【作者】 景立伟；

【导师】 曲成毅；

【作者基本信息】 山西医科大学 ， 流行病与卫生统计学， 2014， 博士

【摘要】 背景艾滋病高危人群被称为隐蔽人群或难以接近的人群，通常是因为①艾滋病高危人群害怕受到歧视，不愿承认其高危行为；②在对艾滋病高危人群进行抽样调查研究的时候，通常难以获得抽样框。这些人群特征使得各国政府无法准确掌握艾滋病高危人群基数，以至于干预工作只能覆盖艾滋病高危人群中的部分个体。鉴于此种情形，联合国艾滋病规划署和世界卫生组织在2003年联合发布了《艾滋病高危人群基数估计方法指南》（所涉及的研究方法有枚举法、捕获再捕获法、提名法和乘数法），并于2010年对该指南进行了更新，在传统方法基础上增加了一种新方法——网络扩大法。该方法由美国人类学家Bernard、Killworth、Johnsen和Robinson等人于1991年提出，之后陆续被美国、乌克兰、摩尔多瓦、哈萨克斯坦、卢旺达、日本、泰国和伊朗用于艾滋病高危人群基数的估计。2011年，Salganik等人校正了网络扩大法中的两个偏倚，此举得到了学术界以及联合国艾滋病规划署和世界卫生组织的肯定，从而将方法升级为广义网络扩大法。同年，Salganik、Mello和Abdo等人首次将该方法与传统方法进行了比较，在巴西估计重度吸毒人群基数时，发现广义网络扩大法与乘数法估计结果相差5倍。尽管，之前就有来自英国和俄罗斯的研究结果表明乘数法存在低估现象。但是，由于缺乏重度吸毒人群基数的金标准，该现象仍有可能是乘数法产生了低估，或广义网络扩大法产生了高估。所以Salganik、Mello和Abdo等人建议进一步对两种方法进行比较研究，因为作为联合国艾滋病规划署和世界卫生组织推荐的两种高危人群基数估计方法，无论是乘数法低估还是广义网络扩大法高估都会影响到两种方法的应用前景，对上述情况进行验证以及对成因进行分析就变得迫在眉睫。本次研究目的在于通过评价和对比乘数法和广义网络扩大法在艾滋病高危人群基数估计中的应用，为两法适用人群的选择、是否满足应用条件的判断以及实际应用提供建设性建议。方法以MOOSE指南为准则，对2003年6月1日至2013年6月1日之间在国内利用乘数法估计男男同性恋人群基数的文献进行Meta分析。对PubMed、万方(Wanfang Data)、CNKI (China National Knowledge Infrastructure)、维普(VIPDatabase for Chinese Technical Periodicals)和CBM (Chinese BioMedical LiteratureDatabase)五个数据库进行中英文检索。中文检索词为：“男男同性恋”或“同性恋”、“基数估计”或“人群基数”、“乘数法”和“中国”。英文检索词为：‘men who have sex with men’ OR ‘homosexual’ OR ‘MSM’ AND ‘prevalence’OR ‘size estimation’ OR ‘population size’ AND ‘multiplier method’ AND ‘China’。以MOOSE指南为准则对2003年6月1日至2013年6月1日之间在国内利用乘数法估计女性性工作者人群基数的文献进行Meta分析。对PubMed、万方、CNKI、维普和CBM五个数据库进行中英文检索。中文检索词为：“女性性工作者”或“暗娼”、“基数估计”或“人群基数”、“乘数法”和“中国”。英文检索词为：‘female sex workers’ OR ‘prostitutes’ OR ‘prostitution’ AND‘prevalence’ OR ‘size estimation’ OR ‘population size’ AND ‘multiplier method’AND ‘China’。利用广义网络扩大法对太原市区男男同性恋以及女性性工作者人群基数进行估计。以太原市区常住345.49万人口为调查总体，采用二阶段按比例分层整群抽样方案，在20个行业中按比例随机抽取174个一级单位、405个二级单位以及8031个三级单位进行调查。此外，对294名静脉注射吸毒人员、319名男男同性恋以及301名女性性工作者人群进行同伴驱动抽样调查。在本次研究中，随机应答技术将首次与广义网络扩大法相结合以消除应答偏倚。结果在国内利用乘数法估计男男同性恋人群基数文献的Meta分析中，剔除5项研究之后（3项重复发表、2项未得到结果），剩下的13项研究（1项英文发表、12项中文发表）来自11个省的13个城市。13项研究乘数法估计的男男同性恋人群基数占其本地15至49岁男性人口比例的中位数(0.486%)低于国家疾病预防控制中心、联合国艾滋病规划署和世界卫生组织联合提出中国平均水平(2%-4%)的4到8倍。在国内利用乘数法估计女性性工作者人群基数文献的Meta分析中，剔除3项重复发表研究之后，剩下的7项研究（2项英文发表、5项中文发表）来自7个城市。7项研究乘数法估计的女性性工作者人群基数占其本地15至49岁女性人口比例的中位数(2.485%)处于国家平均水平(0.2%-2.6%)区间范围内。在利用广义网络扩大法对太原市区男男同性恋以及女性性工作者人群基数估计中，总共有8031名一般人群受访者接受了广义网络扩大法的现场调查，应答率96.4%。294名静脉注射吸毒人员、319名男男同性恋以及301名女性性工作者人群接受了同伴驱动抽样调查。太原市男男同性恋人群基数估计的结果为26870人(95%CI:23556,30184)，占太原市15至49岁男性人口的2.927%(95%CI:2.566%,3.288%)，处于国家平均水平(2%-4%)区间范围内。太原市女性性工作者人群基数估计的结果为11787人(95%CI:10767,12808)，占太原市15-49岁女性人口的1.284%(95%CI:1.172%,1.400%)，处于国家平均水平(0.2%-2.6%)区间范围内。结论乘数法估计艾滋病高危人群基数的Meta分析以及乘数法的实证研究结果表明，在女性性工作者人群基数估计中，乘数法获得了有效估计，然而，在男男同性恋人群基数估计中，由于难以获取高质量历史数据，乘数法频频出现低估现象，所以该方法并不适合用于男男同性恋人群基数的估计。广义网络扩大法的实证研究结果表明，在男男同性恋和女性性工作者人群基数估计中，广义网络扩大法均获得了有效估计。由于在男男同性恋人群基数估计中，乘数法频频出现低估现象，所以当具备了高质量的已知人群信息时，应优先选择广义网络扩大法。此外，我们建议应将随机应答技术与广义网络扩大法结合应用。更多还原

【Abstract】 ObjectivesBehavioral interventions are effective strategies for HIV/AIDS prevention and control.However, implementation of such strategies relies heavily on the accurate estimationof the high-risk population size. Populations at high risk for humanimmunodeficiency virus mainly include female sex workers, men who have sex withmen, and injection drug users). These groups are considered as hidden orhard-to-reach populations with two characteristics that create difficulties forestimating population size:1) They tend to hide their true identities from the publicbecause of the stigma; and2) There is no sampling frame used for surveys. However,without reliable estimates of the size of these at-risk populations, the ability ofgovernments to carry out intervention planning, resource allocation, estimate thenumber of people infected with HIV, project disease burden, measure coverage, andevaluate interventions is limited. With these issues in mind, the Joint United NationsProgramme on HIV/AIDS and the World Health Organization released guidelines in2003, which were updated in2010, for estimating the size of populations at high riskfor HIV. The guideline encompasses enumeration methods, capture-recapture method,nomination methods, multiplier method, population surveys and generalized networkscale-up methods. The most important advantage of the multiplier method is that it ismuch more straightforward to use. However, studies of the multiplier method used toestimate the population prevalence of men who have sex with men in China publishedbetween July1,2003and July1,2013were reviewed. The median of studies usingthe multiplier method to estimate the population prevalence of men who have sexwith men in China was4-8times lower than the national level estimate.The network scale-up method was first proposed by Bernard, Killworth, Johnsen,and Robinson in1991and uses information collected in general population surveys toestimate the size of populations that have a high risk of HIV. The generalized network scale-up method proposed by Salganik et al. based on the network scale-upmethod lies in adjusting information transmission bias with the popularity ratio andinformation transmission rate. The multiplier method and generalized networkscale-up method were used to estimate the number of heavy-drug users in Curitiba,Brazil in2011. However, Salganik et al. found that the estimate yielded by thegeneralized network scale-up method was5times higher than that of the multipliermethod. It is possible that the generalized network scale-up method producedoverestimates or the multiplier method produced underestimates in the absence of agold standard for the number of heavy-drug users. Thus, Salganik et al. recommendedthat additional studies be undertaken to assess these two methods. This study aims toassess and compare the two methods for estimating the size of populations at highrisk for HIV, and to provide practical guidelines and suggestions for implementingthe two methods.MethodsStudies of the multiplier method used to estimate the population prevalence of menwho have sex with men and female sex workers in China published between July1,2003and July1,2013were reviewed. The MOOSE recommendations were used toconduct the reviews. Five electronic databases were searched in Chinese and English:PubMed, Wanfang Data, China National Knowledge Infrastructure, VIP Database forChinese Technical Periodicals, and Chinese BioMedical Literature Database.A survey of the general population using stratified two-stage cluster samplingwas conducted in the urban district of Taiyuan. For the first stage,174primary units(institutions/organizations/companies/governments) from all20industries wereselected from the sampling frame, with the probability of selection being proportionalto industry size. For the second stage,405second-stage units (departments) weredrawn from those chosen in the first stage wherein8,031third-stage units(respondents) aged18and above were interviewed and randomized responsetechnique was applied to eliminate response bias. A national-level estimate proposedby the National Center for AIDS and Sexually Transmitted Disease Control andPrevention in collaboration with UNAIDS/WHO in2007was selected as a reference to assess these two results.ResultsAfter the removal of5studies (3duplicate publications and2studies did not obtainappropriate data),13studies (1in English and12in Chinese) were from13cities.The estimated MSM population prevalence of the adult male population of the13cities were0.066%,0.249%,0.486%,1.220%, and6.884%for the minimum,25thpercentile, median,75th percentile, and maximum, respectively. The median (0.486%)of the13cities was4-8times lower than the national-level estimate (2.0-4.0%).After the removal of3duplicate studies,7studies remained (2in English and5in Chinese) that originated from7cities. The estimated FSW prevalence of the adultfemale population (aged15-49) of the7cities was0.777%,1.766%,2.485%,3.600%and4.365%for the minimum,25th percentile, median,75th percentile, and maximum,respectively. The median (2.485%) of the7cities fell within the range ofnational-level estimate (0.2-2.6%).The generalized network scale-up method allowed us to estimate the number ofMSM in Taiyuan as being26,870(95%CI:23,556,30,184), which corresponds to2.927%(95%CI:2.566%,3.288%) of the adult male population (aged15-49). Theestimates of generalized network scale-up method fall within the range ofnational-level estimate. The generalized network scale-up method allowed us toestimate the number of FSW in the urban district of Taiyuan as being11,787(95%CI:10,767,12,808), corresponding to1.284%(95%CI:1.172%,1.400%) of the femalepopulation aged15-49.ConclusionsWhen high-quality existing data are not readily available, the multiplier methodfrequently yields underestimated results. We thus suggest that the generalizednetwork scale-up method is preferred when sampling frames for the generalpopulation and accurate demographic information are available.更多还原