Data Sharing and Privacy Protection in Digital-Age Environmental Health Studies

描述（由申请人提供）：用于化学生物监测和个人特定采样（如在家中或呼吸空间）的新环境暴露技术是发现疾病环境原因和指导预防的关键进展。这些方法为主要的NIH队列研究和监测提供了有价值的暴露数据，特别是美国人口中200多种化学品的NHANES生物监测和新的EPA ExpoCast。个人暴露测量是昂贵的，有时是不可重复的，激励开放的数据共享，包括在线存储库。与此同时，个人测量引发了新的伦理问题，即即使在被认为是匿名的数据中，研究参与者的身份也可能被泄露，这一过程称为重新识别。公共数据和计算能力的增强将继续增加隐私风险。受到明显的重新识别案例的影响，医疗保健和遗传学研究人员等正在讨论和调查编辑共享数据的新做法，警告参与者隐私风险，有时要求“公开同意”在不保护隐私的情况下共享数据。然而，计算隐私风险尚未调查环境化学品暴露数据。这些数据可能会通过使用诸如真实的房地产、环境合规、许可证、天气和消费者购买等数据的新的联系战略带来风险。重新识别可能导致对“受污染”的个人或社区的污名化，揭示一个人认为隐私的行为（例如，吸烟或使用美国禁止的海外产品），如果测量受管制的化学品，则触发法律的义务，或影响财产价值、保险或就业能力。这个项目将经验性地评估隐私 风险，并为环境健康研究制定解决方案。它将邀请环境卫生科学家、计算机科学家、决策者、社区领袖和生物伦理学家组成的咨询理事会就复杂的伦理和价值观问题提供意见并寻求共识。基于研究者建立的健康数据计算模型，本研究将开发一个模型，用于预测环境健康研究中的再识别风险。通过将该模型应用于10项重要的环境研究，该项目将量化该领域的隐私风险，并确定导致风险的特定数据字段。该模型将通过测试家庭暴露研究中重新识别的实际数量来验证。根据显示风险数据字段的结果，该研究将测试并寻求优化编辑或屏蔽数据的程序，以改善隐私，同时保留数据共享的科学效用。由于数据共享的决定最终取决于参与者的知情同意，该项目通过询问两项大型创新在线研究的参与者对隐私和数据共享的理解和价值观来补充计算分析。该项目的结果将为研究人员提供道德和技术上合理的方法来共享环境数据，有助于更快地发现可预防的疾病原因。

DESCRIPTION (provided by applicant): New environmental exposure technologies for chemical biomonitoring and person-specific sampling, such as in homes or breathing space, are crucial advances for finding environmental causes of disease and guiding prevention. These methods yield valuable exposure data for major NIH cohort studies and for surveillance, notably NHANES biomonitoring for more than 200 chemicals in the US population and the new EPA ExpoCast. Personal exposure measurements are expensive and sometimes non-repeatable, motivating open data- sharing, including in online repositories. At the same time, personal measurements raise new ethical concerns about the possibility that the identity of study participants might be revealed even in data considered anonymized, a process called re-identification. Proliferating public data and computing power will continue to increase privacy risks. Prompted by visible instances of re-identification, healthcare and genetics researchers, among others, are debating and investigating new practices for redacting shared data, warning participants of privacy risks, and sometimes requesting "open consent" to share data without protecting privacy. However, computational privacy risks have not yet been investigated for environmental chemical exposure data. These data may pose risks through novel linkage strategies using data such as on real estate, environmental compliance, permits, weather, and consumer purchases. Re-identification could result in stigma for "contaminated" individuals or communities, reveal behavior a person considers private (e.g., smoking or use of overseas products banned in the US), trigger legal obligations if a regulated chemical is measured, or affect property values, insurance, or employability. This project will empirically evaluate privacy risks and develop solutions for environmental health studies. It will engage an Advisory Council of environmental health scientists, computer scientists, policymakers, community leaders, and bioethicists to provide input and seek consensus on complex ethical and values-based considerations. Building on the investigators' established computational model for health data, this study will develop a model for predicting re-identification risk in environmental health studies. By applying the model to 10 important environmental studies, the project will quantify privacy risks in this field and identify specific data fields that contribute to risk. The model wil be validated by testing the actual number of re-identifications in a household exposure study. Based on results indicating risky data fields, the study will test and seek to optimize procedures to redact or mask data to improve privacy while retaining scientific utility for data-sharing. Because data-sharing decisions ultimately rest on participants' informed consent, the project complements computational analyses by asking participants in two large, innovative online studies about their understandings and values related to privacy and data-sharing. Results from this project will provide researchers with ethically and technically sound methods for sharing environmental data, contributing to more-rapid discovery of preventable causes of disease.