我国吸毒者艾滋病感染的空间特点及影响因素分析

发布时间：2018-02-22 01:42

本文关键词： 艾滋病注射吸毒空间分析多水平模型地理加权回归模型　出处：《中国疾病预防控制中心》2014年硕士论文　论文类型：学位论文

【摘要】：目的：了解并分析1995-2011年我国吸毒者艾滋病感染的空间特点,并在感染及数据可能存在聚集性的情况下,分析影响其感染艾滋病病毒(Human immunodeficieney virus,HIV)的个体因素和宏观因素。研究方法：利用1995-2011年间以注射吸毒为途径感染HIV者和艾滋病病人(Acquired Immune Deficiency Syndrome,AIDS)数据,分别以省份和区县为水平进行空间聚集性分析并生成热点区县的中心位点。收集我国2011哨点监测的吸毒者数据进行我国吸毒者艾滋病感染个体影响因素的分析,利用三水平Logistic模型,分析我国社区吸毒者艾滋病流行的影响因素,计算相关变量的参数估计值及OR值。建立地理加权回归模型探究社会经济因素对我国吸毒者艾滋病感染的影响,为了减少自变量间的共线性,通过主成分分析从经济、交通、社会保障与安全及卫生四方面各提取一个综合变量作为自变量,并定义2007年-2011年全国累计注射吸毒者艾滋病报告密度为应变量拟合地理加权回归模型,分析宏观因素对我国吸毒者艾滋病感染的影响,以得到全国各省份综合变量的局部系数值。研究结果： 1995-2011年间,我国注射吸毒者艾滋病感染分布并不均匀,经空间全局自相关检验,我国累积注射吸毒者艾滋病感染在全国范围内存在聚集性(Moran's I=0.066,Z值=32.629且P0.05),并且逐年注射吸毒者艾滋病感染均具有聚集性。1995-2011年间局部热点多位于西部新疆维吾尔自治区及西南部云南、广西壮族自治区、四川省等。其中,1995-2003年期间,西部地区热点局限于新疆省,西南部地区热点有由从西向东转移的趋势。2004-2011年期间,西部地区热点位置仍处于新疆维吾尔自治区,西南部地区热点有由从边境向内陆转移的趋势。 2011年我国哨点监测共调查社区吸毒人员42011人,HIV感染率为5.01%(95%CI：4.80%,5.22%)。零模型分析结果显示显示社区吸毒者艾滋病感染在区县水平、省份水平上均有聚集性。进一步通过三水平Logistic模型分析影响因素,结果显示,感染丙肝(OR=6.404,95%CI:5.411,7.580)、共用针具(OR=4.043,95%CI：3.545,4.618)、注射吸毒(OR=1.736,95%CI:1.391,2.167)、少数民族(OR=1.728,95%CI:1.446,2.065)等人口学特征及行为学因素是社区吸毒者感染HIV的危险因素。 2007-2011年间,我国累积注射吸毒者艾滋病报告密度存在空间上的聚集性(Moran's1=0.102,Z值=2.472且P0.05)。由地理加权回归模型可得,经济水平系数在大多数省份为负值。交通系数为另一个系数多为负值的社会经济综合变量。卫生水平系数、社会保障与安全系数在省级水平上多为正值。经地理回归模型拟合后,对模型预测各省艾滋病感染数与真实值的差值(残差)进行全局自相关检验,未发现空间自相关性(Moran's1=0.030,Z值=1.970,P=0.56)。模型校正后R2=62.5%、Condition Number均小于30,说明模型拟合效果良好。研究结论： 1995-2011年间我国注射吸毒艾滋病感染存在空间上的聚集性,近年来局部热点数量有所增加出现由边境向内陆转移的趋势。在分析我国吸毒者艾滋病感染的影响因素时,多水平模型可以用于处理我国艾滋病哨点监测数据,并得出较为真实的结论,但仍需在群体水平继续探索相应的影响因素。地理加权回归模型可被用于分析我国宏观因素对吸毒者艾滋病感染流行的影响,可体现出我国各省经济、交通、社会保障与安全及卫生水平对其吸毒者艾滋病感染流行有不同的响应性。
[Abstract]:Purpose : To understand and analyze the spatial characteristics of HIV infection among drug users in China from 1995 to 2011 , and to analyze the individual factors and macro - factors affecting the infection of human immunodeficiency virus ( HIV ) in the presence of infection and data aggregation . Study method : The data of HIV and AIDS patients were infected by injecting drug use from 1995 to 2011 , and spatial clustering analysis was conducted at the provincial and district level , respectively , and the central site of hot spot district was generated . The influence factors of HIV / AIDS infection among drug users in our country were analyzed by collecting the data of drug users monitored in 2011 . Using the three - level Logistic model , the influencing factors of AIDS epidemic among community drug users in our country were analyzed , and the parameters estimation value and OR value of relevant variables were calculated . The effects of socio - economic factors on HIV infection in our country were investigated by establishing a geographical weighted regression model . In order to reduce the collinearity among the variables , an integrated variable was extracted from the four aspects of economy , transportation , social security and safety and health through principal component analysis . Results of the study : During the period from 1995 to 2011 , the distribution of AIDS infection among injecting drug users in China was not uniform , and the global self - correlation test was conducted in China . In the period 1995 - 2011 , the hot spots in the western region were limited to Xinjiang and Southwest Yunnan , Guangxi Zhuang Autonomous Region , Sichuan Province and so on . In 2011 , China sentinel surveillance co - investigated 42011 community drug users with HIV infection rate of 5.01 % ( 95 % CI : 4.80 % , 5.22 % ) . The results of the zero - model analysis showed that the prevalence of HIV infection in community drug users was aggregated at the county level and in the provincial level . The results showed that the factors influencing the infection of HIV by community drug users were the demographic characteristics and behavioral factors such as hepatitis C ( OR = 6.404 , 95 % CI : 5.411 , 7.580 ) , common needle set ( OR = 1.0736 , 95 % CI : 1.391 , 2.167 ) and ethnic minorities ( OR = 1.728 , 95 % CI : 1.446 , 2.065 ) . During the period from 2007 to 2011 , there was a spatial clustering of AIDS report density ( Moran ' s1 = 0.102 , Z = 2.472 and P0.05 ) . Based on the geographic regression model , the economic horizontal coefficient is negative in most provinces . The traffic coefficient is a social and economic comprehensive variable with more negative values . The health level coefficient , social security and safety factor are more positive at the provincial level . After fitting by the georegression model , the difference between the number of AIDS infection and the real value ( residual ) is globally self - correlated , and no spatial autocorrelation is found ( Moran ' s1 = 0.030 , Z = 1.970 , P = 0.56 ) . After model calibration R2 = 62.5 % , Condition Number is less than 30 , which shows that the model fitting effect is good . Conclusions of the study : In the period of 1995 - 2011 , there is a spatial aggregation of HIV / AIDS infection in China . In recent years , the number of local hot spots has increased from the border to the inland . In the analysis of the factors affecting the HIV infection in our country , the multi - level model can be used to deal with the AIDS sentinel surveillance data in China , and to obtain a more realistic conclusion . The geographical weighted regression model can be used to analyze the impact of macro - factors on HIV infection in China .

【学位授予单位】：中国疾病预防控制中心
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R512.91

【参考文献】