上消化道出血患者术后早期再出血风险预测模型的构建及效果验证

Construction and validation of a predictive model for the risk of early postoperative rebleeding in patients with upper gastrointestinal bleeding

目的通过分析筛选上消化道出血患者内镜止血术后早期再出血的危险因素,构建风险预测模型并验证预测效果。方法选择医院2020年1月-2021年12月收治的上消化道出血内镜治疗患者共160例为训练集,选择医院2022年1-10月收治上消化道出血内镜治疗患者共40例为验证集。收集患者的相关资料。利用训练集,采用单因素及多因素Logistic回归分析筛选早期再出血的危险因素,据此构建风险预测模型;采用验证集检验其预测效果。结果训练集早期再出血患者共34例,单因素及多因素Logistic回归分析显示,上消化道出血患者内镜止血术后护理早期再出血危险因素有年龄≥60岁、入院血红蛋白<110 g/L、凝血酶原时间≥17 s、病变直径>2 cm、Blatchford评分≥6分、Rockall评分≥5分、Forrest分级Ⅰ级(P<0.05)。据此构建的预测模型ROC曲线下面积为0.940(95%CI:0.893~0.972),最佳截断值为24.80%,灵敏度为81.82%,特异度为91.20%,最大约登指数为0.730。以>24.80%为预测标准,验证集模型预测的灵敏度为80.00%,特异度为90.00%,Kappa系数为0.677,表明模型预测结果与实际情况具有较好的一致性。结论上消化道出血患者内镜止血术后早期再出血危险因素复杂,据此构建的风险预测模型预测准确率良好,能为临床预测上消化道出血内镜止血术后护理早期再出血风险提供可靠、有效工具。

Objective To construct a risk prediction model and validate the prediction effect by analysing and screening the risk factors for early rebleeding after endoscopic haemostasis in patients with upper gastrointestinal bleeding.Methods A total of 160 patients admitted to the hospital for endoscopic treatment of upper gastrointestinal bleeding from January 2020 to December 2021 were selected as the training set,and a total of 40 patients admitted to the hospital for endoscopic treatment of upper gastrointestinal bleeding from January to October 2022 were selected as the validation set.The relevant data of the patients were collected.Using the training set,single-factor and multifactor logistic regression analyses were used to screen the risk factors for early re-bleeding,and risk prediction models were constructed accordingly,the validation set was used to test its predictive effect.Results There were 34 patients with early rebleeding in the training set,and the unifactorial and multifactorial Logistic regression analyses showed that the risk factors for early rebleeding in the care of patients with upper gastrointestinal bleeding after endoscopic haemostasis were age≥60 years,admission haemoglobin<110 g/L,prothrombin time≥17 s,lesion diameter>2 cm,Blatchford score≥6 points,Rockall score≥5 points,and Forrest classification Grade I(P<0.05).The area under the ROC curve of the prediction model constructed accordingly was 0.940(95%CI 0.893~0.972),with an optimal cut-offvalue of 24.80%,sensitivity of 81.82%,specificity of 91.20%,and maximum Jordon's index of 0.730.To take>24.80%as the prediction criterion,the validation set model predicted a sensitivity of 80.00%,a specificity of with>24.80%as the prediction criterion,the sensitivity of the validation set model prediction was 80.00%,the specificity was 90.00%,and the Kappa coefficient was 0.677,which indicated that the model prediction results were in good agreement with the actual situation.Conclusion The risk factors of early rebleeding after endoscopic haemostasis in patients with upper gastrointestinal bleeding are complex,and the risk prediction model constructed accordingly has a good prediction accuracy,which can provide a reliable and effective tool for clinical prediction of the risk of early rebleeding after endoscopic haemostasis for upper gastrointestinal bleeding.