For the two cross-sectional serosurvey studies that were conducted during July-November of 2005 [2] and between November 2006-February 2007 [5-6] in the same districts in Beijing, study participants were recruited in three following ways: (i) through advertisements at the NCAIDS website (www.chinaids.org.cn) and a nongovernmental AIDS volunteer group (www.hivolunt.net); (ii) 15 peer recruiters were hired and trained to reach out to clubs, bars, parks, and bath houses frequented by the MSMs and to distribute flyers with study-related information to recruit volunteers for this study; (iii) the study participants were encouraged to refer their peers to participate in the study. All potential participants came to a district HIV testing and counseling clinic in downtown Beijing for eligibility assessment. Hence the sampling is partly convenient and partly voluntary. Eligibility criteria included self-reported same-gender sex in the past 6 months, Beijing residency, and a willingness to provide written informed consent.
Written informed consent was obtained from all study participants before they were interviewed.
Those who met the screening criteria then completed an HIV/STD risk assessment interview, received HIV pretest and risk-reduction counseling, and had blood drawn to test for HIV and syphilis antibodies. Participants were also given HIV post-test counseling when they subsequently returned for their HIV test results. A summary of distribution of recruitment methods of all survey participants is given in Table 1.
The study protocol and informed consent form were approved by the institutional review board (IRB) of the NCAIDS. For a detailed description of the data collection and laboratory analysis, as well as the details of the 2005 survey, the readers are referred to [2]. Similarly, results of the second survey, using similar sampling procedures and conducted in the same areas in Beijing, are given in [5-6].
For the present study, the HIV serotest results from these two serosurveys are given in Table 2. Note that, for the second sampling period (11/2006-2/2007), there were in fact 541 individuals sample with 26 HIV-positive. However, one of the HIV-positive individuals had been
tested positive in the first sampling, and hence is not included in the second sampling since our statistical estimation method (to be discussed below) requires that those subjects that were tested positive in one sampling must be removed from all subsequent samplings, since it is reasonable to assume that those tested positive will not be tested again. We also note that 43 other persons in the second sample had participated in the first survey study but were found to be seronegative, and hence they remained in the sampled population for the second sampling.
More detailed socio-demographic characteristics of the sampled MSM subpopulation are given in detail in [2, 6, 8]. Our estimation results are relevant to this MSM subpopulation in Beijing only.
Statistical Method: Generalized removal model for open populations (GERMO)
In order to gain insight into the true magnitude of HIV prevalence among the MSM population in Beijing given the restriction that the actual MSM population size is unknown, we make use of the two above-mentioned HIV serosurveys conducted in Beijing between 2005 and 2007 to estimate the number of HIV-infected persons among the sampled MSM subpopulation in Beijing, by utilizing the “Generalized Removal Model for Open populations” [9-16], or the GERMO methodology, which requires at least two sets of sampling data. We consider the two serotest data as random samples drawn from a certain MSM subpopulation in Beijing that can be reached by the sampling procedures employed, in order to estimate the HIV-infected population size in this subpopulation.
In recent years, capture-recapture method (or multiple-record system method in dealing with human populations) has frequently been used for estimating elusive, hard-to-count population groups, such as the MSMs or the intravenous drug users (IDUs), where the emphasis has always been placed on estimating the sizes of these population groups. However, a more direct question of epidemiological importance is the actual number of seropositives within a particular group. In our framework the population size to be estimated is the number of HIV-infected individuals within a certain hard-to-count population who have not developed AIDS defined Illness (ADI), assuming that those individuals with AIDS symptoms would be known and
under treatment. Moreover, there is no recapture since those tested positive will not be tested again, hence the “removal model” [9-10] is the appropriate choice of model for our estimation.
Clearly, given the limited amount of data it is not possible to obtain a valid estimate using the maximum likelihood estimation. Hence we make use of the Bayesian inference of the HIV-infected population size. The Bayes analysis of the model is implemented by using the Gibbs sampler, an MCMC method [17].
The estimation procedure of GERMO is outlined in detail in [11]. We also assume that the natural mortality rate of the MSM population during this time period is small, since the subjects are of ages 17-54 in the first survey (mean age 26.2) [2] and 18-62 in the second survey (median age 27) [6] when the individuals‟ natural mortality is low compared to AIDS-related death rate. Moreover, it has been shown that small variation in natural mortality does not affect the estimation result [11]. Several application of this procedure to estimation of HIV-infected population sizes among at high risk groups can be found in [12-16]. The Bayes estimates are based on the Monet Carlo samples from the Gibbs Sampler run of 50000 iterations after 30000 burn-in.