大学生MSM人群HIV感染风险预测模型建立及效果验证研究

目的 建立并验证大学生男男性行为者(men who have sex with men,MSM)的HIV感染风险预测模型。 方法 利用我国7大城市(上海、南京、长沙、郑州、济南、沈阳和昆明)MSM横断面调查数据库中在校大学生MSM相关信息,通过单因素和多因素logistic回归建立模型并采用Stepwise逐步回归法筛选模型变量;采用C-statistics检验、Hosmer-Lemeshow检验(H-L检验)对模型进行判别能力和拟合效果的内部验证和评价;采用Bootstrap重抽样技术进行模型内部验证。 结果 共纳入742名大学生MSM参加研究,其HIV感染率为5.9%(44/742)。经logistic回归模型拟合并通过stepwise法筛选,以下变量进入最终模型:有性传播疾病感染史(OR=5.4,95%CI:2.5~11.4,P<0.01)(5分);近半年使用过新型毒品(OR=2.2,95%CI:1.2~4.2,P=0.016)(2分);最近一次男男性行为方式为被插入方或同时为插入方和被插入方(OR=2.3,95%CI:1.0~5.0,P=0.044)(2分)。模型的C-statistic检验的统计量AUC为0.71,H-L检验的卡方值为1.35(P=0.717);各变量在1 000个Bootstrap样本中出现的频率均>50%。 结论 大学生MSM人群HIV感染风险预测模型具有较好的判别能力和拟合效果,有助于大学生MSM人群比较客观地评估其自身HIV感染风险,以及促进其参与后续HIV检测等防控活动。

Establishment of a predictive model for HIV infection risk among college student MSM and validation of its performance

Objective To develop and validate a predictive model for HIV infection risk in college student MSM.Methods The information related to college student MSM was collected from the database of MSM cross-sectional survey in 7 cities(including Shanghai,Nanjing,Changsha,Zhengzhou,Jinan,Shenyang and Kunming).Univariate and multivariate logistic regression was used to develop the model,and a stepwise procedure was used to select variables.C-statistics test was employed to assess the model discrimination,and Hosmer–Lemeshow test was used to estimate the model calibration.Bootstrap resampling technology was used for model validation.Results A total of 742 college student MSM were enrolled in this study,and the prevalence rate of HIV was 5.9%(44/742).According to logistic regression and stepwise procedure,the variables and their scores in the predictive model for HIV infection risk among college student MSM were having a history of infection with sexual transmitted diseases(OR=5.4,95%CI:2.5-11.4,P<0.01)(5 points),new-type drug abuse in the past six months(OR=2.2,95%CI:1.2-4.2,P=0.016)(2 points),and practicing receptive or both insertive and receptive anal intercourse in the most recent homosexual behavior(OR=2.3,95%CI:1.0-5.0,P=0.044)(2 points).The area under the curve(AUC)of C-statistics was 0.71,and the Chi-square value of Hosmer-Lemeshow test was 1.35(P=0.717).The frequency of each variable in the final model appeared in1,000 Bootstrap samples were all greater than 50%.Conclusions The said predictive model for HIV infection risk among college student MSM has better judgment ability and fitting effect.It is conducive to objectively self-evaluating the risk for HIV infection among college student MSM and promoting them to participate in HIV detection and other prevention and control activities.