To calculate health expectancies with the method of Sullivan also implies to make some approximations on the term of life. There are two problems:

- Because of the little sample on the 85 and beyond in the survey, it is accustomed to calculate a single prevalence ratio for this group age (see example of section 1.3). However on this group age, prevalence ratios are not constant but rapidly increasing with age.
- Often life tables are made up to age 100. Few information are
available after this age. Then, the following formulae are used:
with:

Obviously, is not precisely known [9]. More generally, quantification of mortality above the age of 80 and evaluation of survival at these ages raise substantial problems. A number of methodological and practical difficulties are involved. Most important of them are methodological and practical deficiencies of the measurement of old age mortality. For example, errors in data play much more serious role in the case of the elderly than in other bigger groups of the population. Errors are present in both the death and population statistics [5].

However, the biases are generally small compared to the standard errors and can safely be ignored in most cases.

Fri Apr 25 22:40:35 DFT 1997