Jiang Yan, Wang Hong-Wei, Tian Fang-Ying, Guo Yue, Wang Xiu-Mei
Department of Nursing, Shanxi Bethune Hospital, Shanxi Academy of Medical Sciences, Third Hospital of Shanxi Medical University, Tongji Shanxi Hospital, Taiyuan, 030032, China.
Shanxi Medical University, Taiyuan, 030000, China.
Sci Rep. 2025 Jun 5;15(1):19737. doi: 10.1038/s41598-025-04028-x.
The prevalence and spread of carbapenem-resistant Pseudomonas aeruginosa (CRPA) is a global public health problem. This study aims to identify the risk factors of CRPA infection and construct a machine learning model to provide a prediction tool for clinical prevention and control. A total of 1949 patients with P.aeruginosa health care-associated infections (HAIs) were enrolled in this study. A total of 89 patients with CRPA infection and 89 patients with CSPA infection were matched 1:1. LASSO regression was used to screen the variables, and the XGBoost model was established (136 cases in the training set and 60 cases in the test set). Shapley additive explain (SHAP) method was used to explain the importance of variables. The area under the ROC curve (AUC) and calibration curve were used to evaluate the performance of the model. There were 89 cases of CRPA infection, and the CRPA infection rate was 4.57%. Respiratory tract was the most common source of infection, and ICU and hematology department were the high-risk departments. The AUC value of the XGBoost machine learning model in the training set was 0.987 (95%CI: 0.974-1.000), and the AUC value in the test set was 0.862 (95%CI: 0.750-0.974). The clinical decision curve also showed good results of the model. SHAP results showed that ICU admission, duration of central venous catheterization, use of carbapenems and fluoroquinolones were important factors for predicting CRPA infection. The XGBoost machine learning model is helpful for the early prevention and screening of CRPA infection in medical institutions. Infection control and clinical departments should carry out effective prevention and control for high-risk factors to reduce the occurrence of CRPA infection.
耐碳青霉烯类铜绿假单胞菌（CRPA）的流行和传播是一个全球性的公共卫生问题。本研究旨在确定CRPA感染的危险因素，并构建一个机器学习模型，为临床预防和控制提供预测工具。本研究共纳入1949例铜绿假单胞菌医疗保健相关感染（HAIs）患者。将89例CRPA感染患者和89例CSPA感染患者进行1:1匹配。采用LASSO回归筛选变量，建立XGBoost模型（训练集136例，测试集60例）。采用Shapley加法解释（SHAP）方法解释变量的重要性。采用ROC曲线下面积（AUC）和校准曲线评估模型性能。CRPA感染89例，CRPA感染率为4.57%。呼吸道是最常见的感染源，ICU和血液科是高危科室。XGBoost机器学习模型在训练集的AUC值为0.987（95%CI：0.974-1.000），在测试集的AUC值为0.862（95%CI：0.750-0.974）。临床决策曲线也显示了该模型的良好结果。SHAP结果显示，入住ICU、中心静脉置管时间、碳青霉烯类和氟喹诺酮类药物的使用是预测CRPA感染的重要因素。XGBoost机器学习模型有助于医疗机构对CRPA感染进行早期预防和筛查。感染控制和临床科室应对高危因素进行有效防控，以减少CRPA感染的发生。
