The Availability and Utility of Open Access Data from Studies on Food Assistance for Nutrition

The desk study ran from June through December 2020. Of 179 studies published and archived in REFINE over the last 5 years, only 15 (8.4% of this sample) had shared underlying datasets in an open access format. Where datasets were published, many were minimal, and lacked accompanying documentation needed for interpretation. An in-depth assessment of the publications archived with open access datasets showed that there is significant heterogeneity among study designs, which hinder prospects for conducting meta-analyses with pooled data. Studies collected a wide range of program exposure variables, and there is much heterogeneity among the coding structures of key variables. This review concluded that access to 'open access' data is very limited at present, and therefore, the potential for pooled secondary analyses of secondary datasets is very limited. This indicates that robust guidance and infrastructure for data sharing is needed in order to make open access truly meaningful. Key recommendations identified for funding organizations, including USAID, focus on approaches to improve the FAIRness of open data. Funding organizations should provide guidance on prospective planning for data sharing, including requiring detailed data management plans (DMPs) specifying data sharing procedures. DMPs should include data collection methods, roles of research staff, and plans for long-term accessibility of data. Funding organizations should also provide guidance and example DMPs, a mechanism for evaluation of DMPs, and training for evaluators. In addition, funding organizations should include in participant consent forms a statement of intent to openly share de-identified data. Curation of data for dissemination, repository deposition, and inclusion of accompanying documentation is also crucial. Funding organizations should specify a timeframe to archive underlying data following article publication, ensure datasets are comprehensive and de-identified, and archive documentation needed for interpretation of dataset contents. This documentation should include complete codebooks/data dictionaries, data collection instruments, and syntax codes used to run analyses. Funding organizations should also clarify preferred locations for archiving data and monitor data sharing compliance. The study highlights the need for robust guidance and infrastructure for data sharing to make open access truly meaningful. By improving the FAIRness of open data, funding organizations can enhance transparency of study results and facilitate reuse of data to inform further scientific research.