A framework for improving the reproducibility of data extraction for meta-analysis

Extracting data from studies is the norm in meta-analyses, enabling researchers to generate effect sizes when raw data are otherwise not available. While there has been a general push for increased reproducibility throughout the many facets of meta-analysis, the transparency and reproducibility of the data extraction phase are still lagging be-hind. This particular meta-analytic facet is critical because it facilitates error-checking and enables users to update older meta-analyses. Unfortunately, there is little guidance of how to make the process of data extraction more transparent and shareable, in part this is as a result of relatively few data extraction tools currently offering such functionality. Here, we suggest a simple framework that aims to help increase the reproducibility of data extraction for meta-analysis. We also provide suggestions of software that can further help users adopt open data policies. More specifically, we overview two GUI style software in the R environment, shinyDigitise and juicr, that both facilitate reproducible workflows while reducing the need for coding skills in R. Adopting the guiding principles listed here and using appropriate software will provide a more streamlined, transparent, and shareable form of data extraction for meta-analyses.