Trends in Causes of Child Death, 1990-2001
The estimation of cause of death patterns for world regions will, for the foreseeable future, be substantially uncertain given the paucity of data on medically certified deaths in many low- and middle-income countries (Mathers and others 2005; Sibai 2004). Verbal autopsies, that is, structured interviews with relatives of the deceased about symptoms experienced prior to death, will not yield the diagnostic accuracy achievable with medical certification based on good clinical case histories and medical records. This is not to deny that verbal autopsies can meet broad policy needs for information about causes of death, particularly with clinical input into the coding of interviewees' responses, but their reliability for diagnosing leading causes of child death is questionable (Snow and others 1992). Thus, estimates of child mortality derived from proportionate mortality models that are based largely on verbal autopsies need to be viewed with caution (Lopez 2003; Morris, Black, and Tomaskovic 2003).
Yet, despite these concerns about the quality of cause of death data, investigators can more confidently assess the comparative magnitude of causes of death for children than for adults. The fact that the demographic "envelope" of child deaths is reasonably well understood in all regions limits excessive claims about deaths due to individual causes, a constraint that is not a feature of adult mortality given the relative ignorance of age-specific death rates in many countries. In addition, the need for data on cause-specific outcomes to assess and monitor the impact of various child survival programs in recent decades has led to a reasonably substantial epidemiological literature that might permit cause-specific estimation, but under an unacceptably large number of assumptions (Black, Morris, and Bryce 2003).
A critical feature of any estimation exercise is a rigorous assessment of data sets for biases, study methods, and gen-eralizability of results. Investigators have undertaken a number of efforts to estimate the causes of child mortality over the past decade or so (Bryce and others 2005; Lopez 1993; Morris, Black, and Tomaskovic 2004; Williams and others 2002), but undoubtedly the most comprehensive was the study by Murray and Lopez (1996) and its 2001 revision (chapter 3 in this volume). Both the latter Global Burden of Disease (GBD) studies apply methods to force epidemiolog-ical consistency according to the evidence available for each region, and inevitably the constraint of demography has meant that the GBD estimates of cause-specific mortality will differ from those developed largely independently of other causes. That is, the GBD estimates of specific causes of death are constrained to sum to the number of deaths derived from demographic analyses, whereas cause-specific estimates that are derived in the absence of such demographic constraints are unbounded and tend to be inclusive at the margin rather than exclusive. Differences in regional estimates between 1990 and 2001 arise in part because the countries included in the regions differed and, more important, because of better information for more recent periods. Yet, despite improved information, the true level of child death rates from major causes such as malaria and perinatal conditions (birth trauma, birth asphyxia, sepsis, and prematurity) remains largely unknown.
Notwithstanding methodological differences and uncertainties, deriving implied estimates of trends in the leading causes of child mortality is possible by comparing results from the two GBD studies, and these are summarized in table 2.4. These estimates have been simply obtained as the difference between the regional estimates for 1990 and 2001, but the implied pattern of change is interesting nonetheless. The conversion of the 1990 regional GBD estimates (Murray and Lopez 1996) to the regions used for the 2001 assessment was done simply by population weighting, a very approximate procedure. By contrast, the 2001 estimates were prepared as regional aggregates of country-specific estimates (see chapter 3,) and this has undoubtedly affected comparisons further.
[Table .]
Global mortality from malaria increased by 0.5 million during the 1990s, with 80 percent of the deaths occurring in Sub-Saharan Africa. The proportion of all child deaths due to malaria doubled from 5 percent in 1990 to 10 percent in 2001 worldwide and increased from 15 percent in 1990 to 22 percent in 2001 in Sub-Saharan Africa. The only other causes that appear to have increased are HIV/AIDS in Africa, a reasonable conclusion given female prevalence levels, and the category of perinatal conditions, which are strongly dependent on the quality and availability of prena-tal services. Causes that appear to have declined substantially include acute respiratory infections (2.5 million to 1.9 million deaths or 15 percent of all child deaths), diar-rheal diseases (2.4 million to 1.6 million deaths or 13 percent of child deaths), measles (0.8 million to 0.5 million deaths or 5 percent of child deaths), and injuries (0.6 million to 0.3 million deaths or 2 percent of child deaths).
The implied pattern of change in the risk of child death varies across regions for all major conditions listed in table 2.4, particularly with regard to the magnitude of change. This can be seen more clearly from figure 2.4, which summarizes these trends for broad regional aggregates and for Sub-Saharan Africa. In general, the absolute change in risk of death has been greater in Sub-Saharan Africa than elsewhere, both for causes with increased risk (HIV/AIDS, malaria) and where risk has declined (diarrheal diseases, measles).
[Figure
2.4]
While these changes may be in accord with what is known about regional health development and economic growth, they need to be confirmed. Some of the suggested changes warrant further investigation, for example, death rates from perinatal causes appear to have risen in both East Asia and the Pacific and South Asia and remained unchanged in Latin America and the Caribbean, which may or may not be in line with what is known about developments in prenatal care and safe motherhood initiatives. Similarly, measles appears to have disappeared as a cause of child death in Latin America and the Caribbean. The risk of child death from congenital anomalies appears to have risen in both Latin America and the Caribbean and the Middle East and North Africa, but why is unclear. Similarly, the large suggested declines in the risk of child deaths because of injury in South Asia and Sub-Saharan Africa appear unlikely and may largely reflect better data and methods for measuring injury deaths.
