DDRIG in DRMS: The implications of algorithmic decision-making for inequality in long-term care

各州越来越多地使用算法来确定美国老年人和残疾人是否有资格享受医疗补助资助的长期医疗保健。与一般的医疗补助不同，长期护理的资格除了收入和资产外，还取决于功能能力和残疾状况。以前，各州可能只依靠医生来决定长期护理申请者是否在功能上有资格，但现在许多州使用算法来做出这一决定。尽管人们经常声称算法没有人类那么有偏见，但研究表明，算法往往会加剧不平等，特别是人口统计类别之间的不平等。然而，不平等的算法决策在人口层面的后果仍然不清楚。医疗补助长期护理是解决这一证据差距的重要案例；医疗补助是数百万美国人，特别是边缘群体成员的关键医疗保健来源。此外，未来几十年，随着老年人口的增长，需要长期护理的人数可能会增加，这使得这一问题变得特别及时。本文使用混合方法回答了三个主要的研究问题：1)州医疗补助计划如何以及为什么在长期护理资格确定中使用算法？2)医疗补助官僚如何使用算法来执行长期护理资格评估？3)算法实施与长期护理获取中人口统计类别之间的不平等之间有什么关系？通过将医疗补助索赔数据与每个州使用的算法的原始数据联系起来，该项目估计了算法实施如何与长期护理中的不平等相关(RQ3)。对三个州的医疗补助管理人员的深入采访阐明了这种关系背后的机制(RQ1和RQ2)。结果有助于确定组织层面的健康社会决定因素，这些因素可能会被改变以改善公平。此外，为回答RQ3而收集的主要数据将公开，以使未来能够对长期护理中的算法进行研究，并使医疗补助长期护理受益人更好地了解他们的福利是如何确定的。该奖项反映了NSF的法定使命，并已通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

States are increasingly using algorithms to determine eligibility for Medicaid-funded long-term healthcare for elderly people and people with disabilities in the United States. As opposed to Medicaid more generally, eligibility for long-term care is based on functional capacity and disability status in addition to income and assets. Previously, states may have relied solely on physicians to decide whether a long-term care applicant is functionally eligible, but many states now use algorithms to make this determination. Though algorithms are often purported to be less biased than humans, research has shown that algorithms often increase inequality, especially inequality between demographic categories. However, the population-level consequences of algorithmic decision-making for inequality are still unclear. Medicaid long-term care is an important case to address this evidence gap; Medicaid is a crucial source of healthcare for millions of Americans, especially for members of marginalized groups. Additionally, the number of people who require long-term care will likely increase as the elderly population grows over the next few decades, making this issue particularly timely. This research sheds light on how algorithmic decision-making, and which characteristics of algorithms, are better or worse for equity in access to long-term care.This dissertation uses mixed methods to answer three primary research questions: 1) How and why do state Medicaid programs use algorithms in long-term care eligibility determination? 2) How do Medicaid bureaucrats use algorithms to perform long-term care eligibility assessments? 3) What is the relationship between algorithm implementation and inequality between demographic categories in long-term care access? By linking Medicaid claims data with primary data on the algorithms used by each state, this project estimates how algorithm implementation is associated with inequality in long-term care (RQ3). In-depth interviews with Medicaid administrators in three states elucidate the mechanisms underlying this relationship (RQ1 and RQ2). Results help identify organization-level social determinants of health that can potentially be altered to improve equity. Further, the primary data collected to answer RQ3 will be made publicly available to enable future research on algorithms in long-term care and give Medicaid long-term care beneficiaries better insight into how their benefits are determined.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.