Sizing the Problem of Improving Discovery and Access to NIH-Funded Data: A Preliminary Study

Area of Science:

• Biomedical Informatics
• Data Science
• Research Policy

Background:

• Discoverability and access to biomedical datasets are crucial for research reproducibility and advancement.
• A significant portion of data generated by U.S. National Institutes of Health (NIH)-funded research remains

Purpose of the Study:

• To estimate the number and types of

Main Methods:

• Analysis of NIH-funded journal articles published in 2011, indexed in PubMed and archived in PubMed Central (PMC).
• Identification of articles explicitly mentioning data deposition in recognized repositories.
• Random sampling and analysis of remaining articles to quantify and characterize "invisible" datasets.

Main Results:

• Approximately 88% of analyzed articles contained "invisible" datasets, not deposited in known repositories.
• An estimated 200,000 to 235,000 invisible datasets were generated from NIH-funded research in 2011.
• The majority (87%) of invisible datasets comprised newly collected data, with over 50% derived from human or animal subjects.

Conclusions:

• This study provides a preliminary estimate of invisible biomedical dataset generation from NIH-funded research.
• Improving data discoverability requires addressing the definition of a "dataset" and developing better estimation methods.
• Lack of consensus in dataset identification highlights the need for standardized approaches to characterizing biomedical data.