Developing an artificial intelligence-based mHealth intervention to increase HIV testing in Malaysia

项目摘要 艾滋病毒检测开始进入艾滋病毒预防和治疗的级联。然而，艾滋病毒检测水平是 尤其是在男男性行为者(MSM)中，这一比例更低，他们日益加剧了艾滋病毒的传播 在存在高度耻辱和歧视的情况下。对于高危男男性行为者，新的指南建议 经常进行艾滋病毒检测，每隔3至6个月进行一次。然而，在男男性接触者中进行艾滋病毒检测的频率较低，因为 个人(例如，对风险披露的高度关注)、诊所(例如，违反保密规定，以及 来自医疗保健提供者的歧视)和政策(将同性性行为定为刑事犯罪)障碍。艾滋病病毒 马来西亚男男性接触者的流行率已飙升至全国21.6%，吉隆坡超过40.9%。而当 对马来西亚符合PrEP标准的男男性接触者进行的监测调查表明，70.3%的人曾经检测过，过去- 年测试的比例为40.9%，只有9.5%的人每年测试一次以上，尽管自我测试的水平非常高 报告的风险。然而，一旦检测，马来西亚感染艾滋病毒的MSM很可能接受抗逆转录病毒治疗并实现 病毒抑制，使艾滋病毒检测成为艾滋病毒预防和治疗的中心重点。 因此，鼓励和指导马来西亚男男性接触者检测的创新战略 急需之物。使用信息-动机-行为技能(IBM)模型进行干预非常适合 在男男性接触者中克服推荐的艾滋病毒检测障碍。此外，在像马来西亚这样的环境中，艾滋病毒 理论指导下，流行已从主要集中在PWID转变为MSM的不稳定流行 行为改变战略，告知、激励和提供实用技能，以更充分地参与 建议的艾滋病毒检测将加速艾滋病毒预防和护理的连续体。鉴于有几个 许多个人、诊所和政策障碍阻碍了艾滋病毒检测、移动医疗(MHealth)干预，从而减少了 人员“联系并提供行为技能菜单非常适合在 污名化环境，促进推荐的艾滋病毒检测。最近在美国学习，中国，南非， 秘鲁表明，使用智能手机和应用程序的移动健康干预措施有可能增加艾滋病毒检测 同时对男男性接触者保密。这种mHealth干预在男男性接触者中是可行和可接受的， 包括在马来西亚，大多数男男性接触者使用类似界面的社交网络应用程序寻找性伴侣 以及提议的干预的功能。然而，当前的移动健康战略受到它们缺乏 自动化和对高强度和持续的人力投入的需求，这限制了它们的扩大。人造的 使用机器学习(ML)的智能(AI)可以克服这些限制，但尚未应用于 基于移动健康的艾滋病毒检测算法。因此，我们的目标是开发和试点测试AI-聊天机器人(R21阶段)。 R21阶段的结果将通知类型1混合实施科学试验(R33阶段)以进行评估 与常规治疗相比，人工智能聊天机器人对艾滋病毒检测的有效性和实施结果。

Project Summary HIV testing jumpstarts entry into the HIV prevention and treatment cascade. HIV testing levels, however, are especially low in men who have sex with men (MSM), who increasingly contribute to heightened HIV transmission in the presence of high levels of stigma and discrimination. For high risk MSM, new guidelines recommend frequent HIV testing, ranging from every 3 to 6 months. Yet, HIV testing in MSM often occurs less frequently due to individual (e.g., heightened concerns about risk disclosure), clinic (e.g., confidentiality breaches, and discrimination from healthcare providers) and policy (criminalization of same-sex sexual behaviors) barriers. HIV prevalence in MSM in Malaysia has soared to 21.6% nationally, exceeding 40.9% in Kuala Lumpur. While surveillance surveys of MSM in Malaysia who meet criteria for PrEP suggest that ever tested is 70.3%, past- year tested is 40.9%, and only 9.5% were tested more than 1 time per year, despite extraordinary levels of self- reported risk. Once tested, however, MSM with HIV in Malaysia are likely to be treated with ART and achieve viral suppression, making HIV testing a central focus for HIV prevention and treatment. Innovative strategies that motivate and provide guidance for testing among MSM in Malaysia are therefore urgently needed. Intervening using Information-Motivation-Behavioral Skills (IBM) model is ideally suited to overcome barriers to recommended HIV testing in MSM. Moreover, in settings like Malaysia where the HIV epidemic has transitioned from primarily concentrated in PWID to a volatile epidemic in MSM, theory-guided behavioral change strategies that inform, motivate and provide pragmatic skills to more fully engage in recommended HIV testing are poised to accelerate the HIV prevention and care continuum. Given that there are many individual, clinic and policy barriers to HIV testing, mobile health (mHealth) interventions that reduce “in person” contact and offer a menu of behavioral skills is ideally suited to increase access to MSM in highly stigmatized settings and promote recommended HIV testing. Recent studies in the U.S., China, South Africa, and Peru show that mHealth interventions using smartphones and apps have the potential to increase HIV testing while maintaining MSM’s confidentiality. Such mHealth interventions are feasible and acceptable among MSM, including in Malaysia where most MSM find sexual partners using social-networking apps with similar interfaces and functionalities to the proposed intervention. Current mHealth strategies, however, are limited by their lack of automation and need for high-intensity and sustained human inputs, which restricts their scale-up. Artificial intelligence (AI) using machine learning (ML) may overcome such limitations, but has yet to be applied to mHealth-based HIV testing algorithms. We therefore aim to develop and pilot test an AI-chatbot (R21 phase). Findings from the R21 phase will inform a Type 1 Hybrid Implementation Science trial (R33 phase) to evaluate the efficacy and implementation outcomes of the AI-chatbot for HIV testing relative to treatment as usual.