建立并评估HIV自检二次传递仿真模型在识别MSM人群关键人物中的作用

Despite global efforts to control HIV among key populations, HIV persists as a major public health threat among men who have sex with men (MSM) around the world. For example, in China, HIV prevalence among MSM increased from 0.9% in 2003 to 8.0% in 2015, while the HIV incidence among MSM reached 15.0/100 person-years in a big RCT. However, about half of Chinese MSM have never tested for HIV. Although wider availability of facility-based HIV testing services can help increase testing coverage, bold and creative interventions are still needed to promote testing, identify HIV-infected persons, and link them to care. Secondary distribution of HIV self-testing (HIVST), the process of giving one individual multiple self-testing kits to distribute to people in his/her social networks, is a promising approach to substantially increase HIV testing among Chinese MSM. Secondary distribution of HIVST combines two effective strategies in promoting HIV testing— HIV self-testing and social networks-based strategies. Social networks can amplify the spread of behavior through interpersonal ties. By utilizing the wisdom of the crowd and the social capital of influential individuals such as popular opinion leaders (POLs), the secondary distribution of HIVST has a strong potential to influence a large number of individuals and promote HIV testing among them. However, POLs are usually identified through a scale-measured method, which uses self-reported scales and limited empirical analysis. Scale-measured POLs do not adequately account for the overall social network structure of individuals, which may limit the effectiveness of network-based strategies. Optimizing the identification of a subset of key POLs who can connect multiple clusters/sub-communities in the social network through simulation models could solve this problem. This study aims to calibrate a theoretical simulation model and optimization algorithm for the identification of key POLs among Chinese MSM, and to further calibrate the model by using molecular epidemiological data, and then to verify the model using a quasi-experimental study that compares the effectiveness of a crowdsourced secondary distribution intervention distributed through model-identified POLs and scale-measured POLs. The project has two specific aims: First, we will use a theoretical simulation model and an optimization algorithm to identify key POLs among Chinese MSM; second, we will verify the simulation model through a quasi-experimental study, by evaluating whether the model-identified POLs can promote higher HIV testing in a three months period in comparison to the scale-measured POLs. Upon completion of the study, we will provide the scientific community with a verified simulation model for identifying key POLs, and will further evaluate the effectiveness of the secondary distribution of HIVST.

男男性行为者(MSM)是HIV防控的重点人群，但其HIV疫情仍未得到有效控制，而不知晓自身感染状态是一重要原因。HIV自检可提高MSM对自身HIV感染状态的认知。HIV自检二次传递是一种新型、高效的促进HIV检测的方法。识别MSM关键人物并实施此干预可进一步提高干预效能。但关键人物的识别多采用自报或者经验式的量表，未考虑MSM人群社会网络结构特征及HIV传播网络参数，因而识别的关键人物影响力往往有限，干预效果有限。我们拟用已收集的MSM人群社会网络特征建立并优化识别MSM人群关键人物的HIV自检二次传递仿真模型，并在模型构建中纳入HIV传播网络参数，以更准确的通过模型识别MSM关键人物从而实施干预。随后开展类实验，比较模型和量表分别识别的关键人群在接受HIV自检二次传递干预后的不同效果，旨在验证仿真模型，并为我国其他基于社会网络实施的干预及防控策略制定提供新的工具与依据。

HIV自我检测（下称“自检”）二次传递可以有效地提高重点人群的HIV检测率，促进感染者发现，并已经在全球多个地区进行推广应用。性健康关键人物（下称“关键人物”）是指那些能发挥自身影响力从而促进同伴改善其性健康行为的人的总成。识别男男性行为人群（GBMSM）中的关键人物可进一步促进HIV自检二次传递，以增加其在扩大检测和感染者发现方面的效能。过往研究使用了传统量表识别GBMSM中的关键人物，效果有限。本研究拟结合数据科学方法，基于前期收集的HIV自检二次传递数据作为训练集，开发一套机器学习模型以识别GBMSM中的关键人物，并通过和传统量表方法进行比较，评估其所识别出的关键人物在促进自检二次传递中的效能。结合数据科学方法，基于前期收集的HIV自检二次传递数据，本研究首先开发了一套机器学习模型以识别GBMSM中的关键人物。本研究招募符合条件并同意参加实验的GBMSM（≥18岁）并将其随机分到机器学习模型组或传统量表组分别识别关键人物。被识别出的关键人物随后被邀请到HIV自检试剂申领平台上申请自检试剂、生成个性化的专属链接，以完成HIV自检二次传递。受赠者完成自检后将结果照片传回平台即为分享成功。关键人物每成功推荐一位同伴，就可获得平台的20元现金奖励。2021年1月7日至2021年12月11日，本研究共招募了1921名参与者，通过识别，共有393名关键人物（量表组：195，模型组：198）愿参与HIV自检二次传递。最终有229名关键人物（量表组：116，模型组：113）参与了二次传递 ，其平均年龄为28±7.65岁，76%完成了大学以上学历，86%为单身，73%为同性恋者。实验结果表明，模型组的关键人物动员了198名受赠者进行HIV自检，而量表组有103名受赠者（MD=0.86，95%CI：-0.11，-2.02, [aIRR]=1.83，95%CI=1.11，3.05）。此外，在所动员的受赠者中，模型组的关键人物动员了82名首次检测者，量表组为44名（MD=0.35，aIRR=1.54，95%CI=0.77，3.08。相对于量表法，机器学习模型所识别出的MSM关键人物能够更好地促进HIV自检二次传递，提示该方法具有较好的识别关键人物的效能，或可扩大男男性行为人群的HIV检测覆盖面。