Automating the Systematic Review Process: A Bibliometric Analysis Public Deposited

Downloadable Content

Download PDF
Last Modified
  • May 9, 2019
Creator
  • Dodd, Adam
    • Affiliation: University of North Carolina at Chapel Hill. Health Sciences Library
  • Moreton, Elizabeth
    • ORCID: orcid.org/0000-0003-1769-7134
    • Affiliation: University of North Carolina at Chapel Hill. Health Sciences Library
Abstract
  • Objectives : To perform a bibliometric analysis on the body of literature discussing the use of automation tools to facilitate systematic reviews.Methods : A comprehensive search was performed to identify articles describing the automation of systematic reviews using artificial intelligence (e.g. machine learning, text mining, algorithms, etc.). Searches were conducted in biomedical, library & information science, and engineering databases. Results were aggregated and de-duplicated. The number of results indicated that machine learning was an appropriate technique for reviewing such a large body of literature. A random sample of records was screened to identify relevant papers and create a training data set. After the machine learning process was employed to complete screening on the remaining records, final results were uploaded into a visualization & network mapping tool.Results : Our bibliometric analysis found considerable collaboration around a core of prolific authors based in countries where evidence-based medicine has been widely adopted. Over time, representation in publications has moved from computer science journals and conference proceedings to domain-specific application with many newer enhancements focused on health sciences. Additionally, publications have moved from theory to problem-solving in specific parts of the systematic review process. Results will be delivered as a poster at the MLA 2019 Annual Meeting with visualizations including: size and scope of literature, key journals and conferences, collaboration networks, and author demographics including affiliation and role.Conclusions : The field of systematic review automation has grown substantially in recent years, both in the addition of new authors and the expansion of collaboration networks. While adoption has increased in systematic review-focused publications and groups, more work must be done to gain acceptance in biomedical publications, including validations of algorithms and tools, guidelines for use in systematic reviews, and education of researchers. As key partners in systematic review performance and educators in systematic review methods, librarians are poised to play an important role in the adoption and acceptance of this technology.
Date of publication
Keyword
Resource type
Conference name
  • Medical Library Association Annual Meeting
Language
Parents:

This work has no parents.

Items