Complex engineered systems, such as nuclear reactors and chemical plants, have the potential for catastrophic failure with disastrous consequences. In recent years, human and management factors have been recognized as frequent root causes of major failures in such systems. However, classical probabilistic risk analysis (PRA) techniques do not account for the underlying causes of these errors because they focus on the physical system and do not explicitly address the link between components' performance and organizational factors. This paper describes a general approach for addressing the human and management causes of system failure, called the SAM (System-Action-Management) framework. Beginning with a quantitative risk model of the physical system, SAM expands the scope of analysis to incorporate first the decisions and actions of individuals that affect the physical system. SAM then links management factors (incentives, training, policies and procedures, selection criteria, etc.) to those decisions and actions. The focus of this paper is on four quantitative models of action that describe this last relationship. These models address the formation of intentions for action and their execution as a function of the organizational environment. Intention formation is described by three alternative models: a rational model, a bounded rationality model, and a rule-based model. The execution of intentions is then modeled separately. These four models are designed to assess the probabilities of individual actions from the perspective of management, thus reflecting the uncertainties inherent to human behavior. The SAM framework is illustrated for a hypothetical case of hazardous materials transportation. This framework can be used as a tool to increase the safety and reliability of complex technical systems by modifying the organization, rather than, or in addition to, re-designing the physical system.

No abstract

The risk of death or brain damage to anesthesia patients is relatively low, particularly for healthy patients in modern hospitals. When an accident does occur, its cause is usually an error made by the anesthesiologist, either in triggering the accident sequence, or failing to take timely corrective measures. This paper presents a pilot study which explores the feasibility of extending probabilistic risk analysis (PRA) of anesthesia accidents to assess the effects of human and management components on the patient risk. We develop first a classic PRA model for the patient risk per operation. We then link the probabilities of the different accident types to their root causes using a probabilistic analysis of the performance shaping factors. These factors are described here as the "state of the anesthesiologist" characterized both in terms of alertness and competence. We then analyze the effects of different management factors that affect the state of the anesthesiologist and we compute the risk reduction benefits of several risk management policies. Our data sources include the published version of the Australian Incident Monitoring Study as well as expert opinions. We conclude that patient risk could be reduced substantially by closer supervision of residents, the use of anesthesia simulators both in training and for periodic recertification, and regular medical examinations for all anesthesiologists.

No abstract