To provide another crude mortality estimate, we assembled the number of deaths by birth cohort. If we assume that the last member of a birth cohort died at a certain age – say, 115 – and that international migration above age 100 is negligible, we can reconstruct the centenarian cohort population between 100 and 115 years old by adding the number of deaths backward from 115 to 100. This is the same method we used for creating data for Fig. 10.1b above, but here we only create data by cohort. We grouped five cohorts to create data for mortality estimates by birth cohorts, starting with the 1874 cohort: i.e., 1874–1878, 1879–1883, 1884–1888, 1889–1893, and 1894–1898. In addition, we grouped two cohorts, 1899 and 1900, for the analysis.
Note that among the total of 260,220 deaths recorded at age 100 or older from 1951 to 2015 used in this study, only nine cases were reported to have died after the age of 115: five at the age of 116, one at the age of 117, two at the age of 118, and one at the age of 120. This last case, of Shigechiyo Izumi, is known to be misreported. As both cases of deaths at age 118 are old, (reported in 1951 and 1964), it is likely that they were also misreported, and are therefore excluded from our analysis. Our examination of the data by cohort is thus restricted to the cohorts who reached the age of 100 starting in 1974. The case of Misao Okawa, who reportedly died at the age of 117, may be true (see Chap. 21). This case should be added to our study as a survivor at her 117th birthday in order to eliminate any age censorship for the 1898 birth cohort. Among the five cases of individuals who reportedly died at age 116, two are old. These two cases, who were reported dead in 1951 and 1970, may be incorrect, but do not interfere with our cohort study. Another two of these cases are of individuals whose deaths at the age of 116 were reported more recently: one of a woman, Kamato Hongo, who died in 2003; and one of a man, Jiroemon Kimura, who died in 2013 (Gondo et al. 2017). Although Poulain (2010) investigated Hongo’s case and questioned her reported age, we included her case here because there is currently no consensus about her age. The last case, of an individual named Tane Ikai who was reported dead at the age of 116 in 1995, has been verified (Inagaki et al. 1997). These three cases should be included in our study as survivors at their 116th birthdays to eliminate any age censorship in their respective birth cohorts: 1879 for Tane Ikai, 1887 for Kamato Hongo, and 1897 for Jiroemon Kimura. There may also be some missing cases of supercentenarians who belong to the studied cohorts and survived to 2015. We therefore looked carefully at the public lists of Japanese supercentenarians who died in 2016 or 2017 at the age of 116 or older, as well as at the list of Japanese supercentenarians who were known to be alive on August 5, 2017.Footnote 2 Of these living supercentenarians, two are older than 115 years old: Chiyo Miyako (116 years old on August, 5, 2017; see Chap. 21) and Nabi Tajima (117 years old on August, 5, 2017). Chiyo Miyako was born in 1901, and therefore is not part of the present cohort study; but Nabi Tajima was born in 1900, and should be added to our study as having survived to her 117th birthday in August 2017.
By adding these five cases to their respective cohorts, we reconstructed cohort populations from age 100 to age 117, as shown in Table 10.3. Of these five added cases, four were female and one was male. To the best of our knowledge, our information on the Japanese centenarians belonging to the 1874–1900 birth cohorts is complete. Only two of these individuals reached the age of 117, one of whom was still alive on August 5, 2015. We used this information to study empirically the mortality trajectories in Japan above the age of 100.
In total, this part of our study comprises 53,533 centenarians – 42,547 females and 10,986 males – belonging to the cohorts born in 1874–1900. For each group of birth cohorts, the number of survivors was followed from the age of 100 to extinction at age 117, without any age censorship, except for two females who survived to age 117. The survival curves are plotted on Fig. 10.3, with a radix of 10,000 survivors at age 100 for each group of cohorts. We chose 10,000 as a radix for life table construction and for computing the standardized number of survivors (lx) and the probability of dying (qx) for the study. Figure 10.3 shows shifts in the survival curve to the right (higher ages) from one group of cohorts to the other. Obviously, each successive group of centenarians, comprising five birth cohorts, is surviving to higher ages than the previous group. This shift to the right is observed for both the male and the female cohorts.
Table 10.4 shows the computed life table probabilities of dying (qx) by birth cohort from 1874 to 1900 and by sex. When looking at the table, it is very important to note that the actual size of the cohorts is increasing quickly for later cohorts, and that the size of the female cohorts is much larger than the size of the male cohorts. The table also shows that qx is generally lower for later cohorts for both females and males. qx steadily increases with age, and then starts fluctuating.
Figure 10.4 provides a graphical presentation of qx, as shown in Table 10.4. The general trend observed from this figure is that the smaller the studied group of cohorts (earlier group of cohorts and/or male group of cohorts), the sooner fluctuations appear to prevent these cohorts from following the mortality trajectories associated with higher ages.
When we look at individual mortality trajectories by birth cohort, some interesting features appear. The 1874–1878 male cohort displays fluctuations from age 103 onward; while the 1894–1898 female cohort and the 1899–1900 cohort seem to show a steady increase in mortality until age 115 or age 113, respectively. In between, we can see that the age at which the cohorts become extinct tends to shift over time to higher ages, and that the fluctuations tend to decrease with the increase in the number of centenarians in the studied cohorts. More importantly, we can see that the probabilities of dying tend to plateau with age for the cohorts in the middle (1879–1883 female cohorts, 1884–1888 male and female cohorts, and 1889–1893 male and female cohorts). Fluctuations can be observed among the earlier groups of cohorts, especially the 1874–1878 group of cohorts. By contrast, among the most recent 1894–1898 group of cohorts, a regular increase in mortality can be seen up to age 111 for the male cohorts and up to age 115 for the female cohorts. The two most recent extinct cohorts, 1899–1900, display a monotonic increase in mortality until age 107 for males and age 113 for females. This observation suggests that a plateau of mortality occurred among the cohorts born between 1879 and 1893. Among the preceding cohorts, the oldest-old are too rare to allow us to observe anything other than fluctuations in mortality above age 100. Among the cohorts born after 1894, when the numbers of centenarians start to be consequential, the mortality trajectories seem to keep increasing with age, even if the rate of increase is not large.
As a kind of sensitivity analysis, we grouped the cohorts born after 1893 differently. The two female cohorts 1899–1900 seem to follow a lower mortality trajectory than the five previous birth cohorts born in 1894–1898, but they represent only two cohorts of centenarians (n = 9859) versus five cohorts (n = 16,035). The two male cohorts comprise only 2261 centenarians, and suffer from large fluctuations. Therefore, it is tempting to add the two cohorts of 1899–1900 to the previous 1894–1898 cohorts, especially because Fig. 10.4 suggests that the survival patterns of these two groups of cohorts are quite similar. We took this opportunity to perform some sensitivity analyses, proposing alternative groupings (see Fig. 10.5).
Whatever the grouping used starting with the cohort born between 1894 and 1900, mortality seems to increase, at least up to age 107 among males and up to age 111 among females. Then, for the following 2 or 3 years, to age 110 for males and age 113 for females, the trend seems to keep increasing, but with some noise. After these ages, the series only show fluctuations due to the small numbers of Japanese males and females reaching the ages of 111 and 114, respectively.