摘要：

目的: 筛查颅脑创伤后脑疝患者去骨瓣减压术后脑积水的危险因素，并基于危险因素构建贝叶斯网络模型。方法: 纳入2020年3月至2022年1月在东南大学附属南京同仁医院行去骨瓣减压术的77例颅脑创伤后脑疝患者，根据术后是否并发脑积水分为脑积水组（25例）和无脑积水组（52例），单因素和多因素Logistic回归分析筛查颅脑创伤后脑疝患者去骨瓣减压术后脑积水的危险因素，并基于危险因素构建贝叶斯网络模型，绘制受试者工作特征（ROC）曲线和校准曲线并行Hosmer-Lemeshow拟合优度检验。结果: 脑积水组患者入院时Glasgow昏迷量表（GCS）评分（t = 2.178，P = 0.032）、术后腰椎穿刺脑脊液置换术比例（χ² = 8.675，P = 0.003）、术后血清β2微球蛋白水平（t = 11.146，P = 0.000）低于无脑积水组，术前合并蛛网膜下腔出血（χ² = 5.901，P = 0.015）、双侧手术（χ² = 6.441，P = 0.011）、术中未缝合硬脑膜（χ² = 9.759，P = 0.002）、术后脑室积血（χ² = 8.938，P = 0.003）、术后中线移位>10 mm（χ² = 7.589，P = 0.006）、术后并发颅内感染（χ² = 4.519，P = 0.034）比例以及术后昏迷时间（t = 2.709，P = 0.008）高于无脑积水组。Logistic回归分析显示，术前合并蛛网膜下腔出血（OR = 1.885，95%CI：1.432 ~ 2.240；P = 0.012）、术中未缝合硬脑膜（OR = 1.468，95%CI：1.215 ~ 1.930；P = 0.006）、术后昏迷时间长（OR = 1.574，95%CI：1.358 ~ 1.926；P = 0.007）、术后脑室积血（OR = 1.550，95%CI：1.254 ~ 1.768；P = 0.010）和术后血清β2微球蛋白水平升高（OR = 1.622，95%CI：1.165 ~ 1.840；P = 0.004）是颅脑创伤后脑疝患者去骨瓣减压术后脑积水的危险因素。基于上述5项危险因素构建贝叶斯网络模型，ROC曲线下面积为0.886（95%CI：0.823 ~ 0.925，P = 0.000），校准曲线显示预测概率与实际概率之间具有良好的一致性，Hosmer-Lemeshow拟合优度检验显示差异无统计学意义（χ² = 8.760，P = 0.232），表示该模型具有良好的区分度、校准度和准确性。结论: 术前合并蛛网膜下腔出血、术中未缝合硬脑膜、术后昏迷时间长、术后脑室积血、术后血清β2微球蛋白水平升高是颅脑创伤后脑疝患者去骨瓣减压术后脑积水的危险因素，基于上述5项危险因素构建的贝叶斯网络模型对术后并发脑积水风险具有重要预测价值。

关键词: 脑损伤, 创伤性, 脑膨出, 减压颅骨切除术, 脑积水, 手术后并发症, 危险因素, Logistic模型, 贝叶斯定理

Abstract:

Objective: To screen the risk factors of hydrocephalus after decompressive craniectomy in patients with cerebral hernia after traumatic brain injury (TBI), and construct a Bayesian network model based on the risk factors. Methods: A total of 77 patients with cerebral hernia after TBI who underwent decompressive craniotomy in Nanjing Tongren Hospital Affiliated to Southeast University from March 2020 to January 2022 were included. They were divided into hydrocephalus group (n = 25) and non -hydrocephalus group (n = 52) according to whether hydrocephalus was complicated after surgery. The risk factors of hydrocephalus after decompressive craniectomy in patients with cerebral hernia after TBI were analyzed by univariate and multivariate Logistic regression analyses. The Bayesian network model was constructed based on the risk factors, and the receiver operating characteristic (ROC) curve and calibration curve were drawn and Hosmer-Lemeshow goodness-of-fit test was conducted. Results: In hydrocephalus group, the Glasgow Coma Scale (GCS) score at admission (t = 2.178, P = 0.032), the ratio of cerebrospinal fluid replacement after lumbar puncture (χ² = 8.675, P = 0.003), and the level of β2-microglobulin after operation (t = 11.146, P = 0.000) were lower than those in non-hydrocephalus group, while subarachnoid hemorrhage (χ² = 5.901, P = 0.015), bilateral operation (χ² = 6.441, P = 0.011), the ratio of dural unstitched during operation (χ² = 9.759, P = 0.002), postoperative intraventricular hemorrhage (χ² = 8.938, P = 0.003), postoperative midline displacement > 10 mm (χ² = 7.589, P = 0.006), and intracranial infection (χ² = 4.519, P = 0.034), as well as postoperative coma time (t = 2.709, P = 0.008) were higher than those in non-hydrocephalus group. Logistic regression analysis showed that subarachnoid hemorrhage (OR = 1.885, 95%CI: 1.432-2.240; P = 0.012), dural unstitched during operation (OR = 1.468, 95%CI: 1.215-1.930; P = 0.006), long postoperative coma time (OR = 1.574, 95%CI: 1.358-1.926; P = 0.007), postoperative intraventricular hemorrhage (OR = 1.550, 95%CI: 1.254-1.768; P = 0.010), the level of β2-microglobulin increased after operation (OR = 1.622, 95%CI: 1.165-1.840; P = 0.004) were risk factors for hydrocephalus after decompressive craniectomy in patients with cerebral hernia after TBI. Based on these 5 factors, the Bayesian network model was constructed, and the area under ROC curve was 0.886 (95%CI: 0.823-0.925, P = 0.000). The calibration curve showed that there was a good consistency between the predicted probability and the actual probability, while the Hosmer-Lemeshow goodness- of-fit test showed no significant difference (χ² = 8.760, P = 0.232), which indicated that the model had good discrimination, calibration and accuracy. Conclusions: Subarachnoid hemorrhage, dural unstitched during operation, long postoperative coma time, postoperative intra ventricular hemorrhage, and elevated β2-microglobulin level are the risk factors for hydrocephalus after decompressive craniectomy in patients with cerebral hernia after TBI.

Key words: Brain injuries, traumatic, Encephalocele, Decompressive craniectomy, Hydrocephalus, Postoperative complications, Risk factors, Logistic models, Bayes theorem