Commentary: Estimating the causes of child deaths.

Abstract

There is no question that preventing child mortality should be a priority for all populations. There are moral and philosophical reasons to do so (the ‘fair innings’ argument). The replacement of populations depends on sufficient children surviving to reproductive age. It is also critical for indicators of overall population development: reductions in infant and child mortality exert substantially greater impact on levels of life expectancy than similar proportionate declines in death rates at older ages. The long-term decline in child mortality in the industrialized countries that has accompanied economic and social development has depended in part on the implementation of technologies and programmes that target specific diseases, or groups of diseases that often co-occur in the sick child. A reliable information system to monitor progress with programme implementation, and to highlight areas where greater, or new investment is needed, is a key support for public policy to improve child survival. But how reliably do we know the causes of child death in different countries, particularly in poorer populations? Arguably, the most appropriate mechanism for understanding change in the leading causes of death is a nationally representative vital registration system which routinely captures all deaths and assigns an underlying cause, certified by a medical practitioner. Such systems are well established in all industrialized countries, and increasingly in many developing populations as well, but for over half the countries of the world, particularly in Africa and Asia, vital registration of the fact of death, let alone the cause, is woefully inadequate for public policy. 1 But does the absence of a functioning vital registration system imply that little is usefully known about the leading causes of child death in populations? Certainly not, as the article by Morris and colleagues in this issue demonstrates. 2 Vital registration, even on a sample basis as is done in China, 3 provided it is implemented with diligence and a reasonable understanding of its benefits, is an invaluable source of data on causes of death at all ages. Unfortunately, the achievement of sufficient quality of vital registration appears to be largely dependent on statistical infrastructure investments for the routine monitoring of vital events; investments which are generally of lowest priority for Ministries of Health, or even more removed, for Ministries of the Interior charged with maintaining vital registration. What are the most appropriate strategies to meet the needs of countries for monitoring trends in causes of child mortality, and to guide programme implementation and priority setting? A useful first step is to understand, with reasonable certainty, the overall level of child mortality. Knowing how many children die is of itself useful for policy, but, more importantly, this figure will constrain the individual cause-specific estimates within the bounds of additivity. With the substantial investment in measuring child mortality levels through censuses and survey programmes over the past few decades, levels of child mortality are reasonably reliably known for all but a handful of countries, primarily in Africa. 4 The Demographic and Health Surveys programme alone has provided good quality data on levels of child mortality in over 40 countries over the past 5 years. Information on the age distribution of these child deaths can be used to guide the estimation of causes of death, as for example has been done to separate out the main causes of neonatal deaths. 5 Enough is known about the relationship between broad causes of death and overall levels of child mortality to be reasonably confident about leading causes. This was initially modelled by Preston, 6 with subsequent adaptation of his work to child mortality. 7 But such models are unlikely to capture the important and real variations in causes of child death across populations that occur due to such factors as climate, wars, efficiency and coverage of health services, the differential effects of education, and the emergence of new hazards such as human immunodeficiency virus (HIV)/AIDS. This is best assessed on the basis of a systematic review of community studies and other local epidemiological research according to strict evaluation criteria. There are several examples in the literature, ranging from disease-specific reviews, e.g. for malaria, 8 to systematic assessments of leading causes of child mortality at earlier