Modeling fatal injury rates using Poisson regression: a case study of workers in agriculture, forestry, and fishing.
Poisson regression modeling was applied to the analysis of the fatal injury rates among workers in agriculture, forestry, and fishing in order to address the question of whether there is a calendar year trend in fatal injury rates. Information was obtained for the years 1983 to 1992 from the NIOSH National Traumatic Occupational Fatality database and the United States Bureau of Labor Statistics. Poisson regression analysis was then applied to the data. Over the years 1983 to 1992, the fatal injury rate among agriculture, forestry, and fishing workers remained fairly stable, at about 20 fatal injuries per 100,000 workers. In a simple univariate Poisson regression model, the influence of calendar year on the fatal injury rate was determined. Considering neither age nor gender, this model determined that the fatal injury rate decreased insignificantly by approximately 0.7% each year. The confounding effects of age and gender on the relationship between fatal injury rate and calendar year were determined in a more complex Poisson regression model. The point estimate of the year coefficient changed by 32% when gender and age were included in the analysis. The decline in fatal injury rates with calendar year became more significant when age and gender were controlled. In order to determine whether the predictor variables for calendar year varied according to gender and age, the model incorporated effect modification terms, namely the cross product terms year by gender and year by age. While age did not appear to affect the relationship between calendar year and fatal injury rate, gender did exert a significant impact on the calendar year effect. Although males experienced a slight decrease in fatal injury rate with calendar year, females experienced a marked increase in fatal injury rate with calendar year. The authors conclude that the use of Poisson regression modeling offers a thorough analysis of the effects of variables on fatal injury rates.