Estresse e acidentes de trabalho: Estudo Pró-Saúde

Abstract

This thesis has the general aim of investigating the association between stress and accidents at work with public employees at a public university in Rio de Janeiro, by means of multilevel models. To reach this goal, this study has been distributed into two articles. The first one investigates the association between stress and accidents at work taking into consideration hierarchic components in the data structure by means of multilevel models with employees in the first level grouped in working areas in the second level. The second article investigates the behavior of random and fixed coefficients of the multilevel models with cross classification between working areas and occupational groups in relation to the multilevel models which only take into consideration hierarchic components of the working areas, ignoring the adjustment of occupational groups. The psychosocial stress at work was approached from the relation between high psychological demand and low labor process control. These dimensions were taken by means of the summarized version of the Karasek scale, which also contains information about the social support at work. Isolated dimensions of work stress (demand and control), the ratio between psychological demand and work control (ratio D/C) and the social support at work were measured at individual level and at working areas. Generally speaking, the results highlighted the psychological demand measured at individual level with a major aspect associated to accidents occurrence at work. The social support at work, measured at individual level and in the working area, presented inverted association to the prevalence of accidents at work, that association being more present among women. The results also show that the fixed parameters of the models with and without cross classification were similar and that, in general, the stander errors (SE) were a little bigger in the models with cross classification, although this behavior of the SE hasn t been observed when related to the fixed coefficient of the variable measured at working area. The major distinction between the two approaches was observed in relation to random coefficients related to the working areas, which were highly altered after adjusting the effect of the occupation by means of models with cross classification. This thesis reinforces the importance of psychosocial characteristics in the occurrence of accidents at work and contributes to the knowledge of these relations from analytical approaches which refine the gathering of the dependence structure of individuals in their working place. It s suggested that other studies with similar methodologies are made, that would allow us to go deep in the knowledge of stress and accidents at work.