【作者】 张颖；

【导师】 程怡民；

【作者基本信息】 北京协和医学院 ， 流行病与卫生统计学， 2013， 博士

【摘要】 目的本研究的总目的是在陕西省农村地区已婚育龄妇女生殖道感染(Reproductive tract infections, RTIs)流行现状及影响因素资料的基础上,分析生殖道感染患病及相关行为特征在乡镇、区县和城市等群体层面上是否存在聚集性。群体层面特征是否影响农村地区已婚育龄妇女的生殖健康水平和生殖健康服务的利用。具体目的为：1、应用多水平模型分析个体特征因素对陕西农村已婚育龄妇女RTIs疾病、症状及相关的行为,包括避孕行为、性行为和求医行为的影响。2、应用多水平模型分析群体特征因素对陕西农村已婚育龄妇女RTIs疾病、症状及相关的行为,包括避孕行为、性行为和求医行为的影响。3、多水平模型和单水平模型在RTIs研究中的比较。方法本研究所采用的方法包括：流行病学定量研究方法；文献计量学方法；系统评价方法。1、流行病学定量研究方法采用流行病学横断面研究(Cross sectional study)方法和多阶段随机抽样方法调查陕西省已婚育龄妇女生殖健康现状及影响因素。数据的搜集时间在2011年7月-10月。在陕西省10个城市中随机抽取了20个区县,在20个区县中随机抽取了80个乡镇(含80个行政村)。在选定的行政村中,纳入了符合“入选标准”的年龄在20-49岁之间的农村已婚育龄妇女共20878名。利用Epi Data2.1软件进行数据的双份录入和比对。根据本研究设计的特点,在计算RTIs患病率、自觉症状的发生率、避孕行为和求医行为的构成比时,对分析的资料进行权重校正。采用SAS9.1软件的Proc Survyfreq过程进行分析。在对单水平影响因素进行分析时,采用SPSS12.0软件进行单水平多因素Logistic回归分析。当在模型中纳入乡镇、区县、市等水平的分析变量时,采用多水平多因素Logistic回归模型和多水平多因素线性回归模型。分析软件采用MLwiN2.26。2、文献计量学方法采用文献计量学方法,对国内近十年来发表的针对农村已婚育龄妇女RTIs展开的流行病学研究论文进行全面梳理和分析。为本研究的选题和分析提供定量研究的依据。数据采用SAS9.1软件进行分析。3、系统评价方法经过系统回顾和文献的鉴别、甄选过程。对纳入研究的文献进行系统评价。共进行了两个大类(干预前后对照研究和群组对照研究)、4个亚组(每个大类中分别对生殖道感染疾病和生殖道感染相关知识的知晓率)的分析。采用RevMan5.2软件进行主要结果的分析；通过编写SAS程序来完成Z值及失安全系数估算。结果1、过去十年RTIs研究文献统计方法学情况共纳入统计分析的文献有179篇。超过40%的研究在设计时采用了分层整群抽样方法。30%以上的研究使用单水平Logistic回归进行影响因素的统计分析,使用多水平模型的研究仅有2篇。2, RTIs患病及相关行为的影响因素分析农村已婚育龄妇女的RTIs患病率为58.7%。单水平Logistic回归和多水平Logistic回归的结果发现,年龄越大、有RTIs既往史和分娩次数越多是RTIs患病的危险因素。在单水平模型中有统计学意义的教育水平(OR:1.1760,95%CI:1.0662-1.2971)、职业(OR:0.7970,95%CI:0.7084-0.8966).家庭年收入(OR:0.8705,95%CI:0.7869-0.9629)和人流次数(OR:1.1961,95%CI:1.0866-1.3167),在多水平模型中不具有统计学意义(P>0.05)。过去6个月RTIs自觉症状的发生率为70.0%。单水平Logistic回归和多水平Logistic回归的结果发现,年龄越大、有RTIs既往史、分娩次数和人流次数越多出现RTIs症状的危险因素。在单水平模型中有统计学意义的家庭年收入(OR:0.8568,95%CI:0.7659-0.9584)在多水平模型中不具有统计学意义(P>0.05)。在考虑个体水平和区县水平的解释变量之后,区县的从业率与RTIs症状的关联有统计学意义(P<0.05)。在出现RTIs症状后,有60.2%的人会去看病,28.1%的人会自己去买药。单水平Logistic回归和多水平Logistic回归的结果发现,年龄在30-39岁之间、教育程度越高、家庭年收入越高、有RTIs既往史、分娩次数越多的人,在出现RTIs症状后去就医的可能性越大。在单水平模型中有统计学意义的职业(OR:1.2163,95%CI:1.0455-1.4150)和人流次数(OR:1.1882,95%CI:1.0578-1.3346)在多水平模型中不具有统计学意义(P>0.05)。此次调查发现,避孕措施以女扎和上环为主,分别占47.3%和34.2%；避孕套的使用占8.1%。单水平Logistic回归和多水平Logistic回归的结果发现,年龄越小、教育程度越高、初婚年龄越晚、有RTIs既往史、分娩次数越少的人,使用避孕套作为避孕方式的可能性越大。在考虑个体水平和区县水平的解释变量之后,区县的人均第一产业增加值与避孕套使用的关联有统计学意义(P<O.05)。3、健康教育的干预效果对纳入研究的15项研究分4个亚组进行Meta分析。分析结果表明,形式多样的健康教育能够降低干预人群的生殖道感染发病率；能提高人群中生殖健康知识的知晓率。结论1、目前,农村已婚育龄妇女仍然是生殖道感染疾病的易感和高发人群。由于大多数相关流行病学调研数据均具有层次结构和聚集特征,所以,有必要对数据的结构特征进行探索分析,在不满足传统统计模型的前提假设时,尝试使用多水平模型。2、农村已婚育龄妇女生殖道感染及相关行为等存在地区(高水平)上的聚集性说明,已婚育龄妇女的生殖健康水平不仅受其自身及家庭多种因素的影响,而且受其所属地区的经济、社会、卫生服务等更为宏观的因素的影响。不同地区(高水平)的已婚育龄妇女生殖健康的影响因素存在差别,提示在进行生殖道感染干预时,应因地制宜,充分考虑地区的特性。在分析具有聚集特征的数据时,多水平模型较之单水平模型更能获得对效应的无偏估计,可降低统计学Ⅰ类错误的发生的概率。3、健康教育在农村地区对疾病预防具有明显的社会效应,可以作为提高公共卫生水平、RTIs预防和生殖健康促进的主要手段。更多还原

【Abstract】 ObjectivesThe main purpose of this study is to explore whether contextual factors of towns,counties and cities have an significant effect on the prevalence of reproductive tract infections and related behaviors among married reproductive women in Shaanxi Province, based on a cross-sectional epidemiology study. The specific purposes are:1. Using multilevel models to estimate the association between the characteristics of lower level (women) and the prevalence of RTIs and related behaviors, such as contraceptives, sexual behaviors and health-seeking behaviors.2. Using multilevel models to estimate the association between the characteristics of higher level (towns, counties, and cities) and the prevalence of RTIs and related behaviors, such as contraceptives, sexual behaviors and health-seeking behaviors.3. Comparing the estimation results between traditional statistical model and multilevel models.MethodsThe methods related to the research purpose include epidemiological quantitative research,bibliometrics, and system review.1. Epidemiological Quantitative ResearchA cross-sectional study were applied to explore the status of reproductive health and influencing factors among married childbearing aged women in rural areas of Shaanxi province. The data was collected using multi-stage cluster random sampling method from July to Oct.,2011. Firstly,20districts (counties) were selected randomly from10cities of Shaanxi Province. Then,80towns (including80administrative villages) were selected from these20districts and counties. Finally,20878married women of childbearing age from20to49years old were selected according to the inclusion criteria.The data was double entered and cross-checked using Epi Data2.1. The weight of the analysis was corrected when calculating the morbidity of RTIs, the incidence of symptoms, the proportions of contraceptive behavior and health-seeking behavior. The analysis was run in the Proc Survyfreq program in SAS9.1.SPSS12.0was used in traditional multivariate logistic regression analysis. The multi-level logistic regression model and multi-level linear regression model were constructed when the contextual variables were added into the model, such as towns, districts and cities. The analysis software was MLwiN2.26.2. BibliometricsThe comprehensive summarize to the epidemiological studies in the last decade were analysised via bibliometrics. The results provided quantitative and objective evidences for the design and analysis of this study.3. The method of systematical evaluationThe literatures included in this study were evaluated systematically, after systematical reviewed and identified, selected. Four subgroup meta analysis were used to evaluate the effect of intervention of health education.The results were analysised using RevMan5.2. SAS programs were used to compute Z value and Nfs0.05.Results1. The methodologies of RTIs literatures in the past ten years179papers are included in the statistical analysis. More than40%of the study has been designed with a stratified cluster sampling method, and more than30%of the studies analyzed the impact factors using traditional logistic regression. There are only2studies using multilevel models.2. The prevalence and influencing factors of RTIs and related behaviorsThe prevalence of RTIs was58.7%. The results of traditional Logistic and multilevel Logistic regression found, the higher age, RTIs history, and more delivery times were the risk factors for RTIs. The education level (OR:1.1760,95%CI:1.0662-1.2971), occupation(OR:0.7970,95%CI:0.7084-0.8966), family yearly income(OR:0.8705,95%CI:0.7869-0.9629), and induced abortion times(OR:1.1961,95%CI:1.0866-1.3167) had significant realtion with RTIs in traditional Logistic model, but these relationships were not significant in multilevel models(P>0.05).The prevalence of RTIs symptoms in the preceding6months was70.0%. The results of traditional Logistic and multilevel Logistic regression found, the higher age, RTIs history, more delivery times and induced abortion times were the risk factors for RTIs symptoms. The family yearly income(OR:0.8568,95%CI:0.7659-0.9584) had significant realtion with RTIs symptoms in traditional Logistic model, but this relationship was not significant in multilevel models(P>0.05). After including the factors of individual level and county level, the relationship between employment rate and RTIs symptoms was statistically significant(P<0.05).60.2%subjects would seek healthcares when RTIs symptoms occured, and28.1%would buy medication themselves. The results of traditional Logistic and multilevel Logistic regression found, the age between30and39, higher education level, higher family yearly income, RTIs history, and more delivery times were the influencing factors for healthcare seeking behaviors. The occupation(OR:1.2163,95%CI:1.0455-1.4150) and induced abortion times(OR:1.1882,95%CI:1.0578-1.3346) had significant realtion with healthcare seeking behavors in traditional Logistic model, but this relationships were not. significant in multilevel models(P>0.05).Female sterilization and IUD were the most prevalent contraceptive methods, the rates were47.3%and34.2%, respectively. The rate of condom use was8.1%. The results of traditional Logistic and multilevel Logistic regression found, the younger age, higher education level, later first marriage age, RTIs history, and less delivery times were the influencing factors for condom use. After including the factors of individual level and county level, the relationship between per capita added value of the first industry and condom use was statistically significant(P<0.05).3. The effects of health education interventionsMeta analysis was applied to the study of15researches,which were divided into4subgroup by the type of study design and evaluation endpoint. The any type of health education can effectively decrease the morbidity of RTIs and improve the RTI knowledge.Conclusions:1. The rural married women of childbearing age are still the high risk population for RTIs. There are nest or cluster structure in most of current studies. The necessary exploration process will be help to choose appropriate statistical models.2. The significant variance of between-high levels in RTIs prevalence and related behaviors implied that, the reproductive health level of women of childbearing age was affected not only by factors of individual and their families, but also by the contexual factors including regional economic, social, and health services.There are differences on influencing factors of the reproductive health of childbearing age women in different districts (high level). Local conditions and the characteristics of the region have to be considered in the RTI intervention policies.Multilevel models have advantages on avoiding type Ⅰ error compared to traditional model for the data with hierachical structure.3. Health education is an easy-designed intervention with effective outcomes. It can be used as a main method to public health improving, RTI prevention and health promotion.更多还原