Cause-of-Death Decomposition of Old-Age Mortality Compression

Shiro Horiuchi, Hunter College, City University of New York (CUNY)
Siu Lan Karen Cheung, University of Hong Kong
Jean-Marie Robine, Institut National de la Santé et de la Recherche Médicale (INSERM)

Previous studies have shown that the significant decline in old-age mortality in high-income countries during the last few decades was accompanied by mortality compression among the oldest old. The compression is associated with the accelerating pace of age-related mortality increase, which can be more properly measured by the logistic rate of mortality rise, denoted by b, than by the conventional exponential rate. In this paper, we decompose the increase in b in France during the ICD9 period (1979-1994) into level effects and slope effects of cause-specific mortality, using the line-integral method of decomposition. The results show that the rise in all-cause b is due to slope effects: b increased for a wide range of various causes, including both high-b and low-b causes. Because causes with high b's are generally considered to be closely related to senescent processes, we predicted less pronounced declines in mortality from those causes, but found no such tendency.