Accidents are now widely acknowledged to be a systems phenomenon. As part of a proactive approach to safety management, organisations use risk assessment methods to identify the hazards and associated risks that may lead to accidents. Although there is an extensive body of literature on the need for a systems thinking approach in accident analysis, little has been said regarding the theoretical underpinnings of risk assessment methods. The aim of this paper was to systematically review the risk assessment methods presented in the literature and evaluate the extent to which they are underpinned by a systems thinking approach. A total of 342 methods spanning a range of safety-critical domains were evaluated using Rasmussen’s tenets of accident causation. A key finding is that the majority of existing risk assessment methods are not consistent with Rasmussen’s model of accident causation (arguably the most popular model in safety science circles). Instead, the majority of risk assessment methods focus on risks at the so called sharp-end and largely view accidents as emerging from a linear, or chain-of-events process. This overlooks emergent risks at other levels of the system, including supervisory, managerial, regulatory and government levels. The findings therefore suggest that the majority of existing risk assessment methods may be inadequate for identifying hazards and analysing risks within complex sociotechnical systems. The implications for risk assessment practice are discussed.