Application Of Respondent-driven Sampling In Recruiting Community-based Drug Users For HIV Epidemiological Survey In Dongguan, Guangdong Province

Objectives To explore the feasibility of RDS in recruiting community-based drug users for HIV/STI epidemiological survey; to understand the HIV epidemic and its correlates among the community-based drug users; to explore the differences of the HIV prevalence and its correlates between the community-base drug users and the detoxification center-based drug users recruited into the sentinel surveillance to understand the HIV epidemic among the drug users in Dongguan, Guangdong province more accurately; to explore the feasibility of respondent-driven sampling (RDS) with multiplier used to estimate the number of drug users.Methods In 2008 respondent-driven sampling was applied to recruit community-based drug users in Dongguan. A structured questionnaire was used for a face to face interview. Intravenous blood samples were collected for the measurement of HIV, HCV and syphilis antibodies. RDS AT was applied for adjusting the estimates and bivariate analysis. SPSS was used for multivariate analysis. Netdraw was used to draw recruitment tree.Results A total of 303 drug users were recruited within 15 weeks'periods and 300 blood samples were collected; the sample reached the pre-defined sample size and the sample compositions converged to reach equilibrium. After adjusted by RDSAT, the prevalence rates of HIV, HCV and syphilis were 14.7% (95% CI:6.1%-25.4%),63.2%(95% CI:55.7%-71.7%), and 4.7%(95% CI: 1.8%-8.2%) respectively. The rate of injection drug use(IDU) was 83.9% (95% CI: 77.7%-89.6%); of those IDUs, in the last month, the rates of ever injecting drug, sharing needle directly and sharing needle indirectly were 67.3% (95% CI: 57.2%-75.0%),35.7% (95% CI:26.8%-43.8%),27.2%(95% CI:20.7%-34.4%) respectively; 64.4%(95%CI:57.7%-72.7%) of the subjects reported engaging in commercial sex, and 38.7%(95%CI:31.5%-46.7%) reported multiple sexual partners in the past 6 months.Multiple logistic regression model indicated that the following risk factors were associated significantly with an increase risk for HIV seropositivity:Zhuang or Yi ethnicity(AOR:23.64,95%CI:6.80-82.17), needle sharing in the last month(AOR:3.87, 95%CI:1.46-10.28), sexual partners in the past 6 months (AOR:1.10,95%CI: 1.02-1.18); knowledge related to HIV (AOR:0.81,95%CI:0.68-0.98) was independent protective factor of HIV infection.The prevalence of HIV among the drug users recruited into sentinel surveillance in 2008 was 4.0%(16/400), which was significantly lower than that of the community-based drug users (x2=5.84,P=0.022).Based on the compulsive detoxification center, the estimated number of drug users in Dongguan was 52 028 (90%CI:35 196-81590), which was 2.67 (90%CI:1.81-4.19) times the registration number; the estimated number of injection drug users was 43 651(90%CI:29 529-68 454), which accounted for 0.25%(90%CI:0.19%-0.31%) of the population aged 15-64 years.Conclusions RDS is an effective way to recruit community-based drug users for HIV survey. Safe interview site and appropriate incentive are the two most important factors associated with the recruitment efficiency.There are high proportions of needle sharing and risk sexual behaviors among the community-based drug users, while low level of knowledge on HIV. The HIV epidemic among the drug users in Dongguan is rather serious. Dissemination and sharing of information on HIV should be strengthened through media and existing social network among drug users, especially those migrant and minorities. RDS with multiplier can make relatively accurate estimates of the number of drug users and the subgroups.