Building Equity Improvement into Quality Improvement in the use of New Glucose-lowering Drugs (GLDs) through Individualized Drug Value Assessment in People with Diabetes

项目摘要 自2007年以来，已有40多种降糖药物获得美国食品和药物管理局(FDA)的批准 用于治疗糖尿病的药物。这些较新的GLD已被证明对心脏肾脏的益处高于 当应用于心血管疾病和肾脏疾病的高危人群时，更老的类。然而， 这些高成本GLD的引入在糖尿病治疗中引发了重大的质量和公平问题：社交 由于与社会相关的障碍，处于不利地位的个人获得较新的GLD的机会往往有限 属性(例如，收入、教育)，导致在实现最佳健康结果方面存在差距和差距。 因此，迫切需要提高护理质量和公平使用较新的全球初级保健服务 数百万患有2型糖尿病(T2D)的美国人。 先前的研究发现，通过促进有针对性的治疗来提高护理质量的方案 临床上高收益的用户群体会带来公平性的提升，因为社交网络中的高收益用户 弱势群体往往在护理方面有更大的差距，因此从这些方案中受益更多。然而， 在确定新的GLD的临床高收益用户和制定政策方面存在关键的知识缺口- 水平干预措施，可以充分激励患者使用新的GLD，同时保持良好的长期健康 和经济结果。因此，这项拟议项目的目标是确定临床上的高收益 针对较新的GLD的T2D患者亚组，并为设计政策层面生成经验经济证据 旨在改善T2D护理的护理质量和卫生公平的干预措施。 高质量比较有效性研究(CER)要求患者有完整的数据记录 它可以高精度地跟踪事件遭遇和治疗暴露。这些人经常是 被称为“忠诚的病人”。在这个提议的项目中，我们将为“忠诚”开发一个可计算的表型(CP 患者“使用One佛罗里达电子病历和跨网络使用REACHnet电子病历验证CP(目标1)。至 为较新的GLD确定临床高收益的T2D患者亚组，我们将进行比较 新的GLDs与指南推荐的替代方案的有效性和安全性分析 使用严格的因果推理方法和机器学习(ML)方法的子群。高效益 T2D患者亚组将使用来自One佛罗里达州的“忠诚患者”的EHR进行识别，并在 REACHnet(目标2)。最后，我们将评估潜在的政策层面干预的影响，以促进 较新的GLD在健康、经济和公平结果方面的高收益用户中使用。利用先进的 由PI开发的ML算法，我们还将在健康计划层面确定理想的费用分担结构，以 最大限度地遵守药物规定，同时减轻付款人的负担。 这项拟议的研究具有重要意义，因为它将为 糖尿病管理中的护理质量和健康公平。这项研究具有创新性，因为我们将使用切割- EDGE机器学习方法、模拟模型、辅助变量和两个最大的PCORnet EHR解决一个具有挑战性和创新性的研究问题。

Project Summary Since 2007, more than 40 glucose-lowering drugs (GLDs) have been approved by the US Food and Drug Administration to treat diabetes. These newer GLDs have been proven to have higher cardiorenal benefits than older classes when applied in people at high risk of cardiovascular and kidney disease. However, the introduction of these high-cost GLDs has led to significant quality and equity concerns in diabetes care: socially disadvantaged individuals tend to have limited access to newer GLDs due to barriers related to social attributes (e.g., income, education), resulting in gaps and disparities in achieving optimal health outcomes. There is, therefore, an urgent need to improve the quality of care and equity in using newer GLDs among millions of Americans living with type 2 diabetes (T2D). Previous studies have found that programs that improve the quality of care by promoting treatment in targeted clinically high-benefit user groups lead to equity improvement because high-benefit users from socially disadvantaged subgroups often have larger gaps in care thus benefit more from these programs. However, critical knowledge gaps exist in identifying the clinically high-benefit users of newer GLDs and designing policy- level interventions that can adequately motivate patients’ newer GLDs use while having good long-term health and economic outcomes. Thus, the OBJECTIVE of this proposed project is to identify clinically high-benefit T2D patient subgroups for newer GLDs and generate empirical economic evidence for designing policy-level interventions to improve the quality of care and health equity in T2D care. High-quality comparative effectiveness research (CER) requires the patients to have complete data records which can track event encounters and treatment exposure with high accuracy. These individuals were often referred to as “loyal patients.” In this proposed project, we will develop a computable phenotype (CP) for “loyal patients” using OneFlorida EHRs and cross-network validate the CP using REACHnet EHRs (Aim 1). To identify clinically high-benefit T2D patient subgroups for newer GLDs, we will conduct comparative effectiveness and safety analyses of newer GLDs versus guideline-recommended alternatives across patient subgroups using rigorous causal inference methods and a machine-learning (ML) approach. The high-benefit T2D patient subgroups will be identified using EHRs of “loyal patients” from OneFlorida and cross-validated in REACHnet (Aim 2). At last, we will evaluate the impact of potential policy-level interventions for promoting newer GLDs use in high-benefit users on health, economics, and equity outcomes. Leveraging an advanced ML algorithm developed by PI, we will also identify the ideal cost-sharing structure at a health-plan level to maximize drug adherence while reducing the payers' burden. The proposed research is significant because it will provide solutions for an emergent public health issue in quality of care and health equity in diabetes management. This study is innovative because we will use cutting- edge machine-learning methods, simulation models, instrumental variables, and two of the largest PCORnet EHRs to tackle a challenging and innovative research question.