艾滋病高危人群对暴露后预防使用意愿的影响因素研究  

Analysis of influencing factors of willingness to post-exposure prophylaxis among high risk populations for HIV

作　　者：李传玺 林玉玺 王霖 李浩初 焦珂笛 马伟 LI Chuanxi;LIN Yuxi;WANG Lin;LI Haochu;JIAO Kedi;MA Wei(School of Public Health,Cheeloo College of Medicine,Shandong University,Jinan 250012,China)

机构地区：[1]山东大学齐鲁医学院公共卫生学院流行病学系,济南250012

出　　处：《中国艾滋病性病》2021年第10期1096-1101,共6页Chinese Journal of Aids & STD

基　　金：国家重点研发计划(2017YFE0103800);国家科技重大专项(2018ZX10721102-003-004)。

摘　　要：目的利用三种分类算法分析艾滋病高危人群对暴露后预防(PEP)使用意愿的影响因素,并进行预测及预测效果评价。方法采用方便抽样,在中国11个城市招募五类艾滋病高危人群2002人,通过电子问卷收集相关信息,利用Logistic回归、决策树和随机森林分析PEP使用意愿的影响因素,并对验证集进行预测,比较三者的预测性能。结果在五类高危人群中,MSM的PEP使用意愿最高,为82.6%,暗娼的PEP使用意愿最低,为42.8%。Logistic回归结果显示,与MSM相比,吸毒人群(OR=0.472,95%CI:0.306~0.723)和暗娼(OR=0.497,95%CI:0.306~0.804)的PEP使用意愿较低,此外,受教育程度低、艾滋病风险认知高、PEP认知水平低、饮酒、近1个月未发生无套性行为、近1年寻求过艾滋病咨询服务者对PEP的使用意愿较低(P<0.05)。决策树模型结果显示PEP认知、受教育程度、艾滋病风险认知和人群类别是影响PEP使用意愿的主要因素。随机森林结果显示PEP使用意愿的主要影响因素按重要性排序依次为人群类别、PEP认知、艾滋病风险认知和受教育程度。三种分类算法均较好地预测了PEP使用意愿,Logistic回归、决策树和随机森林的预测准确率分别为0.763、0.743和0.810,随机森林的预测效果最佳。结论三种分类算法在预测艾滋病高危人群对PEP使用意愿上表现出较好的准确性。应加强艾滋病高危人群的健康教育,尤其是吸毒人群和暗娼,提高对PEP的认知水平,将有助于PEP使用。Objective To identify the influencing factors of willingness to use post-exposure prophylaxis(PEP)in high-risk populations for HIV using three classification algorithms,to predict the willingness to PEP,and to compare the prediction performances of the three algorithms.Methods A total of 2002 participants from five HIV/AIDS high-risk populations were recruited in 11 cities of China by convenience sampling.Electronic questionnaires were used to collect information.Logistic regression,decision tree,and random forest were used to identify the influencing factors of the willingness of PEP and to predict the willingness of PEP on the validation set,and then the prediction performances were compared.Results Among the five high-risk populations,MSM had the highest willingness to use PEP(82.6%),while female sex workers had the lowest willingness to use PEP(42.8%).Logistic regression model showed that compared with MSM,drug users(OR=0.472,95%CI:0.306-0.723)and female sex workers(OR=0.497,95%CI:0.306-0.804)had lower PEP willingness.In addition,people with lower education levels,higher HIV/AIDS risk perception,lower PEP awareness,drinking,no condomless sex in the past month,seeking HIV/AIDS counseling service in the past year were less likely to use PEP(P<0.05).The results of the decision tree model showed that PEP awareness,education level,HIV/AIDS risk perception,and population type were the main factors influencing PEP willingness.The results of the random forest showed that the main influencing factors of PEP willingness in order of importance were population type,PEP awareness,HIV/AIDS risk perception,and education level.Three classification algorithms predicted PEP willingness well.The prediction accuracy rates of logistic regression,decision tree,and random forest were 0.763、0.743 and 0.810 respectively,and the prediction performance of random forest was the best.Conclusion The three classification algorithms showed good accuracy in predicting the willingness to use PEP among high-risk populations of HIV.Health edu

关 键 词：艾滋病 暴露后预防 LOGISTIC回归 决策树 随机森林 

分 类 号：R512.91[医药卫生—内科学]