Survey of Investigators About Sharing Human Research Data in the Neurosciences

Abstract

Background and objectives: In the neurosciences, significant opportunities for sharing individual-level data are underexploited. Commentators suggest various barriers to data sharing, which may need to be addressed. Investigators' perspectives on the main barriers are unclear. Furthermore, bioethicists have raised concerns about the potential misuse of neuroscience data, although discussions are hampered by uncertainty about the potential risks. It is unclear how common sensitive data are obtained and whether investigators judge them as sensitive.

Methods: An online survey was disseminated among 1,190 principal investigators (PIs) of active National Institute of Neurological Disorders and Stroke, National Institute of Mental Health, or NIH Brain Research Through Advancing Innovative Neurotechnologies Initiative grants involving human subject research.

Results: A total of 397 investigators responded to the survey (response rate 33%). Most investigators (84%) support efforts to increase sharing of deidentified individual-level data. However, investigators perceive many barriers to data sharing. The largest barriers were costs and time; limited interpretation of the data without understanding the context of data collection; lack of incentives; limited standardization and norms for data acquisition, formatting, and description; and heterogeneity of data types. Several types of data described as sensitive in the literature are common among neuroscience studies, for example, neural correlates of behavior, emotions, or decision making (71%) and/or predictive data (54%). Although most investigators consider it unlikely or extremely unlikely for their research data to be misused to harm individual research participants (82%), the majority were at least slightly concerned about potential harm to individuals if their research data were misused (65%). Investigators with more easily reidentifiable data, data from vulnerable groups, and neural data were more concerned about the likelihood of misuse and/or magnitude of harm of misuse of their research data.

Discussion: We hope these data help prioritize the development of tools and strategies to overcome the main barriers to data sharing. Furthermore, these data provide input on what may be sensitive data for which additional safeguards should be considered.

Figure 1. Perceived Likelihood and Degree of Potential Harm From Misuse If the Investigators' Individual-Level Deidentified Data Were Publicly Available

Total number of responses: n = 339–357 per item.

Figure 2. Protections for Sharing Individual-Level Research Data Outside of the Investigators' Research Team

Total number of responses: n = 312 -327 per item. Other protections listed by investigators were a hospital firewall (n = 1) and not sharing variables with small cell counts (n = 1).

Figure 3. Factors That Are Assessed to Determine Whether Individual-Level Deidentified Data Will Be Shared After the Study Has Ended

Total number of responses: n = 311-329 per item. *N = 85 -91 respondents indicated that their data would not be shared on archives; this table represents the proportion of the remaining respondents. Under “other” respondents listed: whether the proposed use overlaps with what their own research team is still planning to analyze themselves (concern about scooping their work; n = 2) and whether they would be offered authorship on any publications resulting from their data (n=1). Others require institutional review board and/or institutional approval (n = 3), a data sharing agreement (n = 5), or consideration of what the authorities consider appropriate factors (n = 2). Finally, some said their willingness to share data depends on the technical feasibility of sharing (e.g., given the size of data files, n = 1), whether they have a clear understanding of how the data may be reused (n = 1), and the requestor's commitment to not share the data further or use it for purposes other than the proposed use (n = 1).

Figure 4. Perceived Current Barriers to Increased Sharing of Deidentified Individual-Level Data in Their Field

Total number of responses: n = 309–323 per item. Other barriers to sharing data noted by investigators included unclear guidelines about data sharing (n = 1), the lack of requirements to share data (n = 1), the need for data sharing agreements or material transfer agreements (n = 2), and international restrictions. (n = 1) Furthermore, an inopportune timing of data sharing may be a barrier (n = 2). Finally, investigators raised concerns about unfairness in who contributes and who benefits from data sharing (n = 4) and specifically warned against placing undue burdens on junior scientists, those who are less politically connected, and scientists of color (n = 2). EU = European Union; GDPR = General Data Protection Regulation.