S de Mateo1, E Regidor. 1. National Centre of Epidemiology, Carlos III, Institute of Health, Madrid, Spain.
Abstract
STUDY OBJECTIVE: To compare the results obtained when estimating a standardised rate using the conventional technique of stratified analysis and using Poisson regression, and to evaluate the speed of the two techniques in making the calculation. DESIGN: Cross sectional study. SETTING AND PARTICIPANTS: The trend in motor vehicle accident mortality in males from 1985 to 1992 in Spain was compared using stratified analysis and Poisson regression. In the stratified analysis the calculations were made using a specially designed spreadsheet while in the Poisson regression the statistical program used was EGRET. RESULTS: The stratified analysis took two hours and the Poisson regression 15 minutes to complete. In the stratified analysis a single estimate for each year was obtained, whereas the model of Poisson regression that best fitted the data included an interaction term between age and year. CONCLUSION: Poisson regression can be considered a serious alternative to stratified analysis when the objective is to compare mortality rates standardised by one or two variables.
STUDY OBJECTIVE: To compare the results obtained when estimating a standardised rate using the conventional technique of stratified analysis and using Poisson regression, and to evaluate the speed of the two techniques in making the calculation. DESIGN: Cross sectional study. SETTING AND PARTICIPANTS: The trend in motor vehicle accident mortality in males from 1985 to 1992 in Spain was compared using stratified analysis and Poisson regression. In the stratified analysis the calculations were made using a specially designed spreadsheet while in the Poisson regression the statistical program used was EGRET. RESULTS: The stratified analysis took two hours and the Poisson regression 15 minutes to complete. In the stratified analysis a single estimate for each year was obtained, whereas the model of Poisson regression that best fitted the data included an interaction term between age and year. CONCLUSION: Poisson regression can be considered a serious alternative to stratified analysis when the objective is to compare mortality rates standardised by one or two variables.