Objective: Web resources can contain high-quality data relevant to peer-reviewed medical publications. However, their online location may change or disappear with time. As medical publication professionals, we are concerned with the ephemeral nature of web resources and the associated qualitative impact on publication integrity of referencing such resources.
Methods: Time-dependence of the phenomenon was probed using a dataset of reference lists from open and free access articles published in 2018. Each reference list was manually screened to tally the number of web pages no longer accessible. The analysis was repeated yearly from 2021 to 2023. Additionally, a set of proofs was analyzed to investigate the proportion of cited web references already inaccessible at publication. A third dataset, consisting of modeling articles published in 2018-2023, was used to quantify -at a single timepoint- the share of web resources cited as model inputs that were inaccessible.
Results: The proportion of inaccessible web resources increased on average from 27.2% to 41.9% three to five years post publication (n = 992 articles), respectively. One out of four analyzed proofs cited at least one inaccessible web resource (n = 50). Five years after publication, 26.1% of web resources used as model inputs were no longer accessible, with one in three modeling articles being impacted (n = 61).
Conclusion: The issue of inaccessible web resources cited in peer-reviewed medical publications is of great concern due to the fast pace and potential impact on research reproducibility. These findings call for the definition of best practices involving all stakeholders and the deployment of robust archiving solutions.