Reducing HIV Vulnerability in High Risks Populations

最近的注意力集中在美国艾滋病毒流行的变化性质上。具体而言，人们日益关切的是，特定人群受到艾滋病毒的影响不成比例。这些人群包括非裔美国人（AA）和拉丁裔性少数群体，变性人，AA异性恋妇女，以及生活在南部和农村社区的人。 有证据表明，使用暴露前预防（PrEP）是最有效的艾滋病毒预防策略之一，如果每天坚持使用，可以将艾滋病毒发病率降低90%。然而，据报道，AA异性恋妇女坚持PrEP的比率要低得多，她们占美国新感染艾滋病毒的约10%。事实上，PrEP在这一人群中的早期临床试验未能显示出疗效，主要是由于依从性低。鉴于坚持是决定PrEP疗效的主要因素，在这个脆弱的人口中，研究吸收和坚持的障碍是必要的。 虽然据报道，AA女性不遵守PrEP的最常见原因是忘记服用药物并且手头没有药物，但这些女性也报告了在个人，人际和机构层面遵守PrEP的一系列障碍。 该项目将利用大数据分析在国家一级进行形成性研究，以确保所有次级社区都有代表性。了解个人的实际行为对于制定有效的干预措施非常重要。使用被动测量技术和数字表型分析技术，我们将尝试更多地了解AA女性的日常生活，包括数字和传统媒体的使用;他们何时何地度过大部分时间;他们与谁联系以及他们的重要关系（数字和现实生活）;以及他们对艾滋病毒风险的看法，PrEP知识，以及获得艾滋病毒预防服务。 从个人在社交媒体平台上留下的数字足迹中，我们可以识别与艾滋病毒风险相关的广泛行为的时间和地理模式，包括物质使用，商业性行为和无保护的性活动。这里采用的方法将以各种方式利用社交媒体数据，从使用指定关键字和主题的出现来识别后续访谈的个人，到应用机器学习技术自动识别与高风险艾滋病毒行为相关的单词，以检测这些行为的个体实例的出现。 为了实时捕捉艾滋病毒风险行为的详细模式，智能手机是完美的工具。这些设备已经嵌入到美国超过79%的成年AA女性的日常生活中。生态瞬时评估（EMA）技术，提示用户在一天中回答一组简短的问题，将被用来收集关于他们的生活方式行为的细粒度信息，实时，大规模。此外，将采用一种更为被动的数据收集形式。我们将使用智能手机传感器和可穿戴设备来捕获行为数据，而不需要用户的任何主动输入。 众包是捕捉行为、信念和环境约束快照的理想方法。这些平台将使我们能够招募大量具有特定人口特征的AA女性，以完成在线调查和实验。将收集代表整个美国的数千名AA女性的数据。使用基于理论的方法，我们将询问有关影响PrEP采用和坚持的因素的问题，包括采取PrEP的优点和缺点（以确定结果预期，成本效益和潜在的行为信念）;来自重要他人的规范压力;以及促进或阻碍PrEP吸收和坚持的因素。 该项目的最终目标是在AA妇女中开发和评估数字艾滋病毒预防干预措施，该干预措施使用被动数据收集方法和机器学习（例如，深度学习，AI）来影响PrEP的采用和遵守。

More recent attention has focused on the changing nature of the HIV epidemic in the United States. Specifically, there are increasing concerns that specific populations are being disproportionately affected by HIV. These populations include African Americans (AA) and Latino sexual minorities, transgender populations, AA heterosexual women, and people living in the southern and rural communities. Evidence suggests that Pre-Exposure Prophylaxis (PrEP) use is one of the most effective HIV prevention strategies which can decrease HIV incidence by as much as 90% when adhered to daily. However, rates of adherence to PrEP have been reported to be much lower in AA heterosexual women, who account for approximately 10% of new HIV infections in the United States. In fact, early clinical trials of PrEP in this population have failed to show efficacy, largely due to low adherence. Given that adherence is the principle factor in determining PrEP efficacy, studying barriers to uptake and adherence are warranted in this vulnerable demographic. While the most common reasons for PrEP non-adherence in AA women are reported to be forgetting to take the medication and not having it on hand, these women have also reported a range of barriers to PrEP adherence at the individual, interpersonal, and institutional levels. This project will use big data analysis to conduct formative research at a national level to ensure all sub-communities are represented. Understanding the actual behavior of individuals is important in the development of effective interventions. Using passive measurement techniques and digital phenotyping techniques, we will attempt to learn more about the daily lives of AA women, including digital and traditional media use; when and where they spend most of their time; with whom they associate with and their significant relationships (digital and real-life); and their perceptions of HIV risk, knowledge of PrEP, and access to HIV prevention services. From the digital footprints left by individuals on social media platforms, we can identify temporal and geographic patterns in a wide range of behaviors related to HIV risks, including substance use, commercial sexual behaviors, and unprotected sexual activities. Approaches employed here will utilize social media data in various ways, from using occurrences of specified keywords and topics to identify individuals for follow-up interviews to applying machine-learning techniques to automatically identify words associated with high risk HIV behaviors to detect occurrences of individual instances of those behaviors. To capture detailed patterns of HIV risks behaviors, in real-time, smartphones are the perfect tool. These devices are already imbedded in the daily lives of more than 79% of adult AA women in the US. Ecological momentary assessment (EMA) techniques, which prompt users to respond to a short set of questions, a few times during the day, will be employed to gather fine-grained information about their life-style behaviors, in real-time, on a large scale. In addition, a more passive form of data collection will be employed. We will use smartphone sensors and wearable devices to capture behavioral data without the need for any active input from the user. Crowdsourcing is an ideal approach to capturing snapshots of behaviors, beliefs, and environmental constraints. These platforms will enable us to recruit large numbers of AA women with specific demographic profiles to complete online surveys and experiments. Data from thousands of AA women will be gathered representing the entire US. Using a theory-based approach, we will ask questions about the factors that influence the adoption of and adherence to PrEP, including the advantages and disadvantages of taking PrEP (to identify outcome expectancies, cost-benefits, and underlying behavioral beliefs); normative pressure from significant other; and factors that would facilitate or hinder uptake and adherence to PrEP. The ultimate goal of this project will be to develop and evaluate a digital HIV prevention intervention among AA women that uses passive data collection methodologies and machine learning (e.g., deep learning, AI) to influence adoption and adherence of PrEP.