Changes in Mortality, 1990-2001
Change in patterns of mortality is a major determinant of the demography of populations and underlies important population differentials. For example, the differences in mortality by sex across regions contribute to the variable pattern of population sex ratios described earlier. The theory of demographic transition suggests that the rapid declines in fertility observed during the 1990s in most regions would be preceded, and perhaps accompanied, by a similarly rapid decline in child mortality. To help interpret the broad regional demographic patterns described earlier, a review of trends in mortality and the causes underlying such trends is useful.
Estimating Mortality
Various methods are available to estimate age patterns and levels of mortality in populations. These fall into three broad categories depending on the available data: direct estimation from complete vital registration, estimates from vital registration corrected for undercounting, and estimates derived from models based on child mortality levels. Mathers and others (2005) review the availability and quality of mortality data and group the 192 member states of the World Health Organization into broad categories according to criteria pertaining to the coverage, completeness, and quality of cause of death data. Their findings indicate that only about 33 percent (64) of World Health Organization member states, mostly high-income countries, have complete mortality data and that another 26 percent (50 countries) have data that can be used for mortality estimation purposes. The approximately 40 percent of remaining countries either have no recent data or no data at all that can be used to estimate causes of death or the level of adult mortality directly.
The situation is somewhat different for levels of child mortality, where decades of interest in monitoring child survival by the global public health community have yielded either direct or indirect estimates of child mortality for all but a handful of countries (Hill and others 1999; Lopez and others 2002). Based on a careful review of the time trend of these estimates of child deaths, which come primarily from censuses and surveys, estimating child mortality levels in 1990 and 2001 is possible for virtually all countries with an acceptable level of uncertainty. Levels of child mortality are unavailable for only about 10 countries that together account for about 2 percent of child deaths (Lopez and others 2002). Formal curve-fitting procedures to estimate time trends in child mortality can be applied to all the data, but given the subjective assessments that are required to judge which data points are plausible and which are not, simple averaging of all plausible observations at any given point in time is likely to be sufficient, and this was the procedure used to estimate child mortality levels for this chapter.
For those countries with complete vital registration data, age-specific and cause-specific death rates are easily derived directly from the registration data and from population censuses. For those countries where registration data are incomplete, demographers have developed indirect demographic methods to correct for underreporting of deaths before estimating age-specific mortality (Bennett and Horiuchi 1984; Hill 1987). These countries include China and India, where application of such methods suggest that data from the disease surveillance points system in China and the sample registration system in India are 85 to 90 percent complete (Mari Bhat 2002; Rao and others 2005).
For countries with no usable data on adult mortality levels, age-specific death rates were predicted from the modified logit life table system (Murray and others 2003). The median level of adult mortality was predicted based on a modeled relationship between adult and child mortality as determined from a historical data set of more than 1,800 life tables judged to be reasonably complete. Uncertainty about these predicted mean values of adult mortality is considerable given the few observations with comparatively high levels of child and adult mortality. The estimated and predicted levels of child and adult mortality, respectively, were then applied to the modified life table system by selecting the best match from among 50,000 life tables to estimate a complete, smoothed set of age-specific death rates (Murray and others 2003). This method was applied for all but about 70 countries.
Obvious uncertainties are associated with this procedure. Hence, the life tables for East Asia and the Pacific, the Middle East and North Africa, and Sub-Saharan Africa (where HIV/AIDS mortality was added to the predicted adult mortality rates) in particular need to be viewed with caution, because the rates for many countries in these regions have been modeled using these methods.
Identical methods were applied to estimate national age-specific mortality rates for both 1990 and 2001; thus, the two sets of estimates are, in principle at least, comparable. Annex 2A provides detailed estimates of summary measures of mortality by country for the two years based on these methods. The annex also shows the percentage decline in child mortality during the period.
[Table .]
Whether these methods correctly describe levels and patterns of mortality is difficult to ascertain given the substantial uncertainties in the data, particularly for adult mortality. The only other systematic attempt to estimate national and global death rates in 1990 is that of the UN Population Division (United Nations 2003). Figure 2.3 presents estimated mortality parameters for 1990 by region. For a comparison of mortality estimates for 2001, see Lopez and others (2002).
[Figure
2.3]
Despite the UN's different model life table approach for estimating age-specific death rates based on child mortality, the two sets of estimates shown in figure 2.3 are remarkably congruent. Regional estimates of child mortality 5q0 (the mortality risk for children under five years of age) are virtually identical, with a possible exception being the UN's slightly higher levels of child mortality for East Asia and the Pacific (which is dominated by China). This congruence is not unexpected given the intense collaborative efforts of the past five years or so by the World Health Organization, the United Nations Children's Fund, the United Nations, and the World Bank to agree upon a common interpretation of the extensive data available on trends in child mortality in low- and middle-income countries.
Somewhat surprisingly given the quite different methodological approaches, regional estimates of adult mortality 45q15 (the mortality risk for adults between the ages of 15 and 60) are remarkably similar, with our estimates tending to be slightly higher in the Middle East and North Africa and South Asia for males and slightly lower in the same regions for females. That is, we have estimated larger sex mortality differentials among adults than the UN on the basis of observed patterns of mortality where data were available (as in the Arab Republic of Egypt and India), and where not, on the basis of observed differences in child mortality for boys and girls. Some investigators expect male excess mortality to increase with social development and economic growth (Bhatia 1983), but whether this is better reflected in our estimates or those of the UN is not clear. In any case, the differences are minor. Significantly more disagreement is apparent for Sub-Saharan Africa, with the UN estimates of adult mortality in 1990 being one-quarter to one-third higher than ours. This is obviously uncertain given the sparse data available on adult mortality in the region and the fact that the HIV epidemic in Sub-Saharan Africa was well established by then, and hence a higher estimate may be justified. Recent evidence, however, has suggested that basing mortality estimates on prenatal clinic data may well lead to an overestimation of death rates due to HIV.
Differences in methodology and adjustment criteria appear to have the greatest effect at older ages, especially for males. The UN estimates indicate significantly higher mortality in high-income countries at ages 60 to 79 even though complete vital registration data are available for virtually all the countries except some of the small Gulf states. Differences in estimated mortality for the Sub-Saharan Africa region are not unexpected given the differences reported for younger adults, and are less extreme than at ages 15 to 59, as one might expect given that HIV/AIDS mortality is not of major consequence for older ages. Otherwise, estimates for females at older ages agree quite closely, but the UN's are significantly higher than ours for Latin America and the Caribbean and significantly lower for East Asia and the Pacific and South Asia. These differences arise because the model life table methods used by the UN tend to shift deaths from younger to older adult ages at lower levels of child mortality (Latin America and the Caribbean) and the converse at higher levels of child death rates (East Asia and the Pacific and South Asia).
Overall, as figure 2.3 demonstrates, the age patterns are largely compensatory, with the result that estimates of life expectancy at birth for the two series are remarkably similar for both males and females, with the notable exception being Sub-Saharan Africa, where the higher adult mortality assumptions favored by the UN result in life expectancies at birth that are about 2.5 years lower than ours for males and 5.0 years lower for females.
Trends in Mortality Levels
The 1990s were characterized by significant economic gains in most regions, with growth in gross national product per capita ranging from 18 percent in South Asia and Sub-Saharan Africa to more than 100 percent in East Asia and the Pacific and the Middle East and North Africa (table 2.2). Overall, gross national product per capita grew by about 35 percent in low- and middle-income countries during the decade. One would expect this to have led to a significant improvement in life expectancy, and this indeed occurred in most regions with the notable exception of Europe and Central Asia and, in particular, Sub-Saharan Africa (table 2.3). In the former region, life expectancy was largely unchanged over the decade, primarily because of the massive rise in adult mortality in countries such as the Russian Federation and its neighbors during the first part of the decade, which negated the declines in child mortality. Much of this extraordinary increase in adult mortality, which rose by about 50 percent between 1987 and 1994, has been attributed to alcohol abuse, particularly among men (Leon and others 1997; Shkolnikov, McKee, and Leon 2001).
[Table .]
Economic development and better coverage of the population with essential child health services have ensured continued declines in levels of child mortality, as measured by the risk of death from birth to age five, in all regions. The notable exception is Sub-Saharan Africa, where child mortality among girls remained unchanged at around 165 per 1,000, with only a modest decline (5 percent) in the risk of death for boys. The absence of significant declines in child mortality in the 1990s in Sub-Saharan Africa is most likely largely due to the impact of HIV/AIDS. Overall, the risk of child death declined from 90 per 1,000 in 1990 to 80 per 1,000 in 2001, with the risk being remarkably similar for males and females (table 2.3); however, the differential in child mortality between the world's richest and poorest populations is stark, with a newborn in Sub-Saharan Africa facing 25 times the risk of death before the age of five than a newborn in a high-income country.
Despite the much greater uncertainty in relation to levels of adult mortality compared with those for children, the estimates shown in table 2.3 nonetheless indicate substantially different trends in adult mortality across different regions between 1990 and 2001. For most regions, the risk of death between ages 15 and 60 fell by about 10 to 17 percent over the decade. This was not the case in Europe and Central Asia, where policy shifts, particularly in relation to alcohol, together with broader social change, have largely been responsible for the 15 percent rise in adult male mortality and the 6 percent increase in the risk of death for women. Note that these estimates mask the large cyclical fluctuations in adult mortality in Russia, in particular, that characterized the region's mortality trends in the 1990s.
Table 2.3 also reveals the large increase in adult mortality in Sub-Saharan Africa, which was due primarily to the unfolding of the HIV/AIDS epidemic in southern Africa. Notwithstanding the substantial uncertainty surrounding these estimates, the epidemic appears to have been of proportionately greater consequence for women, with the rise in their risk of death (67 percent) being twice that of males, among whom other causes of death such as violence were more common. If these estimates are correct, then 52.0 percent of African males reaching age 15 and 44.0 percent of females will die before their 60th birthdays, compared with, for instance, 6.5 percent of women in high-income countries, who despite their already low risk enjoyed a further 11 percent decline in mortality during the 1990s. These reversals in mortality decline have effectively negated gains elsewhere, with the result that the global risk of adult death has remained essentially unchanged for males, and may even have risen slightly for females.
Taken together, the probability of death up to the age of five and between the ages of 15 and 60 are a better reflection of the risk of premature death than either alone, although both have particular public health implications. One might argue that health policy should be equally concerned with keeping adults alive into old age as it is with keeping children alive into adulthood. A convenient metric in this regard is the risk of death between birth and age 60 (table 2.3). In high-income countries, given 2001 mortality rates, only about 7 percent of females and 13 percent of males would be dead by age 60, compared with 55 percent of females and 62 percent of males in Sub-Saharan Africa. Significant improvements in this summary measure of premature death can be observed in all regions except Europe and Central Asia and Sub-Saharan Africa. Worldwide, the index appears to have improved slightly for males and not at all for females.
Other features of global mortality summarized in table 2.3 are worth highlighting. First is the impressive evidence of a continued decline in mortality among older age groups in high-income countries that began in the early 1970s. The risk of a 60-year-old dying before age 80 declined by about 15 percent for both men and women in high-income countries so that at 2001 rates, less than 30 percent of women who reach age 60 will be dead by age 80, as will less than 50 percent of men. Second, crude death rates in East Asia and the Pacific, Latin America and the Caribbean, and the Middle East and North Africa are lower than in high-income countries, reflecting the impact of the older age structure of rich countries, and are particularly low in Latin America and the Caribbean. Third, the proportion of deaths that occur below age five, while declining in all regions, varies enormously across them, from just over 1 percent in high-income countries to just over 40 percent in Sub-Saharan Africa. In some low- and middle-income regions, particularly East Asia and the Pacific, Europe and Central Asia, and Latin America and the Caribbean, the proportion is well below 20 percent. The net effect of these changes in age-specific mortality since 1990 has been to increase global life expectancy at birth by 0.7 years for females and by about twice this for males: a modest scorecard.
