Figure 1Age-adjusted mortality rates for young (30-59) and old (60-79) men and women (world standard population) increases were observed among men. The annual percentage increase in rates for men and women 60-79 years old was, respectively, 12% and 4% in Australia, 7% and 2% in the USA, and 9% and 4% in the Nordic countries. We found a less marked difference between men and women in the CMM mortality increase in Canada (6% for men and 4% for women), Japan (6% and 4%) and the UK (7% and 5%). In France, Italy and Czechoslovakia, CMM mortality rates were low in the 1960s, but afterwards, average annual increases were 24% and 27% in French men and women, 15% and 13% in Italian men and women, and 17% and 33% in Czechoslovakian men and women. Because CMM mortality increased more in older than in younger people, those who died from CMM in 1994 were 10 years older on average than those who died in 1960.
Age-period-cohort modelsThe age-period-cohort models yielded drift terms (measures of the average trend over time) that were all positive and in agreement with the annual percentage increases described above. CMM mortality rose linearly during the period of observation in every population. In every country but Japan, the increase was greater (higher drift) for men than for women. The largest sex differences in drift were observed in Australia, Canada and the USA. Consistent with the analysis of the age-standardized rates, the highest values for drift were found in France, Italy and Czechoslovakia.Trends in mortality rates by birth cohort have been derived from the APC model, and are shown in Figure 2 for three selected countries (Australia, UK and France). These countries were chosen because they represented the three CMM mortality patterns that we observed after APC modelling (see below). Each graph contains three plots that described different scenarios: in the upper plot, all the drift is attributed to a birth cohort effect, in the lower plot all the drift is attributed to a period effect. The rates displayed in Figure 2 simply reflect the trend in rates by birth cohort and are therefore not comparable with actual rates observed in populations. If in each scenario rates are decreasing, we can conclude that there is a real decrease. The trends in CMM mortality rates illustrated by these models describe three patterns:1. increasing rates until generations born around [1930][1931][1932][1933][1934][1935] (for Australia increasing rates ceased earlier), followed by decreasing rates in more recent birth cohorts (in Australia, the Nordic countries and the USA) 2. increasing rates until generations born during World War II, followed by flattening or slightly decreasing rates in more recent birth cohorts (UK, Canada) 3. a steep, almost linear, increase with no major change in this trend (France, Czechoslovakia, and Italy).CMM mortality rates in Japan showed a pattern similar to that described in the third group but the rates of increase were much smaller.
DISCUSSIONWe have succeeded in identifying meaningful patterns in the g...