Compositional model for mortality by age, sex and cause

Reliable information on deaths by cause is an essential input for planning, managing and monitoring the performance of the health sector in all countries. Detailed estimates of mortality patterns disaggregated by age and sex for over 100 specific causes have been published at the global and regional level (Murray and Lopez 1996, 1997) as well as at the national or sub-national level for a number of different countries (Bowie et al 1997, Concha et al 1996, Lozano et al 1995, Mathers et al. 1999, Murray et al. 1997, 1998, Ruwaard and Kramers 1998). A range of different data sources have been exploited for the estimation of mortality by cause, including vital registration systems, sample registration systems, and epidemiological estimates. Where these information sources are not available, however, indirect techniques for estimating cause-of-death patterns can be useful. For the Global Burden of Disease Study 2000, a number of different health indicators will be calculated for every country, including mortality by age, sex and cause and measures of disease burden such as Disability-Adjusted Life Years (DALYs), which require cause-of-death data as inputs. As a critical step in this exercise, an extensive effort has been launched to compile all existing data sources on mortality by cause in as many countries as possible, and WHO continues to augment and refine its comprehensive mortality database. An estimated 17 million deaths, or around thirty percent of all deaths, are currently covered by registration systems. Inevitably, large gaps remain, particularly in high-mortality areas. In order to address these information gaps, models for estimating broad cause-of-death patterns can serve as the starting point for indirect methods of estimating attributable mortality for a comprehensive list of detailed causes. In this paper we describe the development of new models for estimating broad patterns of mortality by age and sex in areas without adequate death registration data. In the following sections, we review previous work on developing cause-of-death models, including the Global Burden of Disease 1990 study, describe the development of new statistical models for cause-of-death data, present one application of these models based on regional variation in cause-of-death patterns, and discuss further directions for research in this area. (Extract of the publication)