Date created: | Last Updated:

Category: Project

Description: Abstract Background and rationale: Despite access to a broad range of electronic databases some works may remain unavailable as full-text for systematic review authors. Searching Google Scholar for multiple different words, sentences or paragraphs from one scientific work can, at least in theory, be used to reconstruct data from previously unavailable studies, excerpt by excerpt, until enough information is found to assess eligibility. Objective: To assess the feasibility of using Google Scholar to reconstruct data from studies previously categorized as “unavailable” Study design: Case study Methods: Attempts were made to reconstruct as much information found within 6 previously “unavailable” articles through iterative Google Scholar searches, combining “intitle:title of the article” and purposively selected keywords (full sentences, keywords specific to different study sections, keywords found by previous searches, etc.). All unique new text/sentences/data found were copied in a Word document, categorized by the study page on which they were found (or suspected to be found when this was not clear) and then collated. Results: Three (n=3) articles could not be reconstructed through this method for Google Scholar searches only helped identify data about their abstracts or no new data at all. 1290 unique words spread over 45 pages could be found for one article, including 253 consecutive words describing the study procedures in enough detail to assess the eligibility of this article (a qualitative study which was therefore ineligible and excluded). The eligibility of the two remaining articles could not be determined based on the data extracted and collated. Among other things the following data could be found before reconstruction attempts were interrupted: title, abstract, people acknowledged, list of keywords, name of some figures, name of some appendixes, some authors/articles cited, total number of pages, study procedures, study design, full sentences from the introduction and methods sections, statistical tests used to analyze the results, research questions, limitations, chapter titles, number of study arms, name of appendixes, etc. Discussion: Reconstruction attempts were interrupted when I was blocked by Google and further searches may have helped reconstruct more complete data which may have made eligibility assessments possible. The most effective search strategies are yet unknown. Potential reasons why three articles could not be reconstructed include: Google either does not have a full-text copy of these works or does not have the authority or rights to crawl their full-texts. In some cases email addresses of study authors (not found otherwise) can also be found with this strategy; this may make reconstruction attempts unnecessary. Limitations: This strategy will not work to reconstruct articles Google Scholar does not have access to; some works will therefore remain unreachable despite use of this method. Fully reconstructed articles may still be reported in insufficient details to assess eligibility. Incompletely reconstructed articles may end up misleading systematic review authors. Repeated attempts to reconstruct studies can get users banned from using Google Scholar for some time. This reconstruction strategy can be time-consuming, it cannot be used to extract data found within images and the feasibility and reliability of extracting study results (particularly if found within tables) is unknown. Excessive or inadequate uses of this strategy may lead Google to implement barriers preventing its use in future projects. Funding: No funding was received for this work. Registration: Not done. Data and materials: Extracts of data collected are included within this manuscript. Further data is not shared due to copyright concerns. Keywords: Google Scholar, systematic reviews, literature searching, data availability