This paper discusses two issues arising from the use of the concept of 'exposure' in accident studies. However 'exposure' is defined or measured, it is used mainly to calculate accident rate, defined as the ratio between average number of accidents and amount of exposure, for a given time period. In practice, exposure is used mainly for the following purposes: (1) equalising for differences in intensity of use, to make comparisons meaningful; and (2) identifying over-representation, the differences between the characteristic accident rates of entities, to find causal factors and perhaps remedies. The relationship between the number of accidents per unit time, the 'safety performance function', is rarely linear. When it is nonlinear, the same entity will have different accident rates for different intensities of use. Knowledge of safety performance functions is as important as knowledge of exposures; both forms of knowledge are essential, and more attention needs to be given to safety performance functions. With reference to over-representation in accident studies, confusion should be avoided between: (1) the epidemiological paradigm, where causes are sought; and (2) the decision-making paradigm, where relevant facts are sought.
Abstract