新型冠状病毒肺炎患者死亡风险列线图模型的构建与验证

目的构建列线图模型以预测新型冠状病毒病2019(COVID-19)的死亡风险，以早期筛选死亡高危患者。 方法收集2020年1月至2020年4月武汉大学人民医院(东院)和2022年4月至2022年5月上海市第九人民医院(北院)收治COVID-19患者的临床资料。以武汉大学人民医院患者(166例)作为训练集，上海市第九人民医院患者(52例)作为验证集。采用先单因素后多因素Logistic回归分析确定死亡的独立危险因素，应用R语言构建列线图模型。采用受试者工作特征曲线(ROC)、C指数及校准曲线评估列线图模型的预测准确性及判断能力，决策曲线分析评估模型的临床应用价值。通过验证集对模型进行外部验证。 结果本研究共纳入重型/危重型COVID-19患者218例，其中67例(30.73%)死亡，多因素Logistic回归分析显示，≥3种基础疾病、APACHE Ⅱ评分(5~40分)、中性粒细胞/淋巴细胞(0~90)、乳酸(0~16mmol/L)均是死亡的独立危险因素。ROC曲线分析显示，训练集的曲线下面积(AUC)为0.869(95%CI：0.811~0.927)，验证集AUC为0.797(95%CI：0.671~0.924)，训练集与验证集校准曲线经Hosmer-Lemeshow拟合优度检验(P＝0.473，P＝0.421)。临床决策曲线分析表明，该列线图预测模型的临床应用价值高。 结论本研究构建COVID-19患者死亡风险列线图模型预测效能良好，可个体化、可视化、图形化预测，有助于医师早期做出合适临床决策及诊疗。

Establishment and validation of a mortality prognostic nomogram model for COVID-19 patients

ObjectiveTo establish a nomogram model for predicting the mortality risk of COVID-19 patients, in order to early screen those who are in higher risk. MethodsAll the clinical data of COVID-19 patients were collected from Eastern Campus of Renmin Hospital of Wuhan University during 2020 January to April and North Campus of Shanghai Ninth People Hospital during 2022 April to May. Patients (n=166) from the Renmin Hospital of Wuhan Universiy were considered as training set, while those(n=52) from Shanghai Ninth Peoplef Wuhan Universiy were considered as training sets of Renmin Hospital of Wuhan University during 2020 January to April and North Campuste logistic regression analysis and the R Programming Language was used to conduct the nomogram model. The prediction accuracy and judgment ability of nomogram model were evaluated by receiver operating characteristic curve (ROC), C index and calibration curve, and the clinical application value was evaluated by decision curve analysis. The model was externally validated by the validation set. ResultsA total of 218 patients with severe/critical COVID-19 were included in this study, among whom 67 of them died (30.73%). Multivariate logistic regression analysis showed that more than three kinds of underlying diseases, APACHEⅡ score, neutrophile granulocyte/lymphocyte, and lactate were all independent risk factors. ROC analysis showed that the area under curve (AUC) of training set was 0.869 (95% CI: 0.811-0.927), while the AUC of validation set was 0.797 (95% CI: 0.671-0.924). The calibration curves between the training set and the validation set were tested by Hosmer lemeshow test (P=0.473, P=0.421). Decision curve analysis showed that the nomogram prediction model had high clinical application value. ConclusionsThe nomogram model presents significantly predictive value for mortality risk of COVID-19, which is individualized, visualized and graphically predicted. Whatpredict, it is benefit for physician to make appropriate clinical decisions and treatment at early stage.