Methods Discussion
3 Results
In the Austrian data set, death after age 50 occurred at an average age of 77.7. The mean life span of people born in specific weeks of the year deviates from this average, and the deviations vary periodically in a 12-month cycle (Figure. 1). There was no significant difference in this pattern between men and women. Average age at death is lowest for those born around week 20 and highest for those born around week 46. I found a similar pattern when I considered deaths after age 70.

Dividing the year into four quarters (weeks 1-13, 14-26, 27-39, and 40-52) I found that the deviation in mean age at death is highly significant (p<0.001) for those born in the second and the fourth quarters. The life span of people born between weeks 14 and 26 is 0.28 0.03 years below average; the life span of those born between weeks 40 and 52 is 0.32 0.03 years above average. For all major groups of causes of death including accidents and suicides, mean age at death of those born in the second quarter is significantly lower than that of individuals born in the fourth quarter (Table 1). As regards suicides the difference can partly be explained by the seasonal distribution of suicides: in the Austrian data set the number of suicides peaks between March and June and is lowest in December. A similar pattern was found by Avaline et al. [2] for France.

In the Danish data set further life expectancy at age 50 is 27.24 years. The average age at death is lowest around week 18, and it peaks in week 51. As was the case for Austrians, the life span of Danes born in specific weeks varies periodically around the mean (Figure 1). For those born in the second quarter, life spans are 0.17 years below average; for those born in the fourth quarter they are 0.13 years above average. This difference is statistically significant (Cox-Mantel statistic: p<0.001).

To test whether the season of birth affects survival up to the oldest ages, I compared the distribution of birth dates in different age groups. An excess mortality among people born during the first part of the year means that with advancing age the proportion of people born during the second part of the year will increase. In both countries the distribution changes significantly with age: at older ages relatively more people celebrate their birthday in the second part of the year than at younger ages (Figure 2a, 3).

In principle, the differences in life expectancy by month of birth could be caused either by factors that influence life span at the end of life, or by factors that work at the beginning of life. Two possible factors that may affect life expectancy at the end of life are the seasonal distribution of deaths and the "birthday-effect".

In both Denmark and Austria deaths peak in late winter and are lowest during the summer. In the Austrian data set the exact dates of death were redistributed such that in each week of the year the same number of people died while the rank order of the death-dates remained unchanged. Despite the transformation of the monthly death distribution, the excess mortality of those born in spring remains (see Figure 4).

Survival models 1a and 1b estimate the effect of the month of birth corrected for the impact of the current month. They reveal that differences according to month of birth remain, even when corrected for seasonally changing mortality risks. Neither the correction for period factors nor for cohort factors changes the result (Tables 2a, 2b). A model which included an interaction term between cohort and month of birth showed that the principal pattern of excess mortality for people born in spring is present in all cohorts. In particular, nothing unusual was found for those cohorts born shortly before or during World War I.

If it is the case that people have a higher risk of death shortly after their birthday, the risk of mortality should be highest in the first three months after their birthday and lowest in the three months before their next birthday. This pattern should be observed independently of the month of birth. The results of the models 2a and 2b (Tables 3a and 3b) show that those born in spring do in fact experience the highest mortality risk within the three months after their birthday. However, those born in autumn and winter tend to have an increased likelihood to die between four and twelve months after their birthday. This result leads to the conclusion that there is no "birthday effect" in Denmark. As a consequence the shorter life span of people born in spring is independent of the "birthday effect".

Methods Discussion

logo70.gif (2450 bytes)

Longevity and Month of Birth:
Evidence from Austria and Denmark

Gabrielle Doblhammer
1999 - 2000 Max-Planck-Gesellschaft ISSN 1435-9871