基于LASSO logistic回归的早泄患者预测模型  

A predictive model for premature ejaculation patients based on LASSO logistic regression

作　　者：张杨 张磊 袁建林 陈长生 ZHANG Yang;ZHANG Lei;YUAN Jianlin;CHEN Changsheng(Ministry-of-Education Key Laboratory of Hazard Assessment and Control in Special Operational Environment,Department of Military Health Statistics,Faculty of Military Preventive Medicine,Air Force Medical University,Xi’an,Shaanxi 710032,China;Xijing Hospital of Air Force Medical University,Xi’an,Shaanxi 710032,China)

机构地区：[1]空军军医大学军事预防医学系军队卫生统计学教研室,特殊作业环境危害评估与防治教育部重点实验室,陕西西安710032 [2]空军军医大学西京医院,陕西西安710032

出　　处：《实用预防医学》2023年第12期1470-1475,共6页Practical Preventive Medicine

基　　金：国家自然科学基金面上项目(61971425、82073663);西京医院学科助推计划(XJZT18D05)。

摘　　要：目的探讨早泄患者相关影响因素并建立基于LASSO logistic回归的早泄预测模型。方法基于西京医院等5家医院门诊招募的男性受试者,通过问卷及量表评分结果,构建基于LASSO logistic回归的早泄患者预测模型,通过交叉验证法选择最优调和系数λ,采用赤池信息准则(Akaike information criterion,AIC)和贝叶斯信息准则(Bayesian information criterion,BIC)与全变量logistic回归和逐步logistic回归进行比较,基于曲线下面积(area under curve,AUC)和校准曲线分别评价模型的区分度和准确度,并绘制列线图。结果本研究共纳入3180例受试者,其中早泄组有2663例(83.7%),非早泄组有517例(16.3%)。LASSO logistic回归模型(λ=0.004),纳入的自变量为:年龄、居住地、职业、IIEF-5评分、PEDT评分和GAD-7评分;AIC=2240.2,BIC=2282.7,均低于全变量logistic回归(2262.9/2292.2)和逐步logistic回归(2257.3/2293.7);ROC曲线分析LASSO logistic回归模型的预测价值,其AUC为0.799,显著高于全变量logistic回归和逐步logistic回归模型,差异均有统计学意义(P<0.05)。校准曲线证实列线图模型具有较高的预测准确度。结论利用年龄、职业、居住地、IIEF-5评分、PEDT评分、GAD-7评分,基于LASSO logistic回归建立无创列线图模型作为临床诊断早泄的量化工具,具有较高的诊断效能,值得推广应用。Objective To explore the related factors affecting premature ejaculation(PE)patients,and to establish a predictive model for PE based on LASSP logistic regression.Methods A predictive model for PE based on LASSO logistic regression was constructed through the results of questionnaire surveys and scale scores of male subjects recruited from outpatient departments of 5 hospitals like Xijing Hospital.The cross validation method was used to select the coefficientλ.Akaike information criterion(AIC)and Bayesian information criterion(BIC)were employed to evaluate the performance of LASSP logistic model,and the evaluation results were compared with the parameters of full-variable logistic regression and stepwise logistic regression.Area under curve and calibration curve were applied to evaluating the discriminability and accuracy of the model,and the nomogram was drawn.Results A total of 3,180 subjects were enrolled into this study,with 2,663(83.7%)cases in the PE group,517(16.3%)cases in the non-PE group.Theλselected by cross validation was 0.004.The independent variables included age,residence,occupation,IIEF-5 score,PEDT score and GAD-7 score.The values of AIC and BIC were 2,240.2 and 2,282.7 respectively,which were lower than those of full-variable logistic regression(2,262.9/2,292.2)and stepwise logistic regression(2,257.3/2,293.7).The ROC curve was used to analyze the predictive value of LASSO logistic regression model,and the results showed an area under the ROC curve of 0.799,which was significantly higher than those of full-variable logistic regression and stepwise logistic regression(P<0.05).Calibration curve confirmed that the nomogram model had high accuracy of prediction.Conclusion The noninvasive nomogram model based on LASSO logistic regression is established by using parameters including age,residence,occupation,IIEF-5 score,PEDT score and GAD-7 score,and as a quantitative tool for clinical diagnosis of PE,it has a high diagnostic efficiency and thus holds promise for clinical application.

关 键 词：早泄 危险因素 预防 LASSO logistic回归模型 

分 类 号：R698[医药卫生—泌尿科学]