基于机器学习变量筛选的老年胸腔镜肺癌手术患者术后谵妄预测模型的构建与验证OA
Construction and validation of a predictive model for postoperative delirium in elderly patients with thoracoscopic lung cancer surgery based on machine learning variable screening
目的 探讨老年胸腔镜肺癌根治术患者发生术后谵妄(POD)的危险因素,并构建列线图模型.方法 采用回顾性研究方法,收集2022年1月—2025年1月邯郸市中心医院收治的597例老年(≥65岁)胸腔镜肺癌根治术患者的资料,并采用随机数字表法按7∶3的比例将患者分为建模集(n=417)和验证集(n=180).根据谵妄评估(CAM)量表将患者分为POD组和非POD组.采用机器学习变量筛选方法(LASSO回归)与传统Logistic回归相结合建立预测模型,构建列线图.采用受试者工作特征(ROC)曲线、校准曲线评价其区分度和校准度,通过Hosmer-Lemeshow检验进行拟合优度评估,并利用决策曲线分析(DCA)量化其临床净获益.结果 LASSO回归分析筛选出 8个非零系数预测变量.多因素Logistic回归分析确定5 个独立危险因素:年龄(OR=1.30,95％CI:1.070～1.193)、教育水平(OR=0.581,95％CI:0.344～0.982)、术前认知功能[蒙特利尔认知评估量表(MoCA)](OR=0.821,95％CI:0.745～0.904)、脑血管病史(OR=2.667,95％CI:1.325～5.367)、手术时间(OR=1.023,95％CI:1.010～1.036).整合上述5个跨生理-认知-社会维度的核心指标,构建列线图预测工具.该模型在建模集中的曲线下面积(AUC)为0.804(95％CI:0.757～0.850),在验证集中的AUC为0.793(95％CI:0.723～0.863),差异无统计学意义(P=0.804).经 Hosmer-Lemeshow 检验,建模集(x2=10.508,P=0.231)和验证集(x2=8.641,P=0.373)校准度良好.DCA显示在较宽的阈值范围内(建模集为0.01～0.56;验证集为0.01～0.50),使用该预测模型具有显著的临床净获益.结论 本研究构建的针对≥65岁老年胸腔镜肺癌根治术的列线图模型,可为早期识别高危患者并实施针对性干预提供量化工具.
Objective To explore the risk factors for postoperative delirium(POD)in elderly patients with thoracoscopic radical resection for lung cancer and construct a nomogram model.Meth-ods A retrospective study was conducted,data from 597 elderly patients(≥65 years old)who un-derwent thoracoscopic radical resection for lung cancer at Handan Central Hospital from January 2022 to January 2025 were collected.Patients were divided into modeling set(n=417)and validation set(n=180)in a 7∶3 ratio using a random number table method.Patients were categorized into POD group and non-POD group based on the Confusion Assessment Method(CAM)scale.A predictive model was established by combining machine learning variable screening methods(LASSO regression)with traditional Logistic regression to construct a nomogram model.The discrimination and calibration of the model were evaluated using the receiver operating characteristic(ROC)curve and calibration curve,respectively.The goodness-of-fit was assessed via the Hosmer-Lemeshow test,and the clini-cal net benefit was quantified using decision curve analysis(DC A).Results LASSO regression a-nalysis identified eight predictive variables with non-zero coefficients.Multivariate Logistic regres-sion analysis determined five independent risk factors:age(OR=1.30,95％CI,1.070 to 1.193),education level(OR=0.581,95％CI,0.344 to 0.982),preoperative cognitive function[Montre-al Cognitive Assessment Scale(MoCA)](OR=0.821,95％CI,0.745 to 0.904),history of cere-brovascular disease(OR=2.667,95％CI,1.325 to 5.367),and operative time(OR=1.023,95％CI,1.010 to 1.036).A nomogram prediction tool was constructed by integrating these five core indicators spanning physiological-cognitive-social dimensions.The area under the curve(AUC)of the model was 0.804(95％CI,0.757 to0.850)in the modeling set and 0.793(95％CI,0.723 to 0.863)in the validation set,with no significant difference(P=0.804).The Hosmer-Lemeshow test indicated good calibration in both the modeling set(x2=10.508,P=0.231)and the validation set(x2=8.641,P=0.373).DC A demonstrated a significant clinical net benefit of using the pre-dictive model across a wide range of threshold probabilities(0.01 to 0.56 in the modeling set and 0.01 to 0.50 in the validation set).Conclusion The nomogram model constructed in this study for elderly patients(≥65 years old)with thoracoscopic radical resection for lung cancer provides a quantitative tool for early identification of high-risk patients and implementation of targeted inter-ventions.
陈士欢;陈永学;侯俊德;程少飞;李立英
邯郸市中心医院麻醉科,河北邯郸,056004邯郸市中心医院麻醉科,河北邯郸,056004邯郸市中心医院麻醉科,河北邯郸,056004邯郸市中心医院麻醉科,河北邯郸,056004邯郸市中心医院麻醉科,河北邯郸,056004
医药卫生
谵妄肺癌胸腔镜风险因素列线图机器学习蒙特利尔认知评估量表决策曲线分析
deliriumlung cancerthoracoscopyrisk factorsnomogrammachine learningMontreal Cognitive Assessment Scaledecision curve analysis
《实用临床医药杂志》 2026 (6)
48-54,83,8
河北省重点研发计划(182777195)
评论