Sharing is hard. Scientists who promise to share their research data do not do so or are unreachable. Three Croatian scientists from Split noticed this when they wrote to almost 1,800 colleagues and asked them to send them research data. It report of that exercise, the researchers published 30 May in the Journal of Clinical Epidemiology†
In theory, disclosing or sharing research data is good practice among scientists. With access to data, other scientists can reproduce results. And it is a safeguard against scientific misconduct. If data is public, third parties can detect fraud or embellishment more quickly.
Publishers of scientific journals would compel the sharing or disclosure of data. But they don’t. At most, they encourage authors to create a so-called data availability statement (DAS) in their article. In it, authors state whether underlying data is retrievable, even if that is not the case. But also the inclusion of a DAS is often not an obligation.
Reasonable requests
Publisher BioMed Central (BMC) does require authors to draw up a DAS. The Croats therefore turned their attention to researchers who publish in BMC journals to see what those DASs are worth. The trio analyzed all articles that appeared in BMC magazines in January 2019. In the most common DAS (1,437 peated), the authors stated that “data is available upon reasonable request.”
The Croats sent 1,792 emails requesting data, stating at once that the purpose was to explore willingness to share. Only 254 answers were returned (14 percent). Less than half of them (122) eventually shared the data. Even when there was contact, the sharing of research data was more often than not: conversations bled to death, some wanted a nondisclosure agreement (but then didn’t respond when sent), two authors wanted compensation, one demanded to be a co-author. are included in follow-up research and one more admitted to having lost the data.
The DAS therefore falls short as an instrument for making data accessible, the researchers conclude. Only a firm demand from the publisher to make data public can make the difference, the Croats think.
That seems like a good idea anyway. Editor-in-chief Tsuyoshi Miyakawa of trade magazine Molecular Brain shared last year a striking insight: when he asked 41 authors of manuscripts for additional data, 20 promptly withdrew their contributions and 19 also did not provide enough data at a second glance, “suggesting the possibility that the raw data did not exist from the start.”
The data of the Croats themselves are not public, by the way: “We did not publish our raw data because it could be interpreted as pillorying authors for not wanting to share their data.”