食管癌术后新发房颤风险预测模型的构建与评价

Risk factors and prediction model of postoperative atrial fibrillation surgery after esophageal cancer surgery

目的探讨食管癌术后新发房颤的危险因素,构建风险预测模型,并对其有效性进行评价。方法收集淮安第一医院食管外科2019年12月至2022年4月期间收治的1509例行食管癌手术患者的临床资料,根据术后是否新发房颤分为两层,每层按7∶3比例随机划分为训练集和测试集。在训练集人群中,采用多因素logistic回归方法建立预测模型,绘制预测模型列线图,绘制ROC曲线及校正曲线,评价预测模型的区分能力和校准能力。测试集人群用于验证预测模型。结果1509例患者中男1067例(70.71%),女442例(20.29%);年龄41~91岁,平均(68.76±7.15)岁。术后新发房颤247例(16.4%),未新发房颤1262例(83.6%)。训练集1039例(68.9%),应用多因素logistic回归模型分析结果显示,年龄、性别、体质量指数(BMI)、肺部感染、使用有创呼吸机、胸腔积液需要额外引流是术后新发房颤的影响因素。训练集预测模型ROC曲线下面积AUC为0.775(95%CI:0.737~0.812,P<0.001),提示模型预测区分能力较高。校正曲线及Hosmer-Lemeshow检验结果P=0.796,提示模型预测能力的一致性好。测试集470例(31.1%),将上述预测模型带入测试集人群中,结果显示预测模型ROC曲线下面积AUC为0.773(95%CI:0.719~0.826,P<0.001),提示预测模型在测试集人群中仍具有较高的区分能力。结论年龄、性别、BMI、肺部感染、使用有创呼吸机、胸腔积液需要额外引流是术后新发房颤的重要影响因素。通过构建临床预测模型,及时预测、预防和管理术后房颤有利于提高食管癌术后患者的预后质量。

Objective To develop a risk prediction lineogram of neooperative atrial fibrillation in patients with esophageal cancer.Methods The clinical data of 1509 patients undergoing esophageal cancer surgery admitted to the department of esophageal surgery of our hospital from December 2019 to April 2022 were gathered,and they were divided into two layers according to whether they had new atrial fibrillation after surgery.In each layer,they were randomly divided into training set and test set in a ratio of 7∶3.In the training population,the multi-factor logistic regression method was used to establish the prediction model,and the line graph of the prediction model was drawn.The ROC curve and calibration curve were drawn to assess the differentiation ability and calibration ability of the prediction model.The test set population is used to validate the prediction model.Results A total of 1509 patients with esophageal cancer were included in the study,and the incidence of new atrial fibrillation after surgery was 247 patients(16.4%).A total of 1039 patients(68.9%)were enrolled in the training set.The multivariate logistic regression model indicated that age,gender,BMI,pulmonary infection,the use of invasive ventilator,and the need for additional drainage of fluid accumulation were the influencing factors for new postoperative atrial fibrillation.The AUC of the training set prediction model under ROC curve was 0.775(95%CI:0.737-0.812,P<0.001),indicating that the model has high predictive discrimination ability.Calibration curve and Hosmer-Lemeshow test results P=0.796,indicating that the model has good consistency of prediction ability.There were 470 subjects(31.1%)in the test set.The results showed that the AUC of the prediction model under the ROC curve was 0.773(95%CI:0.719-0.826,P<0.001),indicating that the prediction model still has a high discriminative ability in the test set population.Conclusion Patients with age,gender,BMI,pulmonary infection,the use of invasive ventilator,and the need for additional drainage of effusion are at higher risk of new atrial fibrillation after surgery.The timely prediction,prevention and management of POAF are crucial to improve the prognostic quality of postoperative patients with esophageal cancer by constructing clinical prediction models.