Assessing the Application of Artificial Intelligence Techniques for Systematic Reviews and Meta-Analyses

OBJECTIVES: Artificial Intelligence (AI) has become an integral component of numerous professional domains. It has been observed that the use of AI in systematic literature reviews (SLRs) and meta-analyses is increasing. This abstract describes the AI techniques used in the SLR and meta-analyses process phases.

METHODS: Medline (OVID SP®) and Embase (OVID SP®) databases were searched for relevant SLRs published in 2024 with the use of any AI tools (including SciSpace, DistillerSR, EPPI-Reviewer, Rayyan, Covidence, ASReview, Nested Knowledge and Easy SLR). There were no restrictions on language, geography, and medical conditions.

RESULTS: Overall, 271 articles were retrieved, and 226 full-text articles were assessed for the stages of SLR in which AI tools were used. Most used AI tool was Covidence (n=135), followed by Rayyan (n=63), Nested Knowledge (n=10), ASReview (n=10), EPPI-Reviewer (n=4) and DistillerSR (n=2). Covidence is the most frequently used tool with AI techniques across all stages, indicating its versatility and widespread acceptance. Title/Abstract Screening (197) and Full-text Screening (175) are the stages with the highest usage of AI tools, reflecting the critical role AI plays in these processes. AI tools has been moderately used of extraction (58) and De-duplication (48) followed by search (15) indicating its effectiveness in these stages of SLR process. Analysis/Tagging (8) and Quality Appraisal (7) have the lowest AI tool usage, suggesting their underdevelopment in these stages. However, none of the study reported usage of AI tool in meta-analysis.

CONCLUSIONS: AI tools are playing an increasingly significant role in enhancing the efficiency and accuracy of SLRs. Wider adoption and training on existing AI tools, however, could further enhance their usage across all stages of SLRs. Continued assessment of the effectiveness of AI tools in different stages of SLRs will be crucial in refining their applications and improving the overall quality and efficiency of literature reviews.