This thesis investigates the use of GraphRAG for complex diagnostic reasoning. The aim is to identify the unique requirements of diagnostic tasks compared to general RAG applications, develop a prototype GraphRAG system for a selected diagnostic domain and derive transferable design and evaluation guidelines that can support future applications in diverse contexts. By grounding the study in a real-world case study (e.g. IT service desk application), the thesis seeks to bridge theoretical advances in GraphRAG with practical diagnostic challenges.