Deceleration in age-related mortality increase: an analysis by cause of death for the United States and an international comparison for the oldest-old

Roland Rau, University of Rostock

The exponential increase in mortality is levelling off at advanced ages. Differential selection effects in heterogeneous populations are one way to explain this decelerated mortality development: while the force of mortality on the individual level may continue to increase exponentially, frailer members of the population are dying on average younger, resulting in a slowing down in the age-related mortality increase on the observed level. In our paper, we investigate two predictions based on this heterogenity hypothesis as formulated by Horiuchi and Wilmoth (1998). (1) Due to general improvements in mortality over time, the population is less selected and, consequently, the observed deceleration in mortality shifts to later ages in more recent periods. (2) Most major causes of death should show a levelling off and, assuming independence of causes of death, the pattern of deceleration should be more pronounced for causes of death with higher death rates than with lower death rates due to the higher selection. We test those two predictions using two sources of data. The first hypothesis is scrutinized employing mortality data for eight selected countries for the oldest-old (80-100) from the Kannisto-Thatcher Database. The second hypothesis is analyzed by combining cause of death data for the United States from the National Center for Health Statistics with population data from the Human Mortality Database. We only find full support for the first hypothesis: the deceleration, which is observed for both sexes in all eight countries, appears to be shifted towards later ages. The second prediction, though, is only partly supported by our results: on the one hand, we find a deceleration for most causes of death. On the other hand, however, we can not find evidence that causes of death with high death rates have a more pronounced deceleration pattern.