Multivariate spatiotemporal models to quantify disparities in COVID-19 health outcomes

项目摘要 严重急性呼吸综合征冠状病毒2（SARS-CoV-2），2019年冠状病毒疾病的原因 自2019年底爆发以来，新冠肺炎疫情已造成全球公共卫生危机。虽然大流行病 最近的工作表明，社会弱势群体受到不成比例的影响， 受到疾病的影响。越来越多的证据表明，这一流行病不成比例地影响到 肤色，年龄较大的人，以及社会经济地位较低的人。然而，迄今为止， 对社会脆弱性与COVID-19结果之间关系的全面时空分析 在一个国家范围内，并在一段较长的时间内，部分原因是这样的统计工具所需的 缺乏分析。该提案的目的是开发多变量模型，以确定时空 相关计数结果的趋势，并使用这些模型来量化COVID-19感染的差异， 死亡、检测、住院和接种疫苗。目标1提出了一个 贝叶斯多变量时空模型，以量化COVID-19感染，死亡，检测， 住院率，以及美国各县随时间的疫苗接种率。社会脆弱性风险是 通过一种新的多变量核机器以非线性和交互的方式并入模型 回归分析目标2通过开发贝叶斯多变量零膨胀设置扩展该方法， 膨胀负二项模型，以量化COVID-19趋势随时间和各县的差异。目的 3开发了计算上可扩展的贝叶斯软件，用于实现这些方法。这一大流行病已 对卫生保健系统造成了持久的破坏，这将对弱势群体造成不成比例的影响。 在未来的岁月里。在此制定的统计方法将在促进卫生公平方面发挥关键作用， 减轻因这一流行病而加剧的长期差距。

PROJECT SUMMARY Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the cause of coronavirus disease 2019 (COVID-19), has created a global public health crisis since its onset in late 2019. Although the pandemic has affected all communities, recent work suggests that socially vulnerable populations have been disproportionately impacted by the disease. Mounting evidence has found that the pandemic disproportionately affects people of color, older individuals, and those of lower socioeconomic status. To date, however, there has been no comprehensive spatiotemporal analysis of the relationship between social vulnerability and COVID-19 outcomes at a national scale and over an extended period of time, in part because the statistical tools needed for such an analysis are lacking. The objective of the proposal is to develop multivariate models to identify spatiotemporal trends in correlated count outcomes, and to use these models to quantify disparities in COVID-19 infection, death, testing, hospitalizations, and vaccinations across socially vulnerable communities. Aim 1 proposes a Bayesian multivariate spatiotemporal model to quantify disparities in COVID-19 infection, death, testing, hospitalization, and vaccination rates over time across US counties. Social vulnerability exposures are incorporated into the model in a nonlinear and interactive manner through a novel multivariate kernel machine regression. Aim 2 extends the method to the zero inflated setting by developing a Bayesian multivariate zero- inflated negative binomial model to quantify disparities in COVID-19 trends over time and across counties. Aim 3 develops computationally scalable Bayesian software for implementation of the methods. The pandemic has caused enduring disruptions to the health care system that will disproportionately impact vulnerable populations for years to come. The statistical methods developed here will play a critical role in promoting health equity and mitigating long-standing disparities exacerbated by the pandemic.