Improved measures of population health: Heterogeneity in active life expectancy.

Syracuse University, Syracuse NY

Investigators

Abstract

PROJECT SUMMARY/ABSTRACT Life expectancy is an important and closely-watched summary measure of population health. Active Life Expectancy (ALE) extends this indicator, partitioning it into two components: the proportion lived without disability (or ?active?) and the remaining proportion, that lived with disability (or ?inactive?). Measures of ALE are used to test the ?compression of morbidity? hypothesis, to investigate between-group differences in health outcomes, to address the consequences of early-life circumstances on health inequalities later in life, and in many other areas of scientific and public-health inquiry. There are two principal methodological approaches in contemporary use with which to calculate ALE, the ?Sullivan? method and a ?transitions? method. Each has strengths and weaknesses, but each also requires the adoption of some unrealistic or restrictive assumptions; moreover, neither permits the quantification of unmeasured heterogeneity, or within-group variability, in ALE. This project will develop a new statistical model with which to compute ALE, one that overcomes many of the limitations of the methods now in widespread use. Specifically, this project will (1) develop a statistical modeling framework, and produce user-friendly estimation software, with which to jointly estimate a latent-class trajectory model of disability and a heterogeneous model of mortality, recognizing the interdependencies between these two phenomena; (2) apply the model to three different data sources?the second Longitudinal Survey on Aging (LSOA 2), the Health and Retirement Study (HRS), and the National Health and Aging Trends Study (NHATS)?in such a way as to carry out comparative analyses of ALE across time periods and for alternative survey design features; and (3) analyze between-group differences in ALE, focusing on both socio- demographic factors (such as sex, education, and race) and on early-life factors shown in past research to have consequences for late-life health and disability. In these applications we will pay particular attention to the relative size of between- and within-group differences in ALE. The new model draws on a large and rapidly-growing literature in biostatistics and epidemiology that deals with joint models of repeated health-outcome measures and a terminal event such as death. For the health-outcome part of the joint model, we will adapt a latent-class trajectory approach, one that recognizes the existence of several distinctive age-profiles of disability prevalence. In our joint model, each of these ?disability groups? will also have its own distinctive age profile of mortality risks. Together these two components will produce a set of ALE measures, representing unmeasured heterogeneity in ALE.