三种不同算法模型对广州市青年学生艾滋病暴露后预防使用意愿影响因素研究  

Study on the influence factors of willingness to take post‐exposure prophylaxis among young students in Guang‐zhou employing three different algorithm models

作　　者：刘珺[1] 林鹏 徐慧芳 李艳[1] 付笑冰[1] 姚芷潞 谢仕兰 何思敏 黎健荣 潘丝媛 杨放[1] LIU Jun;LIN Peng;XU Huifang;LI Yan;FU Xiaobing;YAO Zhilu;XIE Shilan;HE Simin;LI Jianrong;PAN Siyuan;YANG Fang(Guangdong Provincial Center for Disease Control and Prevention,Guangzhou 511430,China;Guangdong Association of STD&AIDS Prevention and Control)

机构地区：[1]广东省疾病预防控制中心,广东广州511430 [2]广东省性病艾滋病防治协会

出　　处：《华南预防医学》2023年第10期1273-1279,共7页South China Journal of Preventive Medicine

摘　　要：目的分析广州市青年学生艾滋病暴露后预防(PEP)使用意愿及影响因素。方法采用横断面研究对广州市5所不同类型高校学生于2021年9—11月开展艾滋病PEP使用意愿的调查。以PEP使用意愿作为预测变量分别构建logistic回归模型、决策树算法模型和随机森林算法模型,ROSE算法用于处理数据不平衡问题。通过AUC(ROC曲线下面积)和混淆矩阵对3种模型预测性能进行评价。结果共回收有效问卷7346份。67.63%调查对象知道PEP;6794例(92.49%)表示愿意使用,552例(7.51%)表示不愿意。综合3种模型结果,PEP的使用意愿受个体因素(性别、年龄)、学校因素(学校类型、专业)、HIV相关因素(HIV知识知晓情况、检测知识、态度、接受HIV教育学段、性行为)、经济条件等因素影响。logistic回归、决策树和随机森林模型预测性能AUC(95%CI)分别为0.77(0.75~0.79)、0.74(0.72~0.76)和0.90(0.89~0.92)。随机森林算法模型的各指标均优于logistic回归模型和决策树模型。结论广州市青年学生对PEP的知晓仍有待进一步加强,PEP使用意愿主要受PEP知晓情况、个体因素、学校因素、HIV相关因素、经济条件等影响。随机森林算法模型对于在该人群中预测PEP的使用意愿具有适用性。Objective To analyze the willingness and influencing factors of taking post‐exposure prophylax‐is(PEP)among young students in Guangzhou.Methods A cross‐sectional study was conducted among five universities in Guangzhou from September to November 2021.Using PEP as predictive variable,logistic regression model,Decision Tree algorithm model,as well as Random Forest algorithm model were constructed respectively.ROSE algorithm was used to handle data imbalance problems.Evaluated the predictive performance of the three models through AUC(area under ROC curve)and confusion matrix.Results A total of 7346 valid questionnaires were collected,during which 67.63%re‐ported ever heard of PEP,92.49%reported willing to take PEP and 7.51%reported unwilling.Based on the results of the three models,the willingness to take PEP was affected by individual factors(gender,age),school factors(school type,major),HIV‐related factors(HIV knowledge,testing knowledge,attitude of testing,acceptance period of HIV educa‐tion,sexual behavior),and economic conditions etc.The predictive performance AUC(95%CI)for logistic regression,Decision Tree,and Random Forest model were 0.77(0.75-0.79),0.74(0.72-0.76),and 0.90(0.89-0.92),respectively,among which Random Forest algorithm model showed the best prediction than the other two models.Conclusions Knowl‐edge of PEP among young students in Guangzhou still need to be strengthened.The willingness to take PEP is mainly affected by PEP knowledge,individual factors,school factors,HIV‐related factors,economic conditions and so on.The Ran‐dom Forest algorithm model is suitable for predicting the willingness to take PEP among young students.

关 键 词：艾滋病 暴露后预防 青年学生 随机森林算法 决策树 

分 类 号：R179[医药卫生—妇幼卫生保健] R512.91[医药卫生—公共卫生与预防医学]