Despite the growth of knowledge and interest into safety and quality in healthcare more generally, the exploration in mental healthcare has been deemed to be in a narrow isolated 'world of its own'. It is possible that relatively little attention is being paid to the processes and interdependencies within the mental health patient safety system. This may result in simplistic static measures of what the system/organisation has, not what it does (or doesn't do). This can limit the potential for learning and affecting change. To investigate systems thinking in mental health patient safety, we conducted a narrative review into the extent of evidence streams supporting systems and complexity thinking approaches. We sourced a total of 89 reports for analysis with six themes identified. These themes included studies evaluating patient safety events that have occurred within mental healthcare, research that has investigated components of the safety system, and studies that have investigated how patient safety incidents are responded to, investigated, and learned from. The review evaluated the use of systems thinking and complexity research in patient safety, and research encapsulating patient and carer involvement. Most research has focused on the analysis of historic approaches to incident investigation and on system-based factors of patient safety, with little attention being paid to systems and complexity thinking approaches. The relationships between components were often ignored in the non-systemic studies sourced, with relationships between components not investigated and unknown. With policymakers recommending changes in patient safety practice through system-based approaches, it is important that its implementation is evaluated robustly with consideration of the multiple levels of the healthcare system. Future research should aim to incorporate systems-thinking approaches to model the safety system, and to improve our understanding of the highly interconnected technical and social entities that dynamically produce emergent behaviour across the system.