A bivariate zero-inflated Poisson regression model to analyze occupational injuries.

The aim of many occupational safety interventions is to reduce the incidence of injury. However, when measuring intervention effectiveness within a period, population-based accident count data typically contain a large proportion of zero observations (no injury). This situation is compounded where injuries are categorized in a binary manner according to an outcome of interest. The distribution thus comprises a point mass at zero mixed with a non-degenerate parametric component, such as the bivariate Poisson. In this paper, a bivariate zero-inflated Poisson (BZIP) regression model is proposed to evaluate a participatory ergonomics team intervention conducted within the cleaning services department of a public teaching hospital. The findings highlight that the BZIP distribution provided a satisfactory fit to the data, and that the intervention was associated with a significant reduction in overall injury incidence and the mean number of musculoskeletal (MLTI) injuries, while the decline in injuries of a non-musculoskeletal (NMLTI) nature was marginal. In general, the method can be applied to assess the effectiveness of intervention trials on other populations at high risk of occupational injury.