基于peg-344基因的CRKP感染预后列线图预测模型的构建与验证

Construction and Validation of a Prognostic Nomogram for Carbapenem-Resistant Klebsiella pneumoniae Infection Based on the peg-344 Gene

摘要: 目的：探索碳青霉烯类耐药肺炎克雷伯菌(carbapenem-resistant Klebsiella pneumoniae, CRKP)感染患者临床结局的影响因素，并构建列线图预测模型。方法 收集安徽省某三甲医院2020年8月至2024年8月内所有的CRKP感染病例，以8:2的比例随机拆分为训练集(122例)和验证集(31例)，根据患者临床结局将训练集和验证集分为预后良好组和预后不良组，使用Logistic回归分析得到关联强度较高的危险因素，将其纳入多因素逻辑回归分析，并构建风险预测列线图，利用受试者工作特征(Receiver Operating Characteristic, ROC)曲线、校准曲线评估预测模型对住院患者CRKP感染临床结局的预测效能。结果：分析显示神经系统疾病、入住重症监护室、败血症或休克以及CRKP菌株peg-344基因阳性携带为CRKP感染患者预后不良的独立影响因素，分析得到相应的列线图模型。模型的训练集ROC曲线下面积(Area Under the Curve, AUC)为0.770 (95%CI: 0.687~0.854)，验证集ROC曲线下面积为0.717 (95%CI: 0.527~0.907)，校准曲线显示模型校准度良好。结论：根据模型评估及图表数据，该CRKP感染患者预后的列线图预测模型具有较好的检验效能及拟合优度，可有效预测CRKP感染患者出现不良预后的风险，有助于尽早采取相关临床干预措施。

Abstract: Objective: To explore the factors influencing the clinical outcomes of patients infected with carbapenem-resistant Klebsiella pneumoniae (CRKP) and to construct a nomogram prediction model. Methods: We collected all CRKP infection cases from a tertiary hospital in Anhui Province from August 2020 to August 2024. The cases were randomly split into a training set (122 cases) and a validation set (31 cases) in an 8:2 ratio. Based on the clinical outcomes of the patients, the training and validation sets were divided into good and poor prognosis groups. Logistic regression analysis was used to identify risk factors with high association strength, which were then included in a multivariate logistic regression analysis to construct a risk prediction nomogram. The prediction model’s performance in predicting the clinical outcomes of CRKP infections in hospitalized patients was assessed using receiver operating characteristic (ROC) curves and calibration curves. Results: The analysis revealed that neurological diseases, admission to the intensive care unit, sepsis or shock, and the presence of the peg-344 gene in CRKP strains were independent factors associated with poor prognosis in CRKP-infected patients. The corresponding nomogram model was developed. The area under the ROC curve (AUC) for the training set was 0.770 (95%CI: 0.687~0.854), and for the validation set, it was 0.717 (95%CI: 0.527~0.907). The calibration curve showed good model calibration. Conclusion: Based on model evaluation and graphical data, the nomogram prediction model for CRKP infection prognosis demonstrated good discrimination and calibration, effectively predicting the risk of poor outcomes in CRKP-infected patients and aiding in the early implementation of relevant clinical interventions.