摘要： 目的 筛查低级别动脉瘤性蛛网膜下腔出血（aSAH）术后短期临床预后不良危险因素，并构建列线图（Nomogram）预后预测模型。方法 纳入2016年1月至2021年12月在重庆市开州区人民医院神经外科行动脉瘤夹闭术或栓塞术的aSAH患者共293例，世界神经外科学会联合会（WNFS）神经功能分级Ⅰ~Ⅱ级，术后3个月采用改良Rankin量表（mRS）评估临床预后，单因素和多因素逐步法Logistic回归分析筛查术后短期临床预后不良危险因素，通过构建Nomogram模型、绘制受试者工作特征（ROC）曲线并计算曲线下面积，评价模型预测效能，采用Bootstrap内部验证法、Hosmer-Lemeshow拟合优度检验和校准曲线对预测模型进行内部验证。结果 预后不良组（mRS评分3~6分，55例）≥ 65岁（χ² = 18.516，P = 0.000）和WFNS分级Ⅱ级（χ² = 9.491，P = 0.002），以及术后并发迟发性脑缺血（χ² = 28.355，P = 0.000）、分流依赖性脑积水（χ² = 33.497，P = 0.000）和颅内出血（χ² = 17.744，P = 0.000）比例均高于预后良好组（mRS评分0~2分，238例）。Logistic回归分析提示，年龄≥ 65岁（OR = 1.241，95%CI：1.021~1.772；P = 0.000）、术后并发迟发性脑缺血（OR = 9.462，95%CI：1.302~23.823；P = 0.010）和分流依赖性脑积水（OR = 6.092，95%CI：2.730~16.201；P = 0.000）是术后短期临床预后不良的危险因素；基于这3项危险因素构建Nomogram模型，ROC曲线显示曲线下面积为0.833（95%CI：0.772~0.897，P = 0.000），当Youden指数为0.520时灵敏度79.28%、特异度72.69%，截断值128分，该模型预测效能优于3项指标单独应用（均P = 0.000）。经Bootstrap内部验证法、Hosmer-Lemeshow拟合优度检验和校准曲线验证，模型区分度较高（一致性指数0.908）、稳定性良好（χ² = 1.078，P = 0.693）、校准度良好。结论 年龄≥ 65岁、术后并发迟发性脑缺血和分流依赖性脑积水是低级别aSAH患者术后短期临床预后不良的重要危险因素，基于这3项危险因素构建的Nomogram模型对低级别aSAH患者术后短期临床预后具有良好的预测价值。

关键词: 颅内动脉瘤, 蛛网膜下腔出血, 预后, 危险因素, Logistic模型, 列线图, ROC曲线

Abstract: Objective To screen the risk factors for short-term poor clinical prognosis in patients with low-grade aneurysmal subarachnoid hemorrhage (aSAH), and to construct the Nomogram model.Methods A total of 293 aSAH patients with World Federation of Neurosurgical Societies (WNFS) grade Ⅰ-Ⅱ treated by aneurysm clipping or aneurysm embolization in The People's Hospital of Kaizhou District of Chongqing from January 2016 to December 2021 were enrolled. The modified Rankin Scale (mRS) was used to evaluate the clinical prognosis at 3 months after surgery. Univariate and multivariate stepwise Logistic regression analysis was used to screen the risk factors for short-term poor clinical prognosis in patients with low-grade aSAH, and Nomogram model was constructed. The receiver operating characteristic (ROC) curve was drawn and the area under the curve (AUC) was calculated to evaluate the prediction efficiency of the model. The Bootstrap internal verification method, Hosmer-Lemeshow goodness of fit test and calibration curve were used to verify the model.Results Total 293 patients were divided into good prognosis group (0-2 score, n = 238) and poor prognosis group (3-6 score, n = 55) according to mRS score. The proportion of poor prognosis group was age ≥ 65 years old (χ² = 18.516, P = 0.000) and WFNS grade Ⅱ (χ² = 9.491, P = 0.002), the incidences of postoperative delayed cerebral ischemia (χ² = 28.355, P = 0.000), shunt dependent hydrocephalus (χ² = 33.497, P = 0.000) and intracranial hemorrhage (χ² = 17.744, P = 0.000) were higher than those of good prognosis group. Logistic regression analysis showed age ≥ 65 years old (OR = 1.241, 95%CI: 1.021-1.772; P = 0.000), postoperative delayed cerebral ischemia (OR = 9.462, 95%CI: 1.302-23.823; P = 0.010) and shunt dependent hydrocephalus (OR = 6.092, 95%CI: 2.730-16.201; P = 0.000) were the risk factors for short-term poor clinical prognosis in patients with low-grade aSAH. The Nomogram model was constructed based on the above 3 risk factors. ROC curve showed the AUC was 0.833 (95%CI: 0.772-0.897, P = 0.000), when Youden index was 0.520, the corresponding sensitivity was 79.28%, specificity was 72.69%, and the cut-off value was 128 points. The prediction efficiency of the model was better than that of the above 3 factors alone (P = 0.000, for all). Furthermore, the Bootstrap internal verification method, Hosmer-Lemeshow goodness of fit test and calibration curve confirmed the good discrimination (CI = 0.908), stability (χ² = 1.078, P = 0.693) and calibration.Conclusions Age ≥ 65 years old, postoperative delayed cerebral ischemia and shunt dependent hydrocephalus are the risk factors for short-term poor clinical prognosis in patients with low-grade aSAH. The Nomogram model constructed based on these 3 factors has good predictive value for short-term clinical prognosis in low-grade aSAH.

Key words: Intracranial aneurysm, Subarachnoid hemorrhage, Prognosis, Risk factors, Logistic models, Nomograms, ROC curve