基于LASSO-logistic回归建立住院患者MRSA血流感染的预测模型

Establishment of MRSA prediction model of bloodstream infection in inpatients based on LASSO-logistic regression

目的:运用LASSO-logistic回归构建住院患者耐甲氧西林金黄色葡萄球菌(MRSA)血流感染的预测模型并进行效果验证。方法:基于216例血流感染金黄色葡萄球菌患者的数据集建立预测模型。使用21项临床特征分析病例危险因素,采用LASSO回归进行优化特征选择。采用多变量logistic回归分析,结合LASSO回归选择的特征建立预测模型。结果:预测模型中包含的预测危险因素为前3个月外科手术史、前3个月内使用过第三代头孢类药物。预测住院患者血流感染MRSA的危险因素列线图的校准曲线显示出良好的一致性,当C指数为0.779时对血流感染MRSA预测具有较好的提示意义。结论:本研究构建的预测模型具有良好的预测价值,可为临床防治MRSA血流感染提供参考依据。

Objective:To develop a prognostic model for methicillin-resistant staphylococcus aureus(MRSA)bloodstream infections in a hospital setting by employing least absolute shrinkage and selection operator(LASSO)logistic regression,and to evaluate its predictive efficacy.Methods:A prediction model was established utilizing a dataset comprising 216 patients diagnosed with blood influenza caused by Staphylococcus aureus infection.A total of 21 clinical features were used to analyze risk factors,and LASSO regression was used to optimize feature selection.A prediction model was established through multivariate logistic regression analysis,incorporating the variables selected via LASSO regression.Results:The predictive risk factors included a history of surgical procedures within the preceding three months and the administration of third-generation cephalosporins within the same time frame.The calibration curve of the risk factors column chart for predicting MRSA in hospitalized patients shows good consistency,with a C-index of 0.779.Such results suggest that the model has robust predictive utility for predicting MRSA bloodstream infections.Conclusion:The risk prediction model,developed through LASSO logistic regression analysis,demonstrates robust predictive efficacy for MRSA bloodstream infections.