Objective There has been growing interest in applying complexity theory to health care systems, both in policy and academic research discourses. However, its application often lacks rigour - authors discuss the properties of complex systems, state that they apply to health care and draw conclusions anchored around the idea of ‘whole system change’. This paper explores the use of whole systems change in a programme to improve the delivery of unscheduled health care in Scotland. Methods Qualitative case-studies of five health boards in Scotland reflecting different demographics, initial performance data and progress towards meeting programme targets. Results The programme's collaborative approach was successful in moving to a culture of mutual understanding and greater awareness of the interdependencies between different functions within the hospitals. There was whole system working at the acute hospital level, leading to improved patient flows. But despite recognizing the need for whole system change overall, it proved hard to address relationships with stakeholders influencing wider out-of-hospital patient flows. This was exacerbated by the structure of the programme, which was designed much more around acute patient flows. Conclusions The programme worked well to improve performance by focusing on interdependencies within a large part of the acute care subsystem but did not have the same impact at the overall health care system level. This has important implications for the design of policy and associated programmes which seek to effect whole system reform, or at least are realistic about the magnitude of change they can achieve.