WHO introduced healthy life expectancy as a summary measure of the level of health attained by populations in the World Health Report 2000 . The World Health Report 2001 reports estimates of total life expectancy (LE) and healthy life expectancy (HALE) by sex for the 191 Member States of WHO. Calculation of HALE for WHO Member States requires three inputs. First, life expectancy at each age is calculated using the standard life table approach. Second, estimates of the prevalence of various states of health at each age are required . Finally, a method of valuing this time compared to full health must be developed .
Part of the utility of LE and HALE is that each provides an easily interpretable summary of numerous different pieces of information age specific mortality rates in the case of LE, with the addition of information on non-fatal health outcomes in HALE. As such, they also represent the product of numerous sources of uncertainty. Presentation of the results in the World Health Report include intervals around the LE and HALE estimates in order to convey the levels of uncertainty in these estimates due to limitations in the data inputs. This discussion paper describes the methods used to quantify uncertainty in the various inputs to LE and HALE estimates, and how these uncertainties were incorporated in intervals around the reported measures.
Modern epidemiological techniques report confidence or uncertainty intervals around all estimates, and there is a growing literature on methods for capturing uncertainty in quantitative policy analyses (see, for example, Morgan and Henrion , King et al. , Vose . Confidence intervals and uncertainty intervals provide explicit characterizations of the precision around estimates derived from limited information sources. For some simple quantities of interest, such as linear combinations of normally distributed random variables, uncertainty may be reasonably captured using analytic methods. For more complex quantities, however, a simulation-based approach offers a more feasible solution.
In the following sections, we introduce our general approach to propagating uncertainty in LE and HALE, then describe in detail the sources of uncertainty in each of the key components of these measures. We describe the methods used to combine the uncertainty from each input into the computations of HALE, and end with a discussion of continuing work and future directions in this area.
Download
PDF Ebook Methods for life expectancy and healthy life expectancy uncertainty analysis
