All-cause mortality

The first step that must be undertaken by any study of the burden of disease is to make or obtain estimates of all-cause mortality by age and sex for every regional or national population considered by the study.

The 2010 GBD study produced its own estimates of mortality for 187 countries for the period 1970 to 2010 based upon a reanalysis of the primary evidence (Wang, Dwyer-Lindgren, Lofgren et al. 2012). In order to calculate DALYs for a more detailed set of age groups than earlier GBD studies, it estimated mortality separately for the early-neonatal, late-neonatal and post-neonatal periods of infancy, for children aged 1-4, and for five-year age groups up to an open-ended age interval of 80+.

The counts of deaths by age and sex obtained by multiplying the age-specific rates by population estimates represent a mould to which the rest of the GBD estimates are fitted. In almost all populations, all-cause mortality can be estimated more accurately than cause-specific mortality and morbidity. Thus, the analysis should produce more reliable results if, instead of calculating the all-cause estimates as the sum of the cause-specific estimates, the latter are constrained to sum to the former. In particular:

• the number of deaths by age and sex from each disease and injury are constrained to sum to the number of deaths from all-causes in each age group and sex
• the amount of age- and sex-specific disability resulting from each type of disease and injury is constrained to lie in an epidemiologically plausible ratio to the number of deaths by age and sex from the same condition.

In combination, these two constraints largely address the problem discussed earlier of disease-specific interest groups making inflated claims about the burden of disease from the condition in which they specialise. Of course, constraining the GBD estimates in this way cannot ensure that they are correct. It should ensure, however, that they are both internally consistent and plausible.

Unfortunately, estimating abridged life tables remains a difficult challenge for countries in which not all deaths are registered routinely. In particular, while survey and census based estimates of under-five mortality are available for almost all countries, the only way of estimating adult mortality rates in many less-developed countries is by resorting to the use of model life tables.