Zero-inflated Poisson modeling to evaluate occupational safety interventions

Abstract

Much of the data collected on occupational safety involves accident or injury counts. Traditional approaches to analyzing this type of population-based count data, including the Poisson model and the Negative Binomial model, do not take into consideration the relatively few events and hence, many observed zeros. In this paper, we propose a Zero-Inflated Poisson (ZIP) model to evaluate the effectiveness of a consultative manual handling workplace risk assessment team (WRATS) in reducing the risk of occupational injury among cleaners within a 600-bed hospital. We modify the ZIP model (which adjusts for the excess zeros) to take account of the extent of exposure in terms of hours worked during the study period. The findings highlight that the modified ZIP approach provided a satisfactory fit to the data, and that the manual handling WRATS intervention was associated with a reduction in the proportion of cleaners injured at work. In the presence of extra zeros, the ZIP model can be an alternative to the Poisson and Negative Binomial models, when appropriate according to statistical criteria.