Clinical trials are essential for evaluating the safety and efficacy of medical interventions and their success depends critically on the timely recruitment of participants. However, recruiting patients can be challenging and time-consuming, and failure to enroll a sufficient number of participants may increase costs for funding agencies while compromising the reliability and scientific validity of trial outcomes. To address these challenges, we aim to develop automated text-analysis methods to streamline and facilitate the recruitment process.

In this talk, I present our work on natural language inference (NLI) for clinical trial recruitment, demonstrating how it can be used to match patient profiles to relevant clinical trials. I describe the dataset we prepared for this task and present a comparative evaluation of large language models in matching patient medical profiles to clinical trials, using both patient-language and medical-language representations. Our results highlight the potential of leveraging automated NLI approaches to enhance efficiency and accessibility in patient recruitment for clinical research.