Time-to-Death Patterns in Markers of Age and Dependency

Pil H. Chung, University of California, Berkeley
Jeroen J. A. Spijker, Center for Demographic Studies (Barcelona)
John MacInnes, University of Edinburgh

We aim to determine the extent to which variables commonly used to describe health, well-being, and disability in old-age vary primarily as a function of years lived (age), years left, or as a function of both. We analyze data from the US Health and Retirement Study to estimate chronological age and time-to-death patterns in 60 such variables. We describe results for individuals with 15 or fewer remaining years of life, after age 70 and whose completed lifespans do not exceed 100. Our results show that most markers used to study well-being in old-age vary along both time dimensions, but that variation over remaining years of life is typically much stronger for the age-ranges and variables examined.