Thesis title: Using LLMs and Argumentation Formalisms to Detect Drug-Drug Interactions for Multimorbid Patients

Supervisor: Atefeh Keshavarzi(a.keshavarzi.zafarghandi@vu.nl)

Managing multimorbidity—where patients have multiple chronic conditions—presents significant challenges in prescribing medications due to the risk of potential drug interactions. Manual checks for these interactions are often time-consuming and error-prone, posing risks to patient safety and adding strain to the healthcare system. To address this issue, we aim to develop a trustworthy AI system that integrates clinical practice guidelines and performs reasoning to provide safe, interaction-free medication recommendations efficiently. The development of this system can be divided into several research components, including: Training large language models (LLMs) using clinical guidelines and Prompting LLMs with specific patient conditions. Developing an argumentation formalism to detect potential drug interactions in multimorbidity cases. Introducing a semantics-based approach to recommend interaction-free drugs for patients with multiple chronic conditions.