Large Language Model-Based Architecture for Automatic Outcome Data Extraction to Support Meta-Analysis

Meta-analyses play a crucial role in synthesizing findings from diverse clinical studies to evaluate the efficacy of a given treatment. Despite their significance, the conduct of meta-analyses is inherently demanding in terms of time and labor. This process involves a meticulous examination of an extensive body of research articles to extract relevant data. The increasing volume of scholarly publications also presents a significant challenge, leading to the rapid obsolescence of many meta-analyses as they struggle to incorporate emerging evidence. To overcome these challenges, this study leverages the significant capability of large language models to propose a fully automated and time-efficient pipeline tailored for extracting data from open-source research articles, leading to streamlining the meta-analysis process. Additionally, the suggested system has the capacity for continuous updates by automating the integration of new findings, thereby augmenting the temporal relevance of meta-analytic evaluations.