Rectifying Disparities Associated with Risk Score-Based MTM Eligibility in Enhanced MTM

医疗保险和医疗补助服务中心(CMS)一直在测试一项为期5年的示范计划， 自2017年起加强药物治疗管理(MTM)，以提升传统的Medicare MTM 方案，其资格标准在公平性和有效性方面表现不佳，正如部分报告所述 由我们的研究团队提供。为了促进市场创新，CMS特意批准了处方药(部分 D)计划完全自由地在加强的MTM中制定新的MTM资格标准。虽然以前的文学作品 据报道，通常通过机器学习产生的风险分数可能会使历史上的种族/民族永久化 健康结果的差异，所有参加强化MTM的计划都使用了MTM资格中的风险评分 决心。解决基于风险分数的MTM资格所产生的潜在差异的一个关键障碍是 所有利益攸关方都对这类倡议的差异影响保持沉默。我们提议的项目 将探索战略，以确定和解决与基于风险分数的MTM资格相关的差异 增强的MTM。我们的长期目标是改善不同老年人口的健康状况 通过减少药物使用方面的差异。我们的目标是主要分析100%的医疗保险部分 A/B/D数据(2018-2021年；之前两年和新冠肺炎期间两年)，以说明COVID的影响- 19)。我们的团队在索赔数据、差异、MTM和机器学习方面拥有丰富的经验。这项研究 结果将是MTM资格和患者健康结果。目标1.确定种族/民族差异 与基于风险分数的MTM资格相关联。我们将首先使用常规的机器学习算法和 使用与改进的MTM相同的模型产出/结果。然后我们将修改机器学习 通过实验替代模型输出、应用深度迁移学习和预处理的算法 解决潜在差距的数据。MTM资格的差异将在常规/修改后进行比较 机器学习和传统的医疗保险MTM计划。目标2.确定种族/民族的影响 基于风险评分的MTM资格与患者健康结果的差异。健康结果的差异更大 在不符合MTM资格的人中，MTM合格的个人将建议MTM资格标准 评估可能会导致更严重的差距。目的3.确定MTM的相对有效性 基于跨种族/民族群体的常规和经修改的机器学习算法的资格标准。这个 MTM资格的有效性将由被认为符合MTM资格的个人的比例来衡量 在有药物利用问题的患者中。目标2和目标3的成功与目标1无关，因为 由于与风险相关的潜在种族/民族差异得到了很好的记录，因此确保了目标1的成功 得分。影响：我们的项目可以防止基于风险分数的MTM资格在以下方面造成更大差异 医疗保险和其他医疗保险中的药物利用。这项研究将有助于NIA实现其 “了解差距，制定改善老年人健康状况的策略。”

The Centers for Medicare & Medicaid Services (CMS) has been testing a 5-year demonstration program, Enhanced Medication Therapy Management (MTM), since 2017 to ramp up the traditional Medicare MTM program, whose eligibility criteria have underperformed in equity and effectiveness as has been reported partly by our research team. To foster marketplace innovation, CMS purposefully granted the prescription drug (Part D) plans complete liberty to devise new MTM eligibility criteria in Enhanced MTM. Although previous literature reported that risk scores often generated through machine learning can perpetuate historical racial/ethnic disparities in health outcomes, all participating plans in Enhanced MTM have used risk scores in MTM eligibility determination. A critical barrier to tackling potential disparities arising from risk score-based MTM eligibility is that all stakeholders have been silent about the disparity implications of such initiatives. Our proposed project will explore strategies to identify and resolve disparities associated with risk score-based MTM eligibility in Enhanced MTM. Our long-term goal is to improve health outcomes among the diverse older adult population by reducing disparities in medication utilization. Our objective is to mainly analyze 100% Medicare Parts A/B/D data (2018-2021; two years before and two years during COVID-19 to account for the impact of COVID- 19). Our team has extensive experience in claims data, disparities, MTM, and machine learning. The study outcomes will be MTM eligibility and patient health outcomes. Aim 1. To determine racial/ethnic disparities associated with risk score-based MTM eligibility. We will first employ regular machine learning algorithms and use the same model outputs/outcomes as in Enhanced MTM. We will then modify the machine learning algorithms by experimenting with alternative model outputs, applying deep transfer learning, and preprocessing data to address potential disparities. Disparities in MTM eligibility will be compared between regular/modified machine learning and the traditional Medicare MTM program. Aim 2. To determine implications of racial/ethnic disparities in risk score-based MTM eligibility on patient health outcomes. Higher disparities in health outcomes among the MTM-ineligible than MTM-eligible individuals would suggest that the MTM eligibility criteria assessed may lead to worse disparities. Aim 3. To determine the comparative effectiveness of the MTM eligibility criteria based on regular and modified machine learning algorithms across racial/ethnic groups. The effectiveness of the MTM eligibility will be measured by the proportion of individuals deemed MTM-eligible among patients with medication utilization issues. Successes of Aims 2 & 3 are independent of Aim 1 because the success of Aim 1 is ensured due to well-documented potential racial/ethnic disparities associated with risk scores. Impact: Our project can prevent risk score-based MTM eligibility from causing more disparities in medication utilization in Medicare and beyond. This study will assist NIA in achieving its strategic goals of “understand[ing] disparities and develop(ing) strategies to improve the health status of older adults.”