Identifying COVID-19 vaccine deserts using Machine Learning and Geospatial Analyses to target Community -engaged testing for vulnerable rural populations to prevent localized outbreaks

项目摘要 截至2021年6月30日，西弗吉尼亚州（WV）55个县中有23%的县在大多数县中排名前20%。 美国易受新型冠状病毒感染的县。该州极端脆弱性的核心是 医疗合并症的患病率，农村人口获得护理的机会减少，疫苗接种减少 与城市同行相比。值得关注的是，全州范围内的测试减少了， 积极传播疫苗。不幸的是，低检测化合物在医学上对Covid-19的脆弱性 疫苗接种率低的服务不足人群，因为他们极易受到持续的局部 病毒的爆发和随后较高的发病率和死亡率。我们的RADx-UP第二阶段提案 在先前资助的RADx-UP第一阶段的基础上，通过识别和瞄准疫苗沙漠社区， 然后根据个人社区的需求定制测试活动服务， 什么是重要的。为持续测试提供动态解决方案至关重要。我们定义疫苗沙漠 使用总体疫苗接种率和两周内疫苗摄取量的变化。机器学习与 时间序列模型是用来表征县级传输率，并在这里首次纳入 疫苗接种率。县一级的风险估计与疫苗沙漠所在的邮政编码重叠， 使用14天期间总体疫苗接种率和疫苗接种变化的底部十分位数确定。 一旦确定了一个社区，研究联络员将联系研究工作人员和倡导者， 面试，以确定合作伙伴网站举办测试活动，并收集数据，以定制促销，食品和媒体 针对每个社区的具体需求进行信息传递。测试活动将包括抽样和调查 收集数据，并提供促销和机会，以激励社区参与。我们建立 在RADx-UP的一项活动中，重点关注每个社区的第一响应者，以帮助举办测试 事件，并在适用的情况下基于利润的信念。我们让合作调查者与 西弗吉尼亚州南部，这是一个资源有限的地区，用于RADx-UP第一阶段。此外，我们还进行了一项试点研究， 在600名参与者中检查ABBOTT ID Now等温PCR系统的性能。效果 通过监测县的测试前和测试后的比率，使用空间回归来评估干预 分析。一个独特的属性的统计框架，我们建议评估我们的测试策略是一种能力 除了目标社区外，还要说明对附近县的影响。该项目将利用 与地方和国家机构建立并发展自己独特的伙伴关系， 在疫苗沙漠中的社区参与测试交付模式。我们方法的一个重要而新颖的方面是 是建立一个基层的第一反应者研究网络，可以利用它来实施 在隔离的医疗服务不足的社区或研究第一反应者的健康结果的筛查计划。

PROJECT ABSTRACT As of June 30, 2021, 23% of West Virginia’s (WV) 55 counties were ranked within the top 20% of most vulnerable counties to Covid-19 in the United States. Central to the state’s extreme vulnerability is higher prevalence of medical comorbidities, lower access to care among rural populations, and decreased vaccine uptake compared to urban counterparts. Of considerable concern, testing has decreased statewide to allow for active dispersal of the vaccines. Unfortunately, low testing compounds vulnerability to Covid-19 in medically underserved populations where vaccine uptake is low, as they are extremely susceptible to persistent localized outbreaks of the virus and subsequently higher morbidity and mortality. Our RADx-UP Phase Two proposal builds upon previously funded RADx-UP Phase One by identifying and targeting vaccine desert communities then tailoring testing event services to the needs of individual communities building upon their perceptions of what is important. Providing a dynamic solution for continued testing is critical. We define vaccine deserts using overall vaccination rate and the change in vaccine uptake over a two-week period. Machine learning with time series modeling is used to characterize county level transmissibility, incorporating here for the first-time vaccination rates. Risk estimates at the county level are overlaid with zip codes where vaccine deserts have been identified using bottom decile for overall vaccination rate and change in vaccination over a 14-day period. Once a community is identified study liaisons will connect study staff to advocates to conduct semi-structured interviews to identify partner sites to host testing events and collect data to tailor promotions, food, and media messaging to the specific needs of each community targeted. Testing events will involve sample and survey data collection, with promotions and chance giveaways to incentivize communities to participate. We build upon RADx-UP one activity by focusing heavily on first responders in each community to aid in hosting testing events, and faith based and on profits where applicable. We involve co-investigators with strong connections to southern WV, an area with limited resources for RADx-UP Phase One. Additionally, we conduct a pilot study to examine the performance of the ABBOTT ID Now isothermal PCR system in 600 participants. Effect of the intervention is evaluated through monitoring of pre and post testing rate for the county using spatial regression analyses. A unique attribute of the statistical framework we propose to evaluate our testing strategy is an ability to describe the impact on nearby counties in addition to the targeted community. This project will leverage existing and develop its own unique partnerships with local and state agencies for implementation of a community engaged testing delivery model within vaccine deserts. A critical and novel aspect of our approach is establishment of a grass roots first responders research network which can be leveraged to implement screening programs in isolated medically underserved communities or study first responder health outcomes.