[Factors of random selection in epidemiological surveys]

In epidemiological surveys the association of variable under study, for example, morbidity and conditions of work, may be distorted by phenomena of selection. The most frequent effect of selection is "the effect of healthy worker" which consists in lower morbidity and mortality of workers as compared to the population at large. This effect can mask the pathogenic action of environmental conditions. Under other circumstances multiple effects of selection, on the contrary, may additionally enhance the association between morbidity and the environmental conditions. The elimination of selection errors is possible, if the analysis of data takes into account the broad range of variables capable of affecting the dependence. The issue of causative nature of dependence should be solved proceeding from the analysis of the aggregate data on the problem.