Improving the patient experience of hemodialysis vascular access decision making

项目摘要/摘要 终末期肾病(ESKD)患者，使用血液透析作为肾脏替代方法， 需要动静脉瘘、动静脉移植物或中心静脉导管形式的血管通路 接受维持性血液透析。提供者和患者面临血管通路类型的选择，没有 有足够的证据证明可能的结果。为了克服这一关键障碍，这份R01提案的目标是优化 通过a)开发交互式、基于证据的 血管通路结果指南，结合了血管通路短期和长期结果的预测模型 B)确定在临床医生-患者会诊期间使用指南的最佳做法。 为此，一种新的、大规模的数据源包含关于血管的多机构细粒度数据 通路手术及其短期和长期结果将通过将血管质量倡议联系起来而创建 美国肾脏数据系统注册中心(USRDS)和联邦医疗保险的血管访问注册中心(VQIVAR) 索赔。将通过使用传统的统计方法(例如，逻辑回归， Kaplan-Meier估计)和机器学习方法(例如，贝叶斯网络、随机森林)来预测 对患者有意义的结果(修订程序，重复血管通路手术)，并比较 这些模型使用技术指标(例如敏感度/特异度)。将选出表现最好的型号 并在加州大学洛杉矶分校的当地人群中进行了外部效度测试。 同时，将使用混合方法让患者和提供者利益相关者参与 协作创建和实施拟议的血管通路结果指南，评估： 1)在血管通路决策过程中，首选与临床医生沟通的方式； 2)将指南(包括预测模型)纳入决策过程的最佳方法； 3)对指南的迭代版本感到满意。具体目标是： 目的1设计、评价和检验血液透析血管预测模型的外部效度 访问结果，用于血管访问决策，从链接到的VQIVAR数据生成 USRDS和Medicare使用统计和机器学习方法进行索赔，并在加州大学洛杉矶分校的队列中得到验证 通过模型校准。 目标2通过以下方式确定临床医生-患者血管通路决策交互的最佳实践 使用混合方法，包括个人访谈、直接观察和量化 满意度和偏好量表。 目标3创建和改进基于最佳表现的血管通路结果的交互式指南 在目标1中创建的预后模型，允许根据每个患者的特征进行个性化，通过参与 患者和提供者利益相关者以迭代的方式合并他们的反馈，并得出最终指南。

Project Summary/Abstract Patients with end-stage kidney disease (ESKD), who use hemodialysis as their kidney replacement method, require vascular access in the form of an arteriovenous fistula, arteriovenous graft, or central venous catheter to receive life-sustaining hemodialysis. Providers and patients face selection of a vascular access type without adequate evidence of likely outcomes. To overcome this key barrier, the goal of this R01 proposal is to optimize the patient experience of vascular access decision-making by a) developing an interactive, evidence-based guide to vascular access outcomes that incorporates a prognostic model for short and long-term outcomes of vascular access and b) identifying best practices for utilization of the guide during the clinician-patient encounter. To do so, a novel, large-scale data source that contains multi-institutional granular data regarding vascular access operations and their short and long-term outcomes will be created by linking the Vascular Quality Initiative Vascular Access Registry (VQIVAR) to the United States Renal Data Systems Registry (USRDS) and Medicare claims. Prognostic models will be developed, by using traditional statistical approaches (e.g., logistic regression, Kaplan-Meier estimates) and machine learning methods (e.g., Bayesian networks, random forests) to predict outcomes that are meaningful to patients (revision procedures, repeat vascular access operation), and compare these models using technical metrics (e.g., sensitivity/specificity). The best-performing models will be selected and tested for external validity in a local UCLA population. Simultaneously, a mixed-methods approach will be used to engage patient and provider stakeholders to collaborate in creation and implementation of the proposed guide to vascular access outcomes, assessing the: 1) preferred means of communication with the clinician during the vascular access decision-making encounter; 2) optimal methods for incorporating the guide (including the prognostic model) into the decision-making process; and 3) satisfaction with iterative versions of the guide. The Specific Aims are: Aim 1 Design, evaluate and test the externally validity of the prognostic models for hemodialysis vascular access outcomes, to be used in vascular access decision-making, generated from VQIVAR data linked to USRDS and Medicare claims using statistical and machine learning methods and validated in a UCLA cohort with model calibration. Aim 2 Identify best practices for the clinician-patient vascular access decision-making interaction by using a mixed methods approach that includes individual interviews, direct observation, and quantitative satisfaction and preference scales. Aim 3 Create and refine an interactive guide to vascular access outcomes based on the best-performing prognostic model created in Aim 1, that allows for personalization with each patient’s characteristics, by engaging patient and provider stakeholders in an iterative fashion to incorporate their feedback and arrive at a final guide.