Abstract：
We aimed to explore the respiratory tract infection after oral and maxillofacial surgery under general anesthesia and related factors.
A total of 494 patients receiving oral and maxillofacial surgery under general anesthesia with tracheal intubation were assigned to a non-infection group (n=469) and an infection group (n=25). Another 494 healthy people undergoing physical examination in the same period were enrolled to establish a classification tree model. The distribution of pathogens, drug resistance of main pathogens, and related influencing factors of postoperative respiratory tract infection were analyzed. The influencing factors of respiratory tract infection were screened by logistic regression analysis. After construction of the classification and regression tree (CART) model based on the influencing factors, the accuracy was evaluated by plotting receiver operating characteristic (ROC) curve.
Pseudomonas aeruginosa was highly resistant to cefazolin and more sensitive to cefoperazone, ciprofloxacin, norfloxacin and imipenem. Staphylococcus aureus was highly resistant to gentamicin and more sensitive to vancomycin. Age ≥ 60 years old, history of lung diseases, operation time ≥ 4 h, anesthesia ventilation time ≥ 120 min, and orotracheal intubation were independent influencing factors of respiratory tract infection (p< 0.05). The results of the gain chart, index map, and Risk value indicated a high predictive value of the CART model for the risk of postoperative respiratory tract infection. The area under the ROC curve was 0.869 [95% confidence interval: 0.795-0.947].
The CART model has a high predictive value and may reduce the risk of postoperative infection.
A total of 494 patients receiving oral and maxillofacial surgery under general anesthesia with tracheal intubation were assigned to a non-infection group (n=469) and an infection group (n=25). Another 494 healthy people undergoing physical examination in the same period were enrolled to establish a classification tree model. The distribution of pathogens, drug resistance of main pathogens, and related influencing factors of postoperative respiratory tract infection were analyzed. The influencing factors of respiratory tract infection were screened by logistic regression analysis. After construction of the classification and regression tree (CART) model based on the influencing factors, the accuracy was evaluated by plotting receiver operating characteristic (ROC) curve.
Pseudomonas aeruginosa was highly resistant to cefazolin and more sensitive to cefoperazone, ciprofloxacin, norfloxacin and imipenem. Staphylococcus aureus was highly resistant to gentamicin and more sensitive to vancomycin. Age ≥ 60 years old, history of lung diseases, operation time ≥ 4 h, anesthesia ventilation time ≥ 120 min, and orotracheal intubation were independent influencing factors of respiratory tract infection (p< 0.05). The results of the gain chart, index map, and Risk value indicated a high predictive value of the CART model for the risk of postoperative respiratory tract infection. The area under the ROC curve was 0.869 [95% confidence interval: 0.795-0.947].
The CART model has a high predictive value and may reduce the risk of postoperative infection.
摘要：
目的探讨口腔颌面手术全麻术后呼吸道感染情况及相关因素。
共有494例患者在气管插管全身麻醉下接受口腔颌面手术，分为非感染组（n=469）和感染组（n=25）。另纳入同期健康体检者494人,建立分类树模型。病原体的分布,主要病原菌耐药性,分析术后呼吸道感染的相关影响因素。采用logistic回归分析筛选呼吸道感染的影响因素。在构建了基于影响因素的分类回归树(CART)模型后,通过绘制受试者工作特征（ROC）曲线来评估准确性。
铜绿假单胞菌对头孢唑林耐药率较高，对头孢哌酮较敏感，环丙沙星,诺氟沙星和亚胺培南.金黄色葡萄球菌对庆大霉素耐药率较高,对万古霉素较敏感。年龄≥60岁，肺部疾病史,运行时间≥4h,麻醉通气时间≥120min,气管插管是呼吸道感染的独立影响因素(p<0.05)。增益图的结果，索引地图，和Risk值表明CART模型对术后呼吸道感染风险具有较高的预测价值。ROC曲线下面积为0.869[95%置信区间：0.795-0.947]。
CART模型具有很高的预测价值，可能会降低术后感染的风险。
共有494例患者在气管插管全身麻醉下接受口腔颌面手术，分为非感染组（n=469）和感染组（n=25）。另纳入同期健康体检者494人,建立分类树模型。病原体的分布,主要病原菌耐药性,分析术后呼吸道感染的相关影响因素。采用logistic回归分析筛选呼吸道感染的影响因素。在构建了基于影响因素的分类回归树(CART)模型后,通过绘制受试者工作特征（ROC）曲线来评估准确性。
铜绿假单胞菌对头孢唑林耐药率较高，对头孢哌酮较敏感，环丙沙星,诺氟沙星和亚胺培南.金黄色葡萄球菌对庆大霉素耐药率较高,对万古霉素较敏感。年龄≥60岁，肺部疾病史,运行时间≥4h,麻醉通气时间≥120min,气管插管是呼吸道感染的独立影响因素(p<0.05)。增益图的结果，索引地图，和Risk值表明CART模型对术后呼吸道感染风险具有较高的预测价值。ROC曲线下面积为0.869[95%置信区间：0.795-0.947]。
CART模型具有很高的预测价值，可能会降低术后感染的风险。
