An ethical framework-guided metric tool for assessing bias in EHR-based Big Data studies

摘要 大数据健康研究的出现使医学和公共卫生领域呈指数级发展 但也面临着许多道德挑战。道德领域最令人担忧但研究不足的方面之一 问题是数据集中潜在偏差的风险（例如，电子健康记录[EHR]数据）以及 策展和收购周期。很少有基于EHR数据的研究报告数据集、数据采集 和/或由于缺乏标准化的测量工具而将采矿作为研究质量的指标， 评估偏见的指标;作为理论基础的伦理框架很少;有效的跨学科 与道德专家、专业数据管理者、数据管理专家、 存储库管理员，医疗保健工作者和国家机构在讨论解决这一道德问题， 挑战.自2021年以来，我们一直受到NIH（R 01 AI 164947）的资助，开发基于机器学习的 基于EHR和其他来自多个国家的相关数据， 南卡罗来纳州的线人母项目遇到的道德挑战之一是如何评估 EHR数据的管理、获取和处理中的潜在偏差。响应NOT-OD-22-065 标题为“推进生物医学和生物医学领域AI/ML的伦理发展和使用的行政补充材料”， 行为科学”，我们建议开发，完善和试点测试道德框架指导的度量工具， 使用EHR数据集评估大数据研究中的偏见。具体而言，我们请求支持：1）开展 进行文献/政策审查和概念分析，以制定公正和包容性的道德框架， 数据研究; 2）创建和修改一个度量工具，以评估EHR数据为基础的研究中的潜在偏见，通过在 对母项目的主要利益相关者进行深入访谈;以及3）完善和传播度量工具 通过跨学科学者（伦理学专家和学科专家）之间的社区Charette讲习班， 专家）和关键利益相关者（数据管理员、数据管理专家和数据存储库管理员）; 医护人员和艾滋病患者），并在母项目中进行试点测试。拟议的研究将推进 我们对大数据研究中的偏见和公平问题的理解，并制定道德框架和 用于评估基于EHR的大数据研究中的偏差的度量工具，从而导致并提供更细致入微的 评估和探索实践中的偏见，以促进大数据健康研究的伦理发展， 父项目。用于大数据研究的偏倚度量工具可重复用作评估工具， 并量化偏见，这可能有助于提高对这一关键伦理问题的认识和探索。 挑战.关于大数据研究中偏见挑战的道德框架可以提供见解， 解决EHR以外其他类型大数据中的偏见问题的指南。

Abstract The emergence of Big Data health research has exponentially advanced the fields of medicine and public health but has also faced many ethical challenges. One of most worrying but still under-researched aspects of ethical issues is the risk of potential biases in datasets (e.g., electronic health records [EHR] data) as well as in the data curation and acquisition cycles. Very few EHR data-based studies report bias in datasets, data acquisition and/or mining as an indicator of research quality because of a lack of a standardized measurement tool or metrics to assess bias; few ethical frameworks as a theoretical ground; and limited effective interdisciplinary collaboration that engages ethical experts, professional data curators, data management experts, data repository administrators, healthcare workers, and state agencies in discussions addressing this ethical challenge. Since 2021, we have been funded by NIH (R01AI164947) to develop a machine-learning based predictive model of viral suppression among HIV patients based on EHR and other relevant data from multiple sources in South Carolina. One of the ethical challenges encountered by the parent project is how to assess the potential biases in the curation, acquisition, and processing of EHR data. In response to the NOT-OD-22-065 titled “Administrative supplements for advancing the ethical development and use of AI/ML in biomedical and behavioral sciences”, we propose to develop, refine, and pilot test an ethical framework-guided metric tool for assessing bias in Big Data research using EHR datasets. Specifically, we request support to: 1) conduct a literature/policy review and concept analysis to develop an ethical framework for unbiased and inclusive Big Data research; 2) create and modify a metric tool to assess potential biases in EHR data-based studies via in- depth interviews of key stakeholders of the parent project; and 3) refine and disseminate the metric tool through a community charette workshop among interdisciplinary scholars (ethics experts and disciplinary experts) and key stakeholders (data curators, data management experts, and data repository administrators; healthcare workers; and HIV patients) and pilot test it in the parent project. The proposed study will advance our understanding of bias and equity issues in Big Data research and develop an ethical framework and a metric tool for assessing bias in EHR-based Big Data studies, thus leading to and informing a more nuanced assessment and exploration of bias in practice for the ethical development of Big Data health research beyond the parent project. The metric tool of bias for a Big Data study can be reused as an assessment tool to detect and quantify biases, which may contribute to improving awareness and exploration of this critical ethical challenge. The ethical framework regarding bias challenges in Big Data research may provide insights and guidance for addressing bias issues in other types of Big Data beyond EHR.