Many journals now require authors to make their data freely available after publication. These include the British Medical Journal (BMJ), which first published an editorial encouraging the sharing of data back in 2009, and which from 1 July 2015, has required authors of all published clinical trials ‘to make the relevant anonymised patient-level data available on reasonable request’. However, a new study in the BMJ suggests that while rates of ‘data promising’ have increased in recent years, rates of data sharing remain low.
Rowhani-Farid and Barnett randomly selected 160 research articles published in the BMJ between 2009 and 2015. Three of the articles included their full dataset within the paper itself. A further 50 indicated that the data were available, 12 in public databases and 38 upon request. However, Rowhani-Farid and Barnett found that only one of the 12 deposited datasets could be accessed in practice – in the repository Dryad – and only six of the 38 authors emailed their data when asked to do so. This equated to a data sharing rate of 10/160 (6.3%). For clinical trials, the rate of data sharing was higher but still low, at only 24%.
Rowhani-Farid and Barnett acknowledge that some of their emails may have ended up in spam folders, and that the wording of the BMJ policy statement leaves room for interpretation, particularly regarding what constitutes a ‘reasonable request’. Nevertheless, given the much higher rate of ‘data promising’ than data sharing, they suggest that a re-evaluation of existing policies may be necessary, along with improved incentives for the sharing of data.