肝硬化患者医院内感染风险预测列线图模型的构建

目的 本研究旨在建立准确、可操作的医院内感染(NI)风险预测列线图模型,将感染风险量化,从而能直观地提示肝硬化患者发生NI的风险。方法 2016年1月～2020年12月我院就诊的肝硬化住院患者,依据多元Logistic回归分析筛选出的NI发生危险因素建立风险预测列线图。采用Bootstrap进行内部验证,应用受试者工作特征曲线(ROC)、校准图、 Hosmer-Lemeshow 检验和决策曲线分析法评估列线图模型的预测效能和临床获益。结果 本研究纳入503例肝硬化患者,其中131例(26.0%)发生NI;多因素Logstic回归模型分析显示存在腹水、侵入性操作、高血小板/淋巴细胞比值 (PLR)和高MELD 评分为肝硬化患者发生NI的独立危险因素;基于上述4个变量建立列线图模型并进行验证,结果ROC曲线下面积(AUC)为0.845,模型诊断效能良好;Hosmer-Lemeshow 检验显示,模型校正曲线与理想曲线拟合良好(P=0.999,P=0.688),该模型具有良好的校准和判别能力,决策曲线分析表明在较大的阈值内具有较高的临床获益。结论 我们建立的列线图可以较准确地预测肝硬化患者发生NI的风险,有助于临床医生及早识别高危患者,为临床干预和优化决策提供依据。

Nomogram model prediction of nosocomial infection in patients with liver cirrhosis: a retrospective Logistic regression analysis

Objective The aim of the present study was attempted to screen out risk factors in cirrhotic patients with nosocomial infections (Nis) and thereby establish an objective and user-friendly risk prediction model. Methods A retrospective analysis of the clinical data of inpatients with liver cirrhosis in our tertiary hospital between January 2016 and December 2020. The univariate and multivariate Logistic regression analyses were applied to screen the possible risk factors and build up a prediction model, which was further developed into a prediction nomogram. The performance of the nomogram model was evaluated by the area under the receiver operating characteristic curve (AUC), calibration diagram and Hosmer-Lemeshow test. The clinical benefit was assessed by using decision curve analysis. Results The nomograms were developed based on the materials of 503 patients with liver cirrhosis, among them, 131 (26.0%) acquired at least one episode of Nis during the hospitalization; the predictive variables screened out by multivariate Logistic regression were the presence of ascites, invasive procedures, platelet/lymphocyte ratio and MELD score; by incorporating these factors, the validation tests showed that the final model had a well-fitted calibration and good discrimination capability with the AUC of 0845, and the Hosmer-Lemeshow test showed that the model calibration curve fitted well with the ideal curve (P=0.999, P=0.688); the analysis of the decision curve demonstrated that the model had a higher net benefit within a larger threshold. Conclusion Our nomogram could accurately predict the risk of Nis in cirrhotic patients, which might help clinicians identify high-risk patients early and provide clinical decision-making basis for intervention and optimization.