There is a growing need worldwide to increase the quality and productivity of healthcare services delivery. To this end, analysing and reducing unwarranted variations in healthcare has attracted much attention in recent times. However, current modelling and simulation approaches to reduce unwarranted variations suffer from numerous limitations. Consequently, service improvement efforts have often failed to deliver expected results. This paper discusses the key issues associated with reducing unwarranted variations in hospital service delivery systems, and proposes a research framework that aims at overcoming these issues. In doing so, it highlights the need for: accurately and efficiently modelling complex service delivery systems; developing systematic knowledge acquisition approaches; and developing scalable simulation models to analyse unwarranted variations on and from care pathways.