RCMI@Morgan: Center for Urban Health Disparities Research and Innovation

项目摘要 非洲血统健康差异的表征和算法偏差的减少 在大数据时代，算法在社会上随处可见，将其转变为更个性化和 奉承。一个主要的趋同是算法在生物、医学和医疗保健方面的应用。 然而，最近的一项研究表明，一种广泛使用的算法，这种全行业方法的典型和 影响数百万患者，表现出显著的种族偏见：在给定的风险分数下，黑人患者 比白人病人病情严重得多，表现为疾病失控的迹象。这是一个特定的 一个被称为算法偏差的更广泛问题的例子，其中算法恢复了文化偏见 在他们接受训练的数据集中进行编码。越来越多的人意识到算法偏差的影响，导致 以响应对算法责任的相应呼吁。Caplan等人。定义“最终的算法责任” 指如何创建算法及其对社会的影响的责任分配；如果损害 发生时，问责制度包括一个补救机制“(2018：10)，即算法问责 在算法的设计及其效果中都包含了减少危害和预防方面的考虑。 这些“算法偏差”和“算法问责”问题也涉及非洲的健康差距。 在家长RCMI奖中暴露的祖先有以下目标：(1)增强密歇根州立大学的健康差距 基础生物医学和行为/公共卫生研究方面的研究基础设施和能力，(2) 加强关于城市健康和健康差距的高质量研究，包括转化研究，通过 增加对社区的外部资金、出版物和科学服务，(3)促进合作 在基础生物医学和社会/行为学院研究人员之间建立协作和支持 教师职业发展的环境，特别是新教师和早期教师的职业发展和(4)建设 与约翰·霍普金斯大学和约翰·霍普金斯大学这两个研究密集型机构建立可持续的伙伴关系 马里兰州，巴尔的摩，以及当地政府和社区组织处理卫生问题 差距。 在本项目的目标1中，将重新定义父RCMI奖的当前指标，并提供以下统计数据 性别、种族和社会经济地位光谱中的概率分布空间。潜在的数据偏差 在去辨认之前，将被过滤掉，以表征“算法偏差”。基于重新定义的指标 目标1，考虑到人群的广泛光谱，人工智能技术将朝着精准医学方向发展 和血统，以减少已识别的偏见。具体来说，将开发推理和学习算法 基于正则化贝叶斯优化的元学习概率空间。 我们将举办公开研讨会和工作坊，提供案例研究，涵盖以下主题： 伦理、人工智能和健康差距。将公开分发分析框架和工具，并将 “我们所有人”的公共门户分析的可扩展性和可集成性。

Project Summary Characterization of health disparities in African ancestry and reduction of algorithmic bias In the epoch of big data, algorithms are present throughout society, transforming it into more personalized and flatter. One primary convergence is the application of algorithms for biology, medicine, and health care. Nonetheless, a recent study shows that a widely used algorithm, typical of this industry-wide approach and affecting millions of patients, exhibits significant racial bias: At a given risk score, Black patients are considerably sicker than White patients, as evidenced by signs of uncontrolled illnesses. This is a specific example of a broader issue known as algorithmic bias, wherein algorithms reinstate the cultural biases encoded in the data sets they are trained on. Increasing appreciation for the impact of algorithmic bias has led to a corresponding call for algorithmic accountability. Caplan et al. define "Algorithmic accountability ultimately refers to the assignment of responsibility for how an algorithm is created and its impact on society; if harm occurs, accountable systems include a mechanism for redress" (2018: 10). That is, algorithmic accountability includes harm reduction and prevention considerations in both the design of the algorithm and its effects. These issues of “algorithmic bias” and “algorithmic accountability” also involve in health disparities in African ancestry exposed in the parent RCMI award with following goals: (1) Enhance MSU's health disparities research infrastructure and capacity in both basic biomedical and behavioral/public health research, (2) Enhance high-quality research, including translational research, on urban health and health disparities through increased external funding, publications and scientific services to the community, (3) Facilitate collaborations between basic biomedical and social/behavioral faculty researchers and create a collaborative and supportive environment for faculty career development, especially for new and early career faculty and (4) Build sustainable partnerships with two research-intensive institutions, Johns Hopkins University and the University of Maryland, Baltimore, as well as local government and community-based organizations dealing with health disparities. In the Aim 1 for this project, the current metrics of the parent RCMI award will be redefined with statistics on the probability distribution space in the spectra of gender, race, and socioeconomic status. Potential data bias in de-identification will be filtered out before characterizing “algorithmic bias.” Based on the redefined metrics of Aim 1, AI techniques will be advanced toward precision medicine considering the broad spectra of population and ancestry to reduce the identified biases. Specifically, inference and learning algorithms will be developed on probability spaces with meta-learning via Bayesian optimization with regularization. We will host public seminars and workshops with case studies covering the topics in the interaction among ethics, AI, and health disparities. The analysis framework and tool will be publicly disseminated and will have extensibility and integrability with the public portal analytics of "All of Us."