Automate Clinical Evidence Synthesis by Linking Trials to Publications with Text Analytics

Abstract

Randomized clinical trials are the core data source for meta- analyses and relative efficacy analyses. As of today, there are missing links between large portions of trials registered in clinical trial registries and their reported outcomes published in scientific journals. This missing citation information makes it difficult to identify all relevant publications for evidence synthesis and decision support. Therefore, we propose a novel natural language processing-based system to establish links between clinical trials and their published outcomes in literature. Different approaches leveraging information retrieval and machine learning with embedding features were systematically developed and evaluated. Results show that shallow machine learning approach with embeddings provide promising results indicating the value it can add to circumvent the limitations of manual search and analyses.