Dead Science: Most Resources Linked in Biomedical Articles Disappear in Eight Years

摘要

Scientific progress critically depends on disseminating analytic pipelines and datasets that make results reproducible and replica-ble. Increasingly, researchers make resources available for wider reuse and embed links to them in their published manuscripts. Previous research has shown that these resources become unavailable over time but the extent and causes of this problem in open access publications has not been explored well. By using 1.9 million articles from PubMed Open Access, we estimate that half of all resources become unavailable after 8 years. We find that the number of times a resource has been used, the international (int) and organization (org) domain suffixes, and the number of affiliations are positively related to resources being available. In contrast, we found that the length of the URL, Indian (in), European Union (eu), and Chinese (cn) domain suffixes, and abstract length are negatively related to resources being available. Our results contribute to our understanding of resource sharing in science and provide some guidance to solve resource decay.