Presentation (CTI)

OBJECTIVES: Systematic literature reviews (SLRs) are essential for evidence-based medicine (EBM). However, growing research volumes make them increasingly time-consuming. Automation tools and the use of artificial intelligence (AI) are emerging but, without consensus on how to use them, their use risks introducing bias and reducing reproducibility. To the authors’ knowledge, no research has been performed to draft best practices for automation and AI in SLRs.
METHODS: A review of guidelines on the use of AI, specifically in EBM, published by health technology assessment (HTA) agencies and SLR methods groups was conducted. This was supplemented with a targeted review of articles from MEDLINE. Publicly available information about SLR automation tools was also reviewed, to better understand industry AI capabilities.
RESULTS: AI policies and guidelines from major HTA agencies and SLR methods groups were identified. These recommend AI for search, screening, and data extraction, but only to augment, not replace human review. However, these do not specify how exactly AI tools need to be utilized. No published article from MEDLINE was identified with best practices for using AI in SLR. Available good practice guidelines were cross-checked with AI capabilities that were detailed in publicly available information for approximately 25 SLR tools. Based on these, a checklist of AI involvement in SLRs was developed. The checklist acknowledges challenges existing in automating search, screening and data extraction. It offers insights on AI use in search term identification, screening, data extraction, evidence summary development and data visualization.
CONCLUSIONS: The use of AI-based tools has started being accepted by research groups and HTA bodies. Although this may change in the future, currently, AI tools should be used to support rather than replace human intervention. Considering and detailing AI use in a proposed checklist will help users to select the tool that best suits their needs.