Correction of Eyman et al. (1990) life expectancies
Citation: Eyman RK, Grossman HG, Chaney RH, Call TL (1990). The life expectancy of
profoundly handicapped people with mental retardation. New England Journal of
Due to an arithmetic error in their 1990 paper, Eyman et al. used mortality
rates about 3 times too high. Details on the mistake are given in the Appendix.
The following life tables were constructed based on correct computation of the
associated mortality rates.
An additional problem with the life tables, both the incorrect and
corrected versions, is the assumption that the functional abilities of
persons does not change (improve or decline). For example, those in the
lowest subgroup are assumed to remain in that subgroup until death (specifically,
permanent tube feeding and immobility are assumed); this may lead to a downward
bias in the results for younger persons, as many do make some improvement.
Appendix: Details on the mistake in the Eyman et al. (1990) article
- The authors examined mortality in a large California population of persons
with mental retardation or developmental disability during a study period of
roughly 3.5 years.
There were approximately 70,000 persons at risk per year.
There were approximately 650 deaths per year, or about 1,950 deaths in all.
The correct mortality rate is: 650/70000 = 9/1000. This means roughly 9 deaths
per year for each 1000 persons at risk.
The authors instead reported a mortality rate of 29/1000 -- about 3 times too
high. It appears that they mistakenly divided the total number of deaths
over the 3.5 year study period by 70,000.
Because the mortality rates in the article are about 3 times too high,
the life expectancies are much too low. In the most severe groups, the life
expectancies are only about 1/3 of their correct values.
This observation is not original to us. The problem had been pointed
out to Dr. Eyman in a letter from Drs. Shannon and Saigal as early as 1993.
In his reply to these correspondents, Dr. Eyman did not address their questions
about the apparent error.