Calculation and comparison of average standardised mortality ratio in occupational cohort study

The average standardised mortality ratios (SMRs) were calculated, by the methods of median, mean, Poisson and meta-analysis, across 85 occupational cohorts. The difference of average SMR between these methods was reduced with increasing number of cases. The SMRs from the Poisson and fixed/random effects models were close, however, the fixed and random effects models should be more appropriate as the Poisson model ignores variation from the characteristics of different cohorts. There were also wider confidence intervals for the random effects model than for the fixed effects model. Based on such calculations and comparisons of average SMRs, we would suggest that if a total cohort consists of several subcohorts, a summary SMR should be calculated by the fixed effects or random effects models, instead of the Poisson model.