中国现代神经疾病杂志 ›› 2025, Vol. 25 ›› Issue (4): 332-344. doi: 10.3969/j.issn.1672-6731.2025.04.011

• 临床研究 • 上一篇    下一篇

2 延髓梗死呼吸心跳骤停危险因素分析

邹璇1, 井奚月2, 赵文娟3,*()   

  1. 1. 300350 天津市环湖医院神经内科
    2. 300350 天津市神经外科研究所
    3. 300350 天津医科大学神经内外科及神经康复临床学院 天津市环湖医院神经内科
  • 收稿日期:2024-11-20 出版日期:2025-04-25 发布日期:2025-05-19
  • 通讯作者: 赵文娟
  • 基金资助:
    天津市医学重点学科(专科)建设项目(TJYXZDXK-052B)

Analysis of risk factors of respiratory and cardiac arrest after medullary infarction

Xuan ZOU1, Xi-yue JING2, Wen-juan ZHAO3,*()   

  1. 1. Department of Neurology, Tianjin Huanhu Hospital, Tianjin 300350, China
    2. Tianjin Neurosurgical Institute, Tianjin Huanhu Hospital, Tianjin 300350, China
    3. Clinical College of Neurology, Neurosurgery and Neurorehabilitation, Tianjin Medical University; Department of Neurology, Tianjin Huanhu Hospital, Tianjin 300350, China
  • Received:2024-11-20 Online:2025-04-25 Published:2025-05-19
  • Contact: Wen-juan ZHAO
  • Supported by:
    This study was supported by Tianjin Key Medical Discipline (Specialty) Construction Project(TJYXZDXK-052B)

摘要:

目的: 筛查延髓梗死呼吸心跳骤停的危险因素,并建立风险预测模型。方法: 共纳入2016年1月至2023年1月天津市环湖医院收治的3168例延髓梗死患者,根据是否发生呼吸心跳骤停分为呼吸心跳骤停组(66例)和非呼吸心跳骤停组(3102例)。收集可导致呼吸心跳骤停的潜在危险因素,并利用欠采样、过采样和合成少数类过采样技术(SMOTE)进行重采样处理。将原始数据和重采样数据按照7∶3比例分为训练集和测试集。对于训练集,利用单因素及多因素逐步法Logistic回归分析筛查延髓梗死呼吸心跳骤停的危险因素;利用训练集和测试集绘制受试者工作特征(ROC)曲线,采用Delong检验比较4组Logistic回归模型的ROC曲线下面积,并建立列线图模型。结果: SMOTE重采样法测试后的曲线下面积最大(SMOTE∶原始数据Z=3.254,P=0.000;SMOTE∶欠采样Z=4.385,P=0.000;SMOTE∶过采样Z=2.701,P=0.007)。且SMOTE重采样数据的Logistic回归分析显示,年龄增长(OR=1.045,95% CI:1.021~1.070;P=0.000)、有吸烟史(OR=22.216,95% CI:10.426~49.920;P=0.000)、吸烟量小(OR=0.943,95% CI:0.915~0.971;P=0.000)、有饮酒史(OR=1.847,95% CI:1.068~3.207;P=0.028)、脑血管病病史(OR=3.104,95% CI:1.842~5.344;P=0.000)、高密度脂蛋白胆固醇(HDL-C)水平高(OR=5.863,95% CI:2.063~16.725;P=0.000)、纤维蛋白原水平高(OR=1.413,95% CI:1.381~1.702;P=0.001)、左侧延髓外侧梗死[无延髓内侧梗死(OR=0.173,95% CI:0.093~0.312;P=0.000),无右侧延髓外侧梗死(OR=0.337,95% CI:0.176~0.634;P=0.001)]、合并延髓外梗死(OR=31.354,95% CI:17.496~59.163;P=0.000)、入院时洼田饮水试验评分高(OR=3.723,95% CI:2.913~4.862;P=0.000)、合并应激性溃疡(OR=5.266,95% CI:2.902~9.813;P=0.000)为延髓梗死呼吸心跳骤停的危险因素。列线图显示,洼田饮水试验评分的预测作用最大,饮酒史的预测作用最小。结论: 年龄增长、高水平HDL-C和纤维蛋白原、有吸烟史、吸烟量小、有饮酒史、脑血管病病史、左侧延髓外侧梗死、合并延髓外梗死、高洼田饮水试验评分及合并应激性溃疡为延髓梗死呼吸心跳骤停的危险因素。利用列线图可以直观地预测延髓梗死呼吸心跳骤停的发生率。

关键词: 脑干梗死, 延髓, 心脏停搏, 危险因素, Logistic模型, 列线图

Abstract:

Objective: To identify the risk factors of respiratory and cardiac arrest after medullary infarction (MI), and to establish a Nomogram model of respiratory and cardiac arrest after MI. Methods: Total of 3168 patients with MI hospitalized in Tianjin Huanhu Hospital from January 2016 to January 2023 were included, including 66 patients in the respiratory and cardiac arrest group, and 3102 patients in the non-respiratory and cardiac arrest group. Potential risk factors of respiratory and cardiac arrest were collected, and samples were resampled using random undersampling (RUS), random oversampling (ROS), and synthetic minority over-sampling technique (SMOTE). Split the raw data and resampled data into training and testing sets. For the training set, univariate and multivariate stepwise Logistic regression models were used to analyze the risk factors of respiratory and cardiac arrest after MI. Drawn receiver operating characteristic (ROC) curve using the training and testing sets, compared the area under the curve (AUC) of 4 Logistic regression models using Delong test, and established a Nomogram model. Results: Use the testing sets to test the Logistic regression models built on the raw data and 3 resampling methods. The results showed that the AUC of SMOTE resampling was the highest after testing (SMOTE∶raw data Z = 3.254, P = 0.000; SMOTE∶RUS Z = 4.385, P = 0.000; SMOTE∶ROS Z = 2.701, P = 0.007). For SMOTE resampling data, age increase (OR = 1.045, 95%CI: 1.021-1.070; P = 0.000), smoking history (OR = 22.216, 95%CI: 10.426-49.920; P = 0.000), the smaller the number of cigarettes smoked (OR = 0.943, 95%CI: 0.915-0.971; P = 0.000), alcohol history (OR = 1.847, 95%CI: 1.068-3.207; P = 0.028), cerebrovascular history (OR = 3.104, 95%CI: 1.842-5.344; P = 0.000), the higher the high-density lipoprotein cholesterol (HDL-C; OR = 5.863, 95%CI: 2.063-16.725, P = 0.000), the higher fibrinogen (FIB; OR = 1.413, 95%CI: 1.381-1.702; P = 0.001), left lateral medullary infarction [LMI; no medial medullary infarction (MMI; OR = 0.173, 95%CI: 0.093-0.312, P = 0.000), no right LMI (OR = 0.337, 95%CI: 0.176-0.634; P = 0.001)], combined with extramedullary infarction (OR = 31.354, 95%CI: 17.496-59.163; P = 0.000), higher Wada Drinking Water Test score (OR = 3.723, 95%CI: 2.913-4.862; P = 0.000), and patients with stress ulcer (OR = 5.266, 95%CI: 2.902-9.813; P = 0.000) were more likely to experience respiratory and cardiac arrest after MI. The Nomogram model showed that the Wada Drinking Water Test score had the greatest predictive effect, while the predictive effect of drinking history was the smallest. Conclusions: Increasing age, high HDL-C, high FIB, smoking history, the smaller the number of cigarettes smoked, alcohol history, cerebrovascular history, left LMI, combined with extramedullary infarction, high Wada Drinking Water Test score and combined with stress ulcer are risk factors for respiratory and cardiac arrest after MI. The Nomogram model can be used to intuitively predict the probability of occurrence of respiratory and cardiac arrest after MI.

Key words: Brain stem infarctions, Medulla oblongata, Heart arrest, Risk factors, Logistic models, Nomograms