Decomposing a Sullivan Expectancy: An Indirect Incidence-Based Approach

Timothy Riffe , Universidad del País Vasco & Ikerbasque (Basque Foundation for Science)
Rustam Tursun-zade, University of the Basque Country

Background: A lifetable and health prevalence are sufficient to calculate health expectancy (HLE) using the Sullivan method (Sullivan, 1971). Available methods for decomposition of differences in Sullivan HLE (Nusselder & Looman, 2004; Shkolnikov, Andreev, et al., 2017), do not isolate health and mortality effects because (i) prevalence could be driven by unobserved mortality given a health state, and (ii) overall mortality is a prevalence-weighted average of health-specific mortality. Objectives: We aim to (i) transform Sullivan inputs into state-specific mortality schedules consistent with a given mortality rate ratio and (ii) convert prevalence into incidence for the case of health deterioration without the possibility of recovery. Decompositions based on these parameters can clearly separate health and mortality effects. Methods: Our approach is based on indirect separation of mortality by state, leveraging an assumed mortality rate ratio. Transition probabilities representing health deterioration are then inferred following multistate lifetable accounting. HLE calculated from these transitions is identical to that derived using the Sullivan method, but decomposition results can now be based on incidence. Results: Both marginal sums and age patterns of health and mortality effects are different when we decompose using indirectly derived multistate parameters versus Sullivan parameters. Conclusions: We give an indirect method to transform Sullivan HLE inputs into a multistate model with an irreversible health state and show how to decompose obtained expectancies in terms of these new parameters. Although we do not directly observe the mortality rate ratio, we argue that our method is nevertheless more informative than existing Sullivan decompositions.

See paper

 Presented in Session 73. Innovations in Demographic Modelling and Projections