We are collecting medical observational data that eventually will be de-identified and made available for secondary data analysis (to comply with the new NIH Data Management & Sharing Policy). I'd like to find a data repository that requires potential researchers to submit a DUA (data use agreement) to the repository's management. Ideally the DUA specifies restrictions like (a) the recipient will not attempt to re-identify the participants, (b) the recipient will not distribute the dataset, and (c) the dataset will be used for only a narrowly-defined research question.

In a previous study, we selected the NDACAN (National Data Archive on Child Abuse and Neglect; it specializes on these types of datasets) and requires an DUA from secondary researchers. I love how the repo assumes responsibility for determining if a secondary data analysis has a justifiable reason to obtain the dataset.

I'm on the university's CRDW (clinical research data warehouse) team, and we're involved in a lot of research that doesn't fit NDACAN's narrow mission and I'd like to find more options. It's been difficult to find something similar. I'm a big fan of open data repositories like Dataverse and OSF, but since they can't require DUAs, I don't like them for granular medical data. I'm comfortable stripping out the ~18 explicit variables, but I'm concerned that (current or future) techniques will be able to re-identify patients from details dataset. So that's why I want a repository with oversight and DUAs.

I reviewed the 9 generalist repositories, individually ruled out each of the 137 domain-specific repositories on NIH's Selecting a Data Repository page, and read the 8 health sciences repos suggested by Nature. I can't find a repo that accepts general medical research and requires a DUA, but I hope I'm simply overlooking one.

Two repos are close to my criteria: NDA (but we're not funded by the NIMH) and Synapse (but the budget can't afford $30k+).

Does anyone have either (1) specific suggestions for conventional NIH-funded medical research or (2) strategies to identify appropriate data repositories (for research that initially included PHI)?


You must log in to answer this question.

Browse other questions tagged .