11. The case of centenarians
In the earlier chapters little has been shown about ages over 100 years. The numbers have been too small for meaningful period rates or time series for individual countries or even for the aggregates which were formed from them. A still larger grouping is needed.
Reasonably long series of good quality data on centenarian mortality are available in the Odense archive for only thirteen countries of Western Europe. The corresponding data are available also for Japan but the resulting time series is unsteady because in the early decades the numbers were small and possibly inaccurate as would be understandable, birth registration having been introduced only in 1872.
Merged data for the 13 countries are given in Annex Table 8. Smoothed with 5year moving averages the y are illustrated in Figure 6. The results appear to be consistent and credible and allow the conclusion that in the low-mortality zone of Western Europe the mortality of centenarians has been declining throughout the postwar period. There is some evidence of a slowdown from about 1965 to 1975, particularly among men, after which the decline again accelerated.
The death rates for the two sexes have moved in an approximately parallel fashion, the male rate consistently higher than the female, without indicating a tendency to converge as is often claimed to happen. The excess male mortality has fluctuated around 14 - 18 percent without a clear trend.
Another view of centenarian mortality is given in Figure 7 with probabilities of dying at individual ages. It may be pointed out that these. probabilities qx are always lower than central death rates mx, on which Figure 6 was based and that when mortality is high, the difference is quite large. Conceptually, qx is a more faithful measure of mortality when it is calculated, as in the present study, by following a group of persons from one exact age to the next. When the numbers are small, the greater precision may be important.
In Figure 7 the curves are drawn as far as the number of persons exposed to risk is not less than 100. It can be seen that under this rule the curves extend decade by decade one or two years further up the age scale, testifying to a rapid increase in observed cases. The underlying data are given for 1980-90 in Table 13 and for all decades in Annex Table 9. The very unequal numbers involved may be appreciated from the fact that the 1980-90 curve for females is at age 109 based on only 116 observations but at age 100 on nearly 57,000.
It is evident from Figure 7, first of all, that mortality keeps increasing with advancing age and that there is no sign, at least by age 109, of it approaching a plateau, even less a downturn. Allowing for fluctuations due to small numbers and possible inaccuracies in the earliest data, the curves have retained their ascending form essentially unchanged.
Decennial reductions in mortality can also be observed though they have been rather small until the very substantial drop between the 1970s and 1980s which has been consistent by age for both sexes. The current decline in oldest-old mortality extends to ages over 105 years.
Updated by V. Castanova, 1 March 1999