The returns of longitudinal information on the elderly: a micro simulation model to estimate errors in cross-sectional estimates of disability-adjusted life expectancy

Mercedes Ayuso, Universitat Autònoma de Barcelona
Montserrat Guillén, Universitat Autònoma de Barcelona
Malena Monteverde, Universitat Autònoma de Barcelona

We evaluate the accuracy of alternative methods to assess life expectancy in disability among elderly people. We pose a three-state model (healthy(H), disabled(D), dead(M)) with four transitions (H to D, D to M, H to M, and D to H) and use Monte Carlo simulations to generate individual trajectories. The rates in the simulations reflect the experience of the US population (HRS). We apply two methods to estimate life expectancy in disability, each one of them using only partially observed information: (a) Sullivan’s calculations; (b) a new procedure to estimate transition probabilities from observed prevalence and overall mortality. We compare resulting estimates with the true values embedded in simulations under conditions departing from the assumptions invoked by the methods. We focus on violation to the assumptions of homogeneity of risks and no recovery. We are also assessing robustness to lack of stationarity.