OBJECTIVE: To evaluate, by analysis of breast cancer mortality data from all the 26 Swedish counties for the years 1971 to 1990, whether the effect of the introduction of mammography screening in Sweden can be assessed by observation from existing mortality data. METHODS: A Poisson regression model was used to study whether a decrease in breast cancer mortality among women aged 50-74 years was associated with the extent of mammography screening in different counties and periods. RESULTS: In regions where mammography screening had been introduced, breast cancer mortality tended to be decreased, on average, compared with regions without screening. If a 10 year time lag between the start of screening and its full effect on mortality is assumed then the estimated reduction in breast cancer mortality associated with introduction of screening was 19% with a 95% confidence interval ranging from -3% to 37%. CONCLUSIONS: The results suggest that the effect of mammography screening may be studied using existing routine mortality data and appropriate statistical modelling. This way of assessing the outcome of the screening is valuable when continuously monitoring a screening programme that has become a public health routine.