[Analysis on prediction power of HIV infection risk assessment tool in men who have sex with men in Guizhou province].

To evaluate the prediction power of HIV infection risk assessment tool and the applicability in MSM in Guizhou province. MSM were recruited through snowball sampling method. Questionnaire surveys were conducted among the MSM using HIV infection risk assessment tool, and combined with HIV serologic test results, the risk prediction power of HIV infection risk assessment tool was evaluated. A total of 3 379 MSM were recruited from January 2018 to December 2019 in Guizhou. The HIV infection rate was 3.3%(111/3 379). The mean risk scores of HIV positive and HIV negative MSM were (12.15±3.08) and (12.07±3.07), respectively. The difference in risk score was significant between MSM with different HIV status (=8.69, <0.001). According to the principle of decision tree, individual risk scores were divided into following three categories: ≤11.96, 11.97-14.80 and >14.80, the HIV infection rate was 0.8%, 4.3% and 8.6% respectively, suggesting that the higher the individual risk score was, the higher the HIV infection rate was (trend =88.18, <0.001). Multivariate logistic regression analysis showed that the higher the individual risk score was, the higher the risk of HIV infection was. Compared to the total score ≤11.96, the a values at total scores of 11.97-14.80 and >14.80 were 6.34 (95%: 3.38-11.88) and 14.07(95%: 7.44-26.61), respectively. The risk of HIV infection in Miao ethnic group was higher than that in Han ethnic group (a=1.83, 95%1.04-3.21), and the risk of HIV infection in those with education level of primary school and below was higher than that in undergraduates or those with education level of junior college and above (a=2.50, 95%1.06-5.88), and the risk of HIV infection was higher in those who had bisexual behaviors than in those who had homosexual behaviors (a=1.95, 95%1.19-3.19). The risk of HIV infection was higher in those who had never received HIV testing (a=1.53, 95%1.01-2.33). The area under the receiver operating characteristic (ROC) curve and area under ROC (AUC) for HIV infection prediction was 0.751 (95%:0.710-0.792, <0.001). The maximum Youden's index was individual risk score of 12.56, and the sensitivity of the risk assessment tool was 0.838, and its specificity was 0.412. The results of HIV infection risk assessment tool in Guizhou indicated that in MSM the higher the individual risk score, the higher the risk of HIV infection is. The tool can be used to evaluate the risk of HIV infection in MSM, but the specificity should be improved.