Sharing sensitive data
Although it may not be possible in all cases, it is a good idea to obtain informed consent from the participants in your study to allow for publication of their anonymized data from the research.
Other pages on this site with related information:
Modifying sensitive data for public release
Sensitive data that contain potentially identifying information -- whether it be human subject data or other types of sensitive data -- will likely need to be modified prior to sharing these data with the public. It is important that these modifications are made in order to protect participant confidentiality, the location of endangered wildlife, or for other relevant reasons. However, these modifications may affect the data to the point where reproducibility or additional subsequent research by others is no loner possible. You might consider retaining multiple versions of the data: one that is suitable for public release, and one that is suitable for further research but that is available on a highly restricted basis.
For patient health information (PHI), HIPAA privacy rules provide two methods for de-identification: the expert determination method and the safe harbor method. See the resources listed below for documentation on these methods from the US Department of Health and Human Services, as well as information on how to satisfying these two methods.
Types of identifying information
Identifying information is classified as one of two types: direct and indirect.
These data point directly to an individual and are typically removed from data sets before sharing with the public.
These may include:
- mailing address
- phone number
- email address
- unique identifying numbers, like Social Security numbers or driver's license numbers
- vehicle identifiers
- medical device identifiers
- web or IP addresses
- biometric data
- photographs of the person
- audio recordings
- names of relatives
- dates specific to individual, like date of birth, marriage, etc.
These may seem harmless on their own, but can point to an individual when combined with other data. It has been recommended (see BMJ article reference below) that datasets containing three or more indirect identifiers should be reviewed by an independent researcher or ethics committee to evaluate identification risk. Any indirect information not needed for the analysis should be removed. It may be reasonable to supply some of these types of data in aggregated form (like ranges of annual incomes instead of exact numbers).
Indirect identifiers may include:
- place of medical treatment or doctor's name
- rare disease or treatment
- sensitive data like illicit drug use or other "risky behaviors"
- place of birth
- socioeconomic data, like workplace, occupation, annual income, education, etc
- general geographic indicators, like postal code of residence
- household and family composition
- birth year or age
- verbatim responses or transcripts
"Guidance Regarding Mehtods for De-identification of Protected Health Information in Accordance with the Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule," US Department of Health and Human Services, Office for Civil Rights.
Hrynaszkiewicz, I, Norton, ML, Vickers, AJ and Altman, DG. "Preparing raw clinical data for publication: guidance for journal editors, authors, and peer reviewers." BMJ 2010;340:c181.
"Preparing Data for Sharing" from the Inter-University Consortium for Political and Social Research (ICPSR). (2012). Guide to Social Science Data Preparation and Archiving: Best Practice Throughout the Data Life Cycle (5th ed.). Ann Arbor, MI.