Geospatial modeling for stroke care

项目总结 在美国，急性缺血性中风(AIS)仍然是导致残疾的主要原因。大血管闭塞(LVO) 占所有缺血性中风的20%，但导致90%的中风相关死亡和严重残疾。 静脉溶栓(IVT)和血管内治疗(EVT)均为有效的时效治疗方法 预防中风相关的发病率和死亡率。EVT对LVO非常有效，在以下方面不提供任何好处 非LVO卒中，在不到20%的美国卒中中心可用。IVT随手可得，具有适度的 对左心室关闭不全有效，是治疗非左心室关闭不全中风的唯一选择。护理人员面临的挑战是 加快符合条件的患者的EVT，而不会损害需要IVT的不合格患者的大部分 最初诊断不确定的背景。目前的系统分类标准已经落后于新兴市场 治疗方法适用于病情最严重的人群，中风预后的差异在农村地区和 为少数族裔服务。 在此，我们提出了一项研究，旨在促进卒中地理空间分类算法的创新发展 通过多学科合作在美国卫生系统提供护理，以最大限度地促进所有人的神经康复 中风患者。该模型将被构建为为个别患者提供最佳预测结果， 使用贝叶斯框架通知治疗决策树的每个链接，同时建立在先前研究的基础上 克服它们的局限性，缩小执行差距。首先，将建立患者结局模型 使用个人和医院水平的数据随机试验，这将使上下文敏感的分诊决策成为可能 算法，而不依赖于对治疗路径的过于宽泛的假设。我们将独一无二地 通过在贝叶斯框架中对个别水平的数据进行建模来纳入不确定性，而不是依赖 关于总水平上的点数估计。此外，我们的模型将是可适应的；我们将能够 结合新的LVO诊断工具，提高诊断准确率，以及新的治疗方法 随着卒中领域的发展而变化的策略。此外，条件结构将允许修改设施 能力，包括引入新的支持EVT的卒中中心。临床和成本效益 算法的影响将通过与当前现实世界中结合局部卒中的分流方法进行比较来评估 中心和支持EVT的中心将卒中流指标的数据从Get-With-Stroke注册表集中到 更好地估计良好结果的可能性并提高分诊能力。最后，分类算法 将集成到随时可供所有EMS使用的护理点决策工具支持中，以建议 在他们的初步评估之后，所有中风人群的最佳目的地。经过适当的提炼后 并在后续研究中充分实施，该工具不仅具有优化行程的潜力 结果，但也缩小了美国实际的地理和种族差距。

PROJECT SUMMARY Acute Ischemic stroke (AIS) remains the leading cause of disability in the US. Large vessel occlusion (LVO) represents up to 20% of all ischemic strokes, but causes 90% of stroke-related death and severe disability. Both intravenous thrombolysis (IVT) and endovascular therapy (EVT) are effective time-sensitive treatments to prevent stroke-related morbidity and mortality. EVT is highly effective for LVOs, does not provide any benefit in non-LVO strokes and is available in less than 20% of US stroke centers. IVT is readily available, has a modest effect for LVOs and is the only therapeutic alternative for non-LVO strokes. The challenge for paramedics is to expedite EVT for eligible patients without harming a large proportion of non-qualified patients in need of IVT, in the context of initial diagnostic uncertainty. The current system triage criteria have lagged behind emerging therapies available to the sickest subset, and the disparity in stroke outcomes is exacerbated in rural areas and for ethnic minorities. Herein, we propose a study to foster the development of an innovative geospatial triage algorithm of stroke care in the U.S. health system through a multidisciplinary collaboration to maximize neurological recovery to all stroke patients. The model will be constructed to provide optimal predicted outcomes for individual patients, using a Bayesian framework to inform each link of the treatment decision tree, building on prior studies while overcoming their limitations and closing the implementation gap. First, the patient outcome model will be built using individual and hospital level data randomized trials, which will enable a context sensitive triage decision algorithm without reliance on overbroad assumptions about the treatment pathway. We will uniquely incorporate uncertainty through modelling of individual level data in a Bayesian framework, rather than relying on point estimates at an aggregate level. Additionally, our model will be adaptable; we will be able to incorporate emerging LVO diagnostic tools with improved diagnostic accuracy, as well as new therapeutic strategies as the stroke field evolves. Furthermore, the conditional structure will allow the modification of facility capabilities, including the introduction of new EVT-capable stroke centers. The clinical and cost-benefit algorithm impact will be assessed by comparing with the current real-world triage by incorporating local stroke center and EVT-capable center data on stroke flow metrics from Get-With-The Guidelines-Stroke registry to better estimate the probability of good outcomes and improve triage capabilities. Finally, the triage algorithm will be integrated into a point-of-care decision tool support readily available for all EMS to recommend the optimal destination for all the entire stroke population after their initial assessment. After appropriate refinement and adequate implementation in subsequent studies, this tool will not only have the potential to optimize stroke outcomes, but also reduce the actual geographic and racial disparities in the U.S.