摘要：

目的 筛查细菌性脑膜炎合并缺血性卒中的危险因素并初步构建风险预测列线图模型。方法 回顾分析2008年6月至2018年6月在空军军医大学西京医院诊断与治疗的176例细菌性脑膜炎患者的基线资料、临床特点、实验室和影像学检查。采用单因素和多因素Logistic回归分析筛查细菌性脑膜炎合并缺血性卒中的危险因素，R软件构建风险预测列线图模型，绘制受试者工作特征（ROC）曲线和校准曲线评价模型的区分度和校准度。结果 176例细菌性脑膜炎患者中15例合并缺血性卒中，发生率约8.52%。Logistic回归分析显示，年龄≥ 55岁（OR=6.350，95% CI：1.750~23.046；P=0.005）、癫发作（OR=5.114，95% CI：1.363~19.193；P=0.016）、神经功能缺损（OR=10.409，95% CI：2.781~39.480；P=0.001）和脑脊液白细胞计数< 1634×10⁶/L （OR=3.538，95% CI：1.014~12.345；P=0.048）是细菌性脑膜炎合并缺血性卒中的危险因素。根据这4项指标构建风险预测列线图模型，细菌性脑膜炎合并缺血性卒中的概率为66.8%。ROC曲线下面积为0.859（95% CI：0.749~0.968，P=0.001），提示模型区分度较好；校准曲线显示模型曲线与理想曲线的趋势较一致，提示模型预测效能较好。结论 初步构建的细菌性脑膜炎合并缺血性卒中的风险预测列线图模型具有良好的区分度和校准度，有一定的临床应用价值，可为早期发现细菌性脑膜炎合并缺血性卒中的高危患者提供线索。

关键词: 脑膜炎, 细菌性, 卒中, 脑缺血, 危险因素, Logistic模型, 模型, 理论, 列线图

Abstract:

Objective To screen the risk factors of bacterial meningitis complicated with ischemic stroke and initially construct a risk prediction nomogram model. Methods A retrospective research analysis was performed for baseline data, clinical characteristics, laboratory or imaging examinations about 176 patients with bacterial meningitis diagnosed and treated in Xijing Hospital, Air Force Military Medical University of Chinese PLA from June 2008 to June 2018. Univariate and multivariate Logistic regression screened the risk factors for bacterial meningitis complicated with ischemic stroke. A prediction nomogram model was established by R software, using receiver operating characteristic (ROC) curve and calibration curve to evaluate the discrimination and calibration of the model. Results Fifteen of the 176 patients with bacterial meningitis complicated with ischemic stroke, the incidence was about 8.52%. Logistic regression analysis showed that age ≥ 55 years (OR=6.350, 95%CI:1.750-23.046; P=0.005), seizures (OR=5.114, 95%CI:1.363-19.193; P=0.016), neurological deficit (OR=10.409, 95%CI:2.781-39.480; P=0.001) and cerebrospinal fluid white blood cell count < 1634×10⁶/L (OR=3.538, 95%CI:1.014-12.345; P=0.048) were risk factors for patients with bacterial meningitis complicated with ischemic stroke. The risk prediction nomogram model was constructed based on the above four indicators, and the probability of bacterial meningitis complicated with ischemic stroke was 66.8%. The area under the ROC curve was 0.859 (95%CI:0.749-0.968, P=0.001), which indicated that the model had excellent performance. The calibration chart showed that the trend of the model curve and the ideal curve was more consistent, which indicated that the model had better prediction performance. Conclusions Prediction of ischemic stroke in patients with bacterial meningitis has an excellent discrimination and calibration based on the currently constructed nomogram model for the risk. This prediction model contributes to the early detection of ischemic stroke in patients with bacterial meningitis, which has clinical significance to make a further study.

Key words: Meningitis, bacterial, Stroke, Brain ischemia, Risk factors, Logistic models, Models, theoretical, Nomograms