The very definition of the word "accident" implies that it is unplanned, unexpected and occurs without forewarning. While that may be so, is the occurrence of an accident nonetheless predictable? Any reader of this paper who has had even a few months of experience as the safety manager in a manufacturing or construction organization, and who has had the opportunity to investigate an accident, has probably heard another worker say words to the effect of "I just knew that, sooner or later, this would happen." While that comment was a generalization and may have been predicated merely upon premonition, it is an important clue that, in fact, accidents are predictable. Those who go to the efforts necessary to make such predictions can take appropriate actions to insure that the prediction will never come true. While this process is neither simple nor quick, its proper execution will prevent human suffering and economic loss that will make the effort a worthwhile expenditure of the necessary time and resources. The purpose of this presentation is to set forth a protocol by which such predictions can be made and demonstrate the use of that process in a few case studies of accidents that have actually occurred.
There are many worldly events for which valid predictions are commonly made through the accumulation of historical data, together with application of the science of the field and the mathematics of statistics and probability. The most common of these are weather, business and finance, sports, gaming, and politics. As will be documented in this presentation, the same principles that enable predictions to be made in those fields can be applied to safety and health. However, there are two primary and important differences between making predictions in those other venues, and doing so with respect to health and safety. First, other predictions can only be made after other related, unpredictable events have already occurred. Second, it is acceptable to sit and observe whether the predictions come true. In the field of health and safety, the object is to make predictions that will prevent those events from ever occurring and, in the assumption that they do not, we will be most satisfied if we never learn whether the events would have occurred.
There are two well accepted ways of performing technical analyses. The first, commonly known as the inductive scientific method, is to start with a hypotheses and then test it with known facts. Unfortunately, this will not work for accident prediction because the hypotheses, being the prediction, must be the end, not the start, of the process. We must therefore rely upon the deductive scientific method, in which packages of known facts (axioms) are aligned to arrive at the answers (theorems). We do not have the luxury of taking years to philosophize as we choose and align our fact packages, as did the ancients who used deductive reasoning to develop the field of mathematics.
