Bridging the Gap Between Theory and Reality in Systematic Literature Reviews Using Artificial Intelligence

OBJECTIVES: In an ideal systematic literature review (SLR), the process begins with finalising a protocol, followed by title-abstract and full-text screenings, with a third reviewer resolving conflicts. However, practical implementation often diverges, leading to protocol uncertainties, less alignment among reviewers, high or low inclusion rates, and potential rescreening needs. This study aims to address these challenges using artificial intelligence (AI) and cluster analysis.

METHODS: We employed an AI-driven approach using the EasySLR platform. Natural language processing and k-means clustering algorithms grouped citations into 20 thematic clusters. Two independent reviewers used a representative subset from these clusters for pilot screening. Conflicts and alignments across clusters were analysed to rapidly refine the protocol and improve reviewers' understanding of target areas.

RESULTS: Key outcomes include reduced inter-reviewer conflicts, improved resource planning through AI-assisted inclusion rate estimation by clusters, and enhanced protocol refinement based on early insights from clustered samples.

CONCLUSIONS: These improvements contribute to more efficient, consistent, and reliable SLRs, bridging the gap between theoretical best practices and practical implementation. Based on our research, PubMed indexed 44,987 new SLRs in 2023 alone, and this number has increased more than five-fold over the last decade. As the volume of published research continues to grow, such AI-enhanced methods may become essential tools in evidence synthesis and informed decision-making in healthcare and beyond. Future research should focus on validating these findings across diverse fields and quantifying the impact on SLR quality and efficiency.