Assessing the impact of an intervention to prevent violence and HIV in Uganda

乌干达拉凯的特点是成熟的艾滋病毒流行和亲密伴侣暴力（IPV）发生率高。2005年，成人艾滋病毒流行率为16%（妇女）和12%（男子），发病率为1.5/100人年（PY）;而在过去一年中，20%和14.5%的妇女报告了身体或性IPV（分别）。这是一个重大的公共卫生问题，因为经历IPV的妇女感染艾滋病毒的风险增加，感染艾滋病毒的妇女感染IPV的风险增加。虽然选择的干预措施已经独立地减少了艾滋病毒和IPV，但没有一种一维方法成功地减少了这两种结果。拟议研究的目的是评估安全家园和尊重每个人（SHARE）项目的影响，这是Rakai健康科学计划（RHSP）进行的艾滋病毒/IPV综合干预措施。SHARE的设计基于行为改变的跨理论模式，包括与社区志愿者合作、卫生动员和教育、能力建设活动、宣传、媒体和特别活动。采用了一种随机分组试验设计，并通过母研究Rakai社区队列研究（RCCS）收集数据，该研究每12-18个月收集一次调查和生物学数据，所有年龄在15-49岁之间的人都同意。在基线时，有47个活跃的RCCS社区分为11个集群，其中4个被分配到干预组，7个被分配到试验的对照组。干预地区接受了SHARE和RHSP服务，包括艾滋病毒检测、提供抗逆转录病毒治疗和抗逆转录病毒治疗前护理、一般医疗护理、预防母婴艾滋病毒传播、计划生育、健康教育和社区动员。控制区域不暴露于SHARE，但暴露于RHSP服务。通过RCCS调查和干预前（2005年2月至2006年6月）和干预后2次随访（2006年8月至2009年11月）进行的艾滋病毒血清学进行了影响评估。基线时，共有14，119名参与者（干预社区3，316名，对照社区10，803名）参加了研究。拟议的研究有三个目标。目标1：与对照社区相比，评估干预措施对过去12个月内身体和/或性IPV流行的影响。调整后的患病风险比（PRR）将使用对数二项或改良泊松回归和随机效应模型进行估计，以解释群内相关性。目的2：评估干预措施对艾滋病发病率的影响，与对照社区相比。将遵循意向治疗方法，使用对数连接和泊松分布估计每100 PY HIV感染的发生率比（IRR），使用随机效应模型解释群内相关性并按群进行调整。目标3：评估干预措施对性风险行为（过去一年的非婚伴侣关系，过去6个月的安全套使用情况和性行为前饮酒情况）的影响，并与对照社区进行比较。将使用对数二项或改良泊松回归和随机效应模型估计PRR，以解释聚类内的相关性。

Rakai, Uganda is characterized by a mature HIV epidemic and high rates of intimate partner violence (IPV). In 2005, adult HIV prevalence was 16% (women) and 12% (men) and incidence was 1.5/100 person-years (PY); while 20% and 14.5% of women reported physical or sexual IPV (respectively) during the past year. This represents a significant public health concern because women who experience IPV are at increased risk for HIV acquisition and HIV-infected women are at increased risk for IPV. Although select interventions have independently reduced HIV and IPV, no one-dimensional approach has successfully decreased both outcomes. The objective of the proposed research is to assess the impact of the Safe Homes And Respect for Everyone (SHARE) Project, a combination HIV/IPV intervention conducted by Rakai Health Sciences Program (RHSP). The SHARE design was based on the Transtheoretical Model of behavior change and included work with community volunteers, health mobilization and education, capacity building activities, advocacy, media and special events. A cluster randomized trial design was used and data were collected via the parent study, the Rakai Community Cohort Study (RCCS), which collects survey and biological data every 12-18 months with all consenting people aged 15-49. At baseline there were 47 active RCCS communities grouped into 11 clusters, 4 of which were assigned to the intervention arm and 7 to the control arm of the trial. Intervention areas were exposed to SHARE and RHSP services including HIV testing, provision of ART and pre-ART care, general medical care, prevention of mother-to-child HIV transmission, family planning, health education and community mobilization. Control areas were not exposed to SHARE but were exposed to RHSP services. Impact assessment was done through RCCS surveys and HIV serology done pre-intervention (Feb 2005-June 2006) and at 2 post-intervention follow-ups (Aug 2006-Nov 2009). A total of 14,119 participants (3,316 in intervention communities and 10,803 in control communities) were enrolled at baseline. The proposed research has 3 aims. Aim 1: To assess the impact of the intervention on the period prevalence of physical and/or sexual IPV in the past 12 months compared to the control communities. Adjusted prevalence risk ratios (PRR) will be estimated using log-binomial or modified Poisson regression with random effects models to account for within-cluster correlation. Aim 2: To assess the impact of the intervention on HIV incidence compared with control communities. An intention-to-treat approach will be followed using a log link and Poisson distribution to estimate the incidence rate ratios (IRR) of HIV acquisition per 100 PY, using random effects models to account for intra-cluster correlation and adjusting by cluster. Aim 3: To assess the impact of the intervention on sexual risk behaviors (non-marital partnerships in the past year, condom use over the past 6 months and use of alcohol before sex) compared to the control communities. PRRs will be estimated using log-binomial or modified Poisson regression with random effects models to account for correlation within clusters.