06- Responsible sharing to minimize harmful secondary use of large datasets

Nearly all federally funded human genome studies must broadly share their data. To promote secondary use, funders typically want participants’ consent to share data for any type of future research, without restriction. This paper develops and defends the claims that 1) there are technical and governance mechanisms for minimizing the likelihood of harmful or unjust secondary data use; and 2) responsible data sharing requires researchers and funders to employ such mechanisms.
Harmful or unjust secondary uses can occur because secondary users are not well-versed in the dataset or the proper analytical methods for working with the dataset, or because secondary users have pernicious socio-political agendas. During research design, producers of shareable data should structure data collection and datasets so commonly used analysis methods will likely produce valid results. They should consider secondary uses when deciding how to include demographic information in the released dataset and what non-genetic data to collect. Harm-minimization also requires controlled access to individual-level data using well-structured data use agreements; revocation of access for researchers who violate a data use agreement; and mandatory education for secondary users about the dataset’s production, features, and limitations. Data producers can provide precomputed analyses that use appropriate mathematical models and include all relevant variables, and they can provide validated, transparent analysis pipelines that make relevant variables easily accessible to users. This paper will discuss additional harm-reduction measures and will argue that researchers and funders have ethical obligations to minimize harm and injustice to participants and their communities by deploying such measures.