Analytical sensitivity analysis of the multistate life table: a new tool in public health research

Frans Willekens, Netherlands Interdisciplinary Demographic Institute (NIDI)
Mieke Reuser, Netherlands Interdisciplinary Demographic Institute (NIDI)
Luc Bonneux, Netherlands Interdisciplinary Demographic Institute (NIDI)

The sensitivity of life expectancy to changes in the mortality rate and the related concept of entropy were introduced in demography by Demetrius and Keyfitz in the 1970s. To analyze the sensitivity of healthy or diseased life expectancy, one can use a multistate life table and assess the effect of changes in underlying rates. The object of this paper is to demonstrate analytical sensitivity analyses of a multistate life table applied to an illness-death model to quantify the effect of a health change in terms of compression or expansion of disability. Most studies express the effect of a health change in terms of hazard ratios or relative risks compared to a standard without change. Policy makers are more interested in the effects of health changes on future disability, and how disability can be controlled. The object of this paper is to demonstrate analytical sensitivity analyses of a multistate life table applied to an illness-death model to quantify health changes in terms of compression or expansion of disability. We will demonstrate the strength of analytical sensitivity analysis by an application of a multistate illness-death model to the U.S. Health and Retirement Study, showing its usefulness for prevention and intervention decisions.