Estimating health expectancy in presence of missing data: an application using HID survey

The debate on ageing in Europe is currently paying considerable attention to the health expectancy (HE) of elderly. When the calculation of health expectancy is based on longitudinal surveys, the method of multi-state tables could be used, based on the analyses of the transitions between successive states of health. Information on health status is given by interview surveys and can't be measured after “last” wave and statistical analyses of such highly “censored data” is now well performed. But loss of follow-up “within” two waves is another matter which can induce biases in the statistical results. The aim of this study is to estimate the probability of transition among different health states in the presence of missing data, comparing different estimation methods, and discuss the effect on the estimation of health expectancy.