脓毒症患者相关脑病预测模型的建立和验证

Establishment and validation of a predictive model for sepsis-associated encephalopathy

目的探讨脓毒症患者发生脓毒症相关脑病(SAE)的危险因素,建立简便、易用的预测模型并进行验证。方法回顾性分析徐州医科大学附属医院2017年1月至2021年12月入住重症监护病房(ICU)脓毒症患者的临床资料,根据纳入排除标准,确定最终入选病例,将2017年1月至2019年12月收集的病例作为训练队列组(n=640),将2020年1月至2021年12月收集的病例作为验证队列组(n=300)。将训练队列组患者资料进行Logistic回归分析,确定SAE发生的危险因素,建立回归方程,并可视化为列线图。验证队列组对建立的回归方程进行验证,通过绘制受试者工作特征(receiver operating characteristic,ROC)曲线及计算ROC曲线下面积(area under the curve,AUC)评价模型的区分度,通过Hosmer-Lemeshow检验和校准图评价模型的校准度。结果本研究共纳入940例患者,单因素及多因素Logistic回归结果表明,高龄、使用升压药、高中枢神经特异蛋白(S100β)水平、低脉搏血氧饱和度(SpO_(2))和低蛋白血症5个因素为SAE发病的独立危险因素(P<0.05),纳入预测模型,该预测模型的AUC在训练和验证队列组分别为0.810(95%CI 0.763~0.857)和0.813(95%CI 0.740~0.885),模型的校准曲线在训练和验证队列组均与平面直角坐标系中45°的直线重合度较高,提示该模型的表现良好。结论本研究建立的预测模型可以科学、有效地对SAE的发生进行预测,操作简便、快速,具有重要的临床价值。

Objective To explore the independent risk factors for predicting the incidence of sepsis-associated encephalopathy(SAE),and to establish a simple and easy-to-use prediction model.Methods Patients diagnosed with sepsis who were admitted to ICU of the Affiliated Hospital of Xuzhou Medical University from January 2017 to December 2021 were collected retrospectively.The final cases were determined according to the inclusion and exclusion criteria,and the cases from January 2017 to December 2019 were included into training cohort and the cases from January 2020 to December 2021 were included into validation cohort.In the training cohort,independent predictive factors related to SAE were determined by Logistic regression,and a nomogram was established.The nomogram was validated in the validation cohort,and the performance of the prediction model was evaluated according to the area under receiver operating characteristic(AUC)curve,Hosmer-Lemeshow test and calibration curve.Results In this study,a total of 940 patients were included.Independent risk factors of SAE were older age,vasopressor,high S100βlevel,low SpO_(2)and hypoalbuminemia.Based on these factors,a nomogram was established.The AUC of the nomogram was 0.810(95%CI 0.763-0.857)and 0.813(95%CI 0.740-0.885)in the training cohort and validation cohort respectively.The calibration curve had a high degree of coincidence with a straight line of 45°in the plane Cartesian coordinate system,indicating that the model performed well.Conclusions The nomogram can predict the occurrence of SAE scientifically and effectively,which is simple and fast,and has important clinical value.