摘要：

目的: 筛查重型颅脑创伤患者并发颅内感染的危险因素，并基于危险因素构建风险预测列线图（Nomogram）模型。方法: 纳入2021年1月至2022年6月山东第一医科大学第二附属医院收治的130例重型颅脑创伤患者，根据是否并发颅内感染分为颅内感染组（27例）和无颅内感染组（103例），单因素和多因素逐步法Logistic回归分析筛查重型颅脑创伤并发颅内感染的危险因素，并基于危险因素构建Nomogram模型，绘制受试者工作特征曲线（ROC曲线）和校准曲线并行Hosmer⁃Lemeshow拟合优度检验。结果: 颅内感染组糖尿病（χ² = 5.356，P = 0.021）、开放性颅脑创伤（χ² = 4.248，P = 0.039）、合并脑脊液漏（校正χ² = 4.731，P = 0.030）、手术治疗（χ² = 8.284，P = 0.004）、并发重症感染（校正χ² = 6.479，P = 0.011）、气管插管（χ² = 6.487，P = 0.011）和气管切开（χ² = 4.072，P = 0.044）比例均高于无颅内感染组。Logistic回归分析显示，糖尿病（OR = 2.748，95%CI：1.417 ~ 8.654；P = 0.047）、合并脑脊液漏（OR = 4.483，95%CI：1.852 ~ 8.341；P = 0.031）、手术治疗（OR = 1.941，95%CI：1.483 ~ 8.842；P = 0.031）、并发重症感染（OR = 1.614，95%CI：1.113 ~ 5.682；P = 0.041）是重型颅脑创伤并发颅内感染的危险因素。基于这4项危险因素构建Nomogram模型，ROC曲线下面积为0.758（95%CI：0.641 ~ 0.875，P = 0.001），该模型预测重型颅脑创伤并发颅内感染的最佳截断值为175分；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 = 4.613，P = 4.412），表明该模型具有良好的区分度、校准度和稳定性。结论: 糖尿病、合并脑脊液漏、手术治疗、并发重型感染的重型颅脑创伤患者更易并发颅内感染，据此构建的Nomogram模型可以较好地预测重型颅脑创伤并发颅内感染风险。

关键词: 脑损伤，创伤性, 感染, 危险因素, Logistic模型, 列线图

Abstract:

Objective: To screen the risk factors of intracranial infection in patients with severe traumatic brain injury (sTBI) and establish a Nomogram model based on these risk factors. Methods: A total of 130 patients with sTBI admitted to The Second Affiliated Hospital of Shandong First Medical University from January 2021 to June 2022 were enrolled. They were divided into a group with intracranial infection (n = 27) and a group without intracranial infection (n = 103) according to whether complicated with intracranial infection. To analyze the risk factors of intracranial infection in patients with sTBI by univariate and multivariate stepwise Logistic regression, and construct a Nomogram model based on the risk factors to draw receiver operating characteristic (ROC) curve and calibration curve of this model and perform Hosmer-Lemeshow goodness of fit test. Results: The proportion of diabetes (χ² = 5.356, P = 0.021), open traumatic brain injury (χ² = 4.248, P = 0.039), cerebrospinal fluid (CBF) leakage (adjusted χ² = 4.731, P = 0.030), surgical treatment (χ² = 8.284, P = 0.004), severe infection (adjusted χ² = 6.479, P = 0.011), tracheal intubation (χ² = 6.487, P = 0.011) and tracheotomy (χ² = 4.072, P = 0.044) in intracranial infection group were higher than those in non - intracranial infection group. Logistic regression analysis showed diabetes (OR = 2.748, 95%CI: 1.417-8.654; P = 0.047), CBF leakage (OR = 4.483, 95%CI: 1.852-8.341; P = 0.031), surgical treatment (OR = 1.941, 95%CI: 1.483-8.842; P = 0.031) and severe infection (OR = 1.614, 95%CI: 1.113-5.682; P = 0.041) were risk factors for sTBI complicated with intracranial infection. The area under the curve (AUC) of ROC curve was 0.758 (95%CI: 0.641-0.875, P = 0.001), and the optimal cut-off value for predicting sTBI complicated with intracranial infection was 175. The calibration curve showed good consistency between the predicted probability and the actual probability, while the Hosmer-Lemeshow goodness of fit test showed no statistically significant difference (χ² = 4.613, P = 4.412), indicating the Nomogram model has good differentiation, calibration and stability. Conclusions: Diabetes, CBF leakage, surgical treatment and severe infection can increase the risk of sTBI complicated with intracranial infection. The Nomogram model can better predict the risk of sTBI complicated with intracranial infection.

Key words: Brain injuries, traumatic, Infections, Risk factors, Logistic model, Nomograms