Mortality estimates for South East Asia, and INDEPTH mortality surveillance: necessary but not sufficient?

In the June 2012 issue of the IJE, two diverse aspects of the global status of epidemiology were presented , in the form of a summary of health indicators for the World Health Organization (WHO) SouthEast Asia Region (SEAR) 1 and an overview of the International Network for the Demographic Evaluation of Populations and Their Health (INDEPTH) mortality surveillance activities in developing countries. 2 The former includes epidemiological estimates derived mainly from statistical models for about a quarter of the world's population, and the latter describes meticulous demographic surveillance activities to measure mortality and its causes, implemented in relatively small population clusters. Both approaches are of limited value in informing about population health status, the very purpose for which they have been presented. A pragmatic appraisal of the reasoning behind this judgement is essential, so that adequate remedial activities can be undertaken to improve the understanding of mortality patterns in developing countries. The mortality estimates presented for SEAR countries are not adequately qualified in terms of data sources and/or methods. Table 1 shows the primary data sources and methods used for deriving mortality estimates for SouthEast Asian countries. The extensive modelling used to develop these estimates was driven by the critical lack of current local vital registration data for SEAR countries. 3 Clearly there is very little credibility in these estimates, but this has been masked by the absence of any mention by Dhillon et al. of the details presented in Table 1. These details are at the least worrisome in view of the limited applicability of the demographic and epidemiological models used in the estimation process. The WHO life-table system uses two types of input parameters to predict a complete schedule of age-specific death rates for specific countries, depending on the availability of input data. 4–6 Type A inputs consist of country specific measures of the risks of under-5 mortality and adult mortality. Several countries only have local measures of under-5 mortality, in which case, this is first used in a separate regression to estimate a risk of adult mortality, and the two are then together used in the life-table system. Hence, Type B inputs consist of a country-specific measure of under-5 mortality, and a regression-based estimate of adult mortality. The WHO life table prediction model is based on empirical population and mortality data derived largely from western countries, which do not represent recent phenomena such as rapid declines …