In the past decade, there has been a marked increase in the development, calculation and use of summary measures of population health, which combine information on mortality and non-fatal health outcomes. This study seeks to improve the conceptual, methodological and empirical basis for the calculation of summary measures. The starting point is the evaluation of different summary measures along a set of basic criteria, such as the requirement that a summary measure should improve if, for example, age-specific mortality in a population declines, ceteris paribus.

    New population-based data are being collected on one of the critical inputs to all summary measures of health, namely valuations of health states worse than perfect health. Valuations of a range of different states, along with a rich set of data on self-reported health status in various domains, self-reported general health status, socio-demographic characteristics and, through record linkage, medical diagnoses and treatment outcomes, also are being collected in a nationally-representative sample survey in Denmark. Analyses are being undertaken on the relationships between health state valuations and performance in various domains of health. Critical components of the project are multivariate analyses to examine how valuations of different health domains may vary as a function of age and other socio-economic variables. The national survey in Denmark also provides a link to the project on comparing self-reported and observed measures of health status, allowing estimation of the relationships between self-reported health and health state valuations. The findings from the Denmark study will inform the next phase of this project, which is a series of pilot tests in field sites in Tanzania and Mexico with the objective of extending health state valuation protocols to developing countries.

    Using the data on health state valuations and epidemiological data produced by the WHO Global Burden of Disease 2000 Network, various summary measures are being computed. Simulation algorithms are being developed to estimate confidence intervals around summary measures as a function of uncertain epidemiological and preference inputs.