Healthy life expectancy, mortality, and age prevalence of morbidity

In calculating period healthy life expectancy, the use of age-specific morbidity prevalence patterns assumes that age captures the important time-variation in the given health condition, i.e. that the disabling process is related to how long an individual has lived. However, many morbidity patterns are better classified by time-to-death. At advanced ages the conflation of an increasing chronological-age mortality pattern and a time-to-death morbidity pattern produces an apparent morbidity pattern that increases with advancing age. Differences in period healthy life expectancy over time or between populations cannot easily be partitioned into morbidity and mortality components because the period morbidity pattern may depend on an unknown future time-todeath process not captured by period mortality. We illustrate these concepts formally and empirically, using morbidity data from the U.S. Health and Retirement Study. While holding the time-to-death morbidity pattern fixed, we show that mortality reduction alone reduces the total life years with disability. We estimate the magnitude of this bias for different realistic morbidity patterns. This has implications for any betweenor within-population comparisons of period healthy life expectancy conditioned on different age patterns of mortality.