基于术后早期颅内压相关参数构建脑室出血患者LTEVD术后预后不良风险预测模型及验证

Construction and validation of a risk prediction model for poor postoperative prognosis in patients with ventricular hemorrhage based on early postoperative intracranial pressure-related parameters

摘要:

目的 探讨脑室出血患者行长皮下隧道脑室外引流(LTEVD)术后早期颅内压(ICP)相关参数与预后的关系,构建并验证预后不良风险预测模型。方法 选取2023年1月-2025年12月167例于我院接受LTEVD术治疗的脑室出血患者为研究对象。术后随访3个月,根据格拉斯哥预后(GOS)评分将患者分为预后良好组(n=95)和预后不良组(n=72)。术后监测ICP参数[ICP、压力反应指数(PRx)、平均颅内压波幅(MWA)、20 mmHg阈值下颅内压剂量(Dicp20)]。采用受试者工作特征曲线(ROC)分析各参数对预后不良的预测价值；经LASSO回归筛选影响因素,通过多因素Logistic回归分析确定独立危险因素；基于独立危险因素构建预后预测模型,并以Bootstrap自抽样法完成内部验证。结果 预后不良组年龄、血肿体积高于预后良好组,术前GCS评分低于预后良好组(P<0.05)；术后早期 ICP、PRx、MWA、Dicp20 均显著高于预后良好组(P<0.05)。ROC曲线显示,PRx预测预后不良的AUC最高(0.825)。LASSO回归筛选出年龄、术前GCS评分、血肿体积、ICP、PRx、MWA、Dicp20 为预后不良影响因素。多因素Logistic回归表明,术前GCS评分降低、血肿体积增大、PRx 升高、Dicp20升高是脑室出血患者LTEVD术后预后不良的独立危险因素(P<0.05)。基于上述4项指标构建预测模型,Hosmer-Lemeshow 检验示拟合优度良好(χ2=6.324,P=0.612)；内部验证后模型AUC为0.859,灵敏度83.56%,特异度80.21%,区分度与稳定性良好。结论 术后早期ICP相关参数可有效评估脑室出血患者LTEVD术后预后,其中PRx预测价值最优；由术前GCS评分、血肿体积、PRx、Dicp20构建的预测模型,能准确识别术后预后不良高风险患者,为临床个体化诊疗提供依据。

Abstract:

Objective To investigate the relationship between early intracranial pressure (ICP)-related parameters and prognosis in patients with intraventricular hemorrhage undergoing long subcutaneous tunnel external ventricular drainage, and to construct and validate a risk prediction model for poor prognosis. Methods A total of 167 patients with intraventricular hemorrhage who underwent long subcutaneous tunnel external ventricular drainage in our hospital from January 2023 to December 2025 were enrolled. After 3 months of follow-up, patients were divided into good prognosis group (n=95) and poor prognosis group (n=72) based on the Glasgow Outcome Scale (GOS) score. Postoperative ICP parameters [ICP, pressure reactivity index (PRx), mean wave amplitude (MWA), and intracranial pressure dose above 20 mmHg (Dicp20)] were monitored. The predictive value of each parameter for poor prognosis was analyzed using receiver operating characteristic (ROC) curves. Influencing factors were screened by LASSO regression, and independent risk factors were identified by multivariate logistic regression. A prognostic prediction model was constructed based on independent risk factors and internally validated using the Bootstrap resampling method. Results The poor prognosis group had significantly higher age and hematoma volume, and lower preoperative GCS score compared to the good prognosis group (P<0.05). Early postoperative ICP, PRx, MWA, and Dicp20 were significantly higher in the poor prognosis group (P<0.05). ROC curve analysis showed that PRx had the highest AUC (0.825) for predicting poor prognosis. LASSO regression identified age, preoperative GCS score, hematoma volume, ICP, PRx, MWA, and Dicp20 as factors influencing poor prognosis. Multivariate logistic regression indicated that decreased preoperative GCS score, increased hematoma volume, elevated PRx, and elevated Dicp20 were independent risk factors for poor prognosis after long subcutaneous tunnel external ventricular drainage in patients with intraventricular hemorrhage (P<0.05). A prediction model based on these four indicators was constructed. The Hosmer-Lemeshow test indicated good goodness of fit (χ2=6.324, P=0.612). After internal validation, the model’s AUC was 0.859, with a sensitivity of 83.56% and specificity of 80.21%, demonstrating good discrimination and stability. Conclusion Early postoperative ICP-related parameters can effectively assess the prognosis of patients with intraventricular hemorrhage undergoing long subcutaneous tunnel external ventricular drainage, with PRx having the best predictive value. The prediction model based on preoperative GCS score, hematoma volume, PRx, and Dicp20 can accurately identify patients at high risk of poor postoperative prognosis, providing a basis for individualized clinical diagnosis and treatment.