severe COVID-19 during hospitalization. The least absolute shrinkage and selection operator (LASSO) regression model was used to reduce data size and select relevant features. Multivariable logistic regression analysis was applied to build a prediction model incorporating the selected features. The performance of the nomogram regarding the C-index, calibration, discrimination, and clinical usefulness was assessed. Internal validation was assessed by bootstrapping. The median age of the cohort was 43 years. Severe patients were older than mild patients by a median of 6 years. Fever, cough, and dyspnea were more common in severe patients. The individualized prediction nomogram included seven predictors: body temperature at admission, cough, dyspnea, hypertension, cardiovascular disease, chronic liver disease, and chronic kidney disease. The model had good discrimination with an area under the curve of 0.862, C-index of 0.863 (95% confidence interval, 0.801-0.925), and good calibration. A high C-index value of 0.839 was reached in the interval validation. Decision curve analysis showed that the prediction nomogram was clinically useful. We established an early warning model incorporating clinical characteristics that could be quickly obtained on admission. This model can be used to help predict severe COVID-19 and identify patients at risk of developing severe disease.