F1000Research (Sep 2022)
Understanding the funding characteristics of research impact: A proof-of-concept study linking REF 2014 impact case studies with Researchfish grant agreements [version 3; peer review: 1 approved, 2 approved with reservations]
Abstract
Background: All parts of the research community have an interest in understanding research impact whether that is around the pathways to impact, processes around impact, methods for measurement, describing impact and so on. This proof of concept study explored the relationship between research funding and research impact using the case studies submitted to the UK Research Excellence Framework (REF) exercise in 2014 as a proxy for impact. Methods: The paper describes an approach to link the REF impact case studies with the underpinning research grants present in the Researchfish dataset, primarily using the publications captured in both datasets. Where possible the methodology utilised unique identifiers such as Digital Object Identifiers and PubMed ID’s, and where this was not possible the funding information within each publication was used. Results: Through this automated approach 21% of the non-redacted case studies could be linked to a specific research grant. Additional qualitative analysis was then done for unlinked REF impact case studies, which involved reading the document to identify additional information to make the linkage. This approach was taken on 100 REF impact case studies selected at random and resulted in only seven having no identifiable research grants funding associated. The linked research grants were analysed to identify characteristics that are more frequently associated with these grants, than non-linked ones. Conclusions: This analysis did point to some interesting observations such as the grant funding linked to REF impact case studies are more likely to be longer, higher financial value, have more publications and be more collaborative (amongst other characteristics). These findings should be used with caution at present and not be over interpreted, this is due to the sample size for this proof of concept study and some potential limitations on the data which were not addressed at this stage.
