高血压脑出血微创穿刺引流术后并发颅内感染的临床预测模型建立和验证

目的研究基于临床资料建立并验证高血压脑出血（HICH）微创穿刺引流术后并发颅内感染的临床预测模型。方法纳入采用微创穿刺引流术治疗的231例HICH患者。遵循术后有无并发颅内感染，将患者分成感染组和非感染组；分析比较与HICH穿刺引流术后并发颅内感染可能相关的16项因素。采用套索回归算法（LASSO）实现特征选择和发现预测因子，多元逻辑回归（Logistic）分析建立预测模型，并以列线图显示。对于模型的识别能力，通过受试者工作特征（ROC）曲线实现；而校准能力，则通过校准曲线来评估。决策曲线分析（DCA）用以评价预测模型的临床适用性。结果LASSO回归筛选出4个预测因子（双根引流管、脑脊液漏、引流时间、尿激酶注射次数），纳入Logistic回归并形成列线图。经过验证，模型的ROC曲线下面积为0.779；校准曲线显示预测概率与实际概率有很好的一致性；DCA则显示模型在一定范围内的临床适用性。结论本研究建立的HICH微创穿刺术后并发颅内感染的预测模型具有良好的区分度、拟合优度和临床适用性，有利于术后可能出现颅内感染的高风险患者的早期识别和针对性治疗。

Objective To establish and verify the prediction model of intracranial infection after minimally invasive puncture drainage for hypertensive intracerebral hemorrhage（HICH） based on clinical data. Methods A total of 231 patients with HICH treated by minimally invasive puncture drainage were included. According to postoperative intracranial infection occurrence, patients were divided into infected group and non-infected group. 16 factors that may be associated with intracranial infection after puncture and drainage of HICH were analyzed and compared. The least absolute shrinkage and selection operator(LASSO) regression was used to practice features selection and discover predictors. The multivariate logistic regression was used to establish the prediction model and constructe nomogram. Receiver operating characteristic(ROC) curve and calibration curve were respectively used to detect the discrimination and degree of fitting of the model. Decision curve analysis(DCA) was used to evaluate the clinical applicability of the prediction model. Results Four predictors(double drainage tubes, cerebrospinal fluid leakage, drainage duration, and urokinase injection times) were screened by LASSO regression, which were incorporated into logistic regression and mapped. It has been verified that the area under the curve of ROC was 0.779. The calibration curve showed that the predicted probability was in good agreement with the actual probability. DCA showed the clinical applicability of the model in a certain range. ConclusionThe prediction model of intracranial infection after minimally invasive puncture for HICH established in this study has good differentiation, degree of fitting and clinical applicability, which is conducive to the early identification and targeted treatment of high-risk patients with postoperative intracerebral infection.