Human and management factors in probabilistic risk analysis: the SAM approach and observations from recent applications

Most severe industrial accidents have been shown to involve one or more human errors and these are generally rooted in management problems. The objective of this paper is to draw some conclusions from the experience that we have acquired from three different studies of this phenomenon: (1) the Piper Alpha accident including problems of operations management and fire risks on-board offshore platforms, (2) the management of the heat shield of the NASA space shuttle orbiter, and (3) the roots of patient risks in anaesthesia. This paper describes and illustrates the SAM approach (System-Action-Management) that was developed and used in these studies to link the probabilities of system failures to human and management factors. This SAM model includes: first, a probabilistic risk analysis of the physical system, second, an analysis of the decisions and actions that affect the probabilities of its basic events, and third, a study of the management factors that influence those decisions and actions. In the three initial studies, the analytical links (conditional probabilities) among these three submodels were coarsely quantified based on statistical data whenever available, or most often, on expert opinions. This paper describes some observations that were made across these three studies, for example, the importance of the informal reward system, the difficulties in the communication of uncertainties, the problems of managing resource constraints, and the safety implications of the short cuts that they often imply.