截至2023年4月30日,COVID-19大流行已导致全球690多万人死亡。病毒继续传播和变异,导致不断发展的病理和生理过程。必须重新评估预测因素,以确定早期疾病进展的风险。
对2022年12月15日至2023年2月15日从上海交通大学医学院附属新华医院出院的1379名COVID-19患者进行了回顾性研究。患者症状,合并症,人口统计,生命体征,和实验室测试结果进行了系统记录。数据集被分成测试集和训练集,并采用15种不同的机器学习算法构建预测模型。评估这些模型的准确性和接受者工作特征曲线下面积(AUROC),并选择表现最好的模型进行进一步分析。
■由15种机器学习算法生成的模型的AUROC均超过90%,其中10个的准确率也超过了90%。光梯度提升模型成为最优选择,准确度为0.928±0.0006,AUROC为0.976±0.0028。值得注意的是,对住院死亡率影响最大的因素是生长刺激表达基因2(ST2,19.3%),白细胞介素-8(IL-8,17.2%),白细胞介素-6(IL-6,6.4%),年龄(6.1%),NT-proBNP(5.1%),白细胞介素2受体(IL-2R,5%),肌钙蛋白I(TNI,4.6%),轻度梯度提升模型中的充血性心力衰竭(3.3%)。
■ST-2,IL-8,IL-6,NT-proBNP,IL-2R,TNI,年龄和充血性心力衰竭是COVID-19患者院内死亡率的重要预测因素.
UNASSIGNED: As of 30 April 2023, the COVID-19 pandemic has resulted in over 6.9 million deaths worldwide. The virus continues to spread and mutate, leading to continuously evolving pathological and physiological processes. It is imperative to reevaluate predictive factors for identifying the risk of early disease progression.
UNASSIGNED: A retrospective study was conducted on a cohort of 1379 COVID-19 patients who were discharged from Xin Hua Hospital affiliated with Shanghai Jiao Tong University School of Medicine between 15 December 2022 and 15 February 2023. Patient symptoms, comorbidities, demographics, vital signs, and laboratory test results were systematically documented. The dataset was split into testing and training sets, and 15 different machine learning algorithms were employed to construct prediction models. These models were assessed for accuracy and area under the receiver operating characteristic curve (AUROC), and the best-performing model was selected for further analysis.
UNASSIGNED: AUROC for models generated by 15 machine learning algorithms all exceeded 90%, and the accuracy of 10 of them also surpassed 90%. Light Gradient Boosting model emerged as the optimal choice, with accuracy of 0.928 ± 0.0006 and an AUROC of 0.976 ± 0.0028. Notably, the factors with the greatest impact on in-hospital mortality were growth stimulation expressed gene 2 (ST2,19.3%), interleukin-8 (IL-8,17.2%), interleukin-6 (IL-6,6.4%), age (6.1%), NT-proBNP (5.1%), interleukin-2 receptor (IL-2R, 5%), troponin I (TNI,4.6%), congestive heart failure (3.3%) in Light Gradient Boosting model.
UNASSIGNED: ST-2, IL-8, IL-6, NT-proBNP, IL-2R, TNI, age and congestive heart failure were significant predictors of in-hospital mortality among COVID-19 patients.