Objective: Smoking is a potential confounder in studies of workplace exposures and smoking-related disease, but little data exist to quantitatively adjust for smoking in statistical models.
Methods: We estimated smoking prevalence trends between 1950 and 1999 for 12,299 female and 43,307 male hourly and salaried petrochemical workers using company physical examination data.
Results: Nearly half of hourly male and female employees smoked during the study period, compared with 38% of salaried males and 29% of females. Smoking prevalence in the 1950s reached 80% and 66% among female and male hourly workers, respectively, significantly higher than the US general population.
Conclusions: As hourly workers typically comprise higher exposure groups and expected case counts are typically generated from the US general population, biased risk estimates may result from standardized mortality ratio analyses if smoking rate differences are not accounted for.