摘要目的:构建脑肿瘤术后患者ICU谵妄风险预测模型，并对模型在脑肿瘤患者ICU谵妄风险预测中的应用价值进行验证。方法:采用便利抽样法，选取2020年12月—2021年7月同济大学附属东方医院神经外科ICU收治的336例脑肿瘤术后患者为建模组，根据是否发生ICU谵妄，将患者分为谵妄组（ n=101）和非谵妄组（ n=235），采用ICU意识模糊评估法、镇静程度评估量表对患者进行评估。采用单因素分析、二项Logistic回归分析探讨脑肿瘤术后患者发生ICU谵妄的影响因素，并构建预测模型；采用受试者工作特征曲线下面积（ AUC）、Calibration校准曲线评价模型的区分度、校准度；采用1 000次Bootstrap自抽样法对模型进行验证。按照同样标准，选取2021年8—11月的144例脑肿瘤患者为验证组，对模型进行验证。 结果:最终进入模型的预测因子为年龄（ OR=1.033）、神经外科ICU入住时长（ OR=1.298）、手术时长（ OR=1.006）、使用苯二氮卓类药物（ OR=5.850）、身体约束（ OR=2.820）、肿瘤直径（ OR=1.385）、双侧脑肿瘤占位（ OR=3.604）。预测模型的 AUC为0.935（95% CI：0.911~0.960， P<0.01），约登指数为0.747，灵敏度为92.1%，特异度为82.6%；Bootstrap法的内部验证一致性指数为0.916，且校正曲线与理想曲线拟合良好。模型验证结果显示，灵敏度为86.4%，特异度为85.0%。 结论:脑肿瘤术后患者ICU谵妄风险预测模型具有较好的预测效能，可为医护人员早期采取预防性管理措施提供参考。

abstractsObjective:To construct a risk prediction model for ICU delirium in patients with brain tumor surgery, and to verify the application value of the model in predicting the risk of ICU delirium in patients with brain tumor.Methods:Using the convenient sampling method, a total of 336 postoperative patients with brain tumors who were admitted to Neurosurgery ICU of East Hospital Affiliated to Tongji University from December 2020 to July 2021 were selected as the modeling group. Patients were divided into the delirium group ( n=101) and the non-delirium group ( n=235) according to the occurrence of ICU delirium. The patients were evaluated using Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) and Richmond Agitation Sedation Score (RASS) . Univariate analysis and binomial logistic regression analysis were used to explore the influencing factors of ICU delirium in patients after brain tumor surgery and to construct a predictive model. The area under receiver operating characteristic curve ( AUC) and Calibration curve were used to evaluate the discrimination and calibration of the model. The model was validated by 1 000 Bootstrap self-sampling methods. According to the same criteria, a total of 144 patients with brain tumors from August to November 2021 were selected as the validation group to verify the model. Results:The predictors that finally entered the model were age ( OR=1.033) , length of stay in neurosurgical ICU ( OR=1.298) , length of operation ( OR=1.006) , use of benzodiazepines ( OR=5.850) , physical restraint ( OR=2.820) , tumor diameter ( OR=1.385) and bilateral brain tumor mass ( OR=3.604) . The prediction model of AUC was 0.935 (95% CI: 0.911-0.960, P<0.01) , the Youden index was 0.747, the sensitivity was 92.1% and the specificity was 82.6%. The internal validation consistency index of the Bootstrap method was 0.916, and the calibration curve fit the ideal curve well. The model validation results showed that the sensitivity was 86.4%, the specificity was 85.0%. Conclusions:The prediction model of delirium risk in ICU patients after brain tumor surgery has good predictive performance, which can provide reference for medical staff to take preventive management measures in the early stage.