Sensitivity analysis in multistate models: applications in health research
Mieke Reuser, Netherlands Interdisciplinary Demographic Institute (NIDI)
Frans Willekens, Netherlands Interdisciplinary Demographic Institute (NIDI)
Luc Bonneux, Netherlands Interdisciplinary Demographic Institute (NIDI)
Multistate models are becoming increasingly popular in health research including active aging, the dynamics of chronic diseases and disability. The dynamics are governed by transition rates and transition probabilities that are estimated from clinical or observational data. Changes in transition rates have an impact on survival, life expectancy and health indicators, such as the healthy life expectancy and the quality of life indicators. Sensitivity analysis studies the effects of changes in transition rates on health indicators. This paper applies matrix differentiation techniques to sensitivity analysis in multistate models of health and disability. Although these techniques were developed decades ago and have a great potential, they are rarely applied in population studies and health research. We demonstrate the sensitivity analyses using data from the US Health and Retirement Survey.