Theme Papers
The decline in child mortality: a reappraisal
Omar B. Ahmad,1 Alan D. Lopez,2 & Mie Inoue 3
The present paper examines, describes and documents country-specific trends in under-five mortality rates (i.e., mortality among children under five years of age) in the 1990s. Our analysis updates previous studies by UNICEF, the World Bank and the United Nations. It identifies countries and WHO regions where sustained improvement has occurred and those where setbacks are evident. A consistent series of estimates of under-five mortality rate is provided and an indication is given of historical trends during the period 1950-2000 for both developed and developing countries.
It is estimated that 10.5 million children aged 0-4 years died in 1999, about 2.2 million or 17.5% fewer than a decade earlier. On average about 15% of newborn children in Africa are expected to die before reaching their fifth birthday. The corresponding figures for many other parts of the developing world are in the range 3-8% and that for Europe is under 2%.
During the 1990s the decline in child mortality decelerated in all the WHO regions except the Western Pacific but there is no widespread evidence of rising child mortality rates. At the country level there are exceptions in southern Africa where the prevalence of HIV is extremely high and in Asia where a few countries are beset by economic difficulties. The slowdown in the rate of decline is of particular concern in Africa and South-East Asia because it is occurring at relatively high levels of mortality, and in countries experiencing severe economic dislocation. As the HIV/AIDS epidemic continues in Africa, particularly southern Africa, and in parts of Asia, further reductions in child mortality become increasingly unlikely until substantial progress in controlling the spread of HIV is achieved.
Keywords: infant mortality, trends; cause of death, trends; health surveys; models, statistical; European region, South-East Asia, Africa, Americas, Latin America, Western Pacific.
Introduction
The 20th century witnessed dramatic declines in mortality in almost all countries of the world, regardless of initial levels, socioeconomic circumstances and development strategies. In the advanced economies the declines were already apparent at the end of the 19th century. In the developing countries, substantial declines did not take place until shortly after the end of the Second World War. The magnitude of the initial declines in the developing countries was so impressive as to engender widespread speculation during the 1960s and 1970s that the mortality gap between the developed and developing countries would narrow significantly by the end of the 20th century (1). This view was encouraged by the fact that some of the most pronounced reductions were occurring in countries with relatively low gross national product (1, 2).
Since the downward trend in mortality appeared to be a worldwide phenomenon, little interest was shown in critically defining the specific factors responsible for it. Consequently, both national and donor resources were aimed more broadly at achieving further reductions. For instance, various child survival programmes, such as USAIDs Child Survival Initiative, set loosely defined strategies for achieving certain child mortality targets among participating countries. The achievement of these targets was principally sought through improvements in immunization coverage, greater use of oral rehydration therapy, improvement in the health and nutritional status of mothers and children, and a reduction in the number of high-risk births (e.g. GOBI,a EPI,b and CCCDc). The specific strategies for achieving these broad sets of objectives were left to individual countries.
General optimism gave way to some pessimism when surveys conducted during the early 1980s were interpreted as showing that the initial rapid decline in mortality had begun to taper off in some countries, and that in a few there was strong evidence of the onset of stagnation or reversal (3-6). However, these conclusions were viewed with a degree of scepticism (7). While some levelling off in the rate of decline in mortality was to be expected at relatively low levels of mortality, the slowdown in these countries appeared to be occurring at relatively low levels of life expectancy. It was suggested that the rapid rate of decline observed earlier was not sustainable, given the slow rate of economic development, the impact of the AIDS epidemic, and the infusion of a very narrowly defined set of sophisticated technology-driven public health interventions (3, 4). It was argued further that the focus on vaccine-preventable diseases and oral rehydration therapy, as opposed to a more comprehensive, broadly based set of community health programmes, had led to substitution effects in morbidity and mortality. In other words, children were saved from measles and diarrhoea only to die from causes not covered by these interventions (8-10). Where substantial, substitution effects were likely to have a significant effect on the rate of decline in mortality.
Such concerns prompted research into the nature, pattern and determinants of mortality decline among children. Some workers sought an explanation for the phenomenon at the individual level by examining individual biological, social, behavioural, demographic and economic characteristics (11-13). Others set out to elucidate the mechanisms through which various determinants affected child mortality (14, 15). Attempts were also made to explain observed differences in terms of differentials in environmental, cultural and material living conditions of children or families (16-18).
Failure to identify the causal pathways linking child mortality and its proximate and distal determinants led to a further proliferation of child survival interventions. For instance, in 1990 the World Summit for Children laid down a target for the year 2000 of 70 deaths per 1000 live births among children under 5 years of age or of reductions of a third if these yielded lower mortality rates. This was to be achieved through improved birth weights and a reduction in the prevalence of malnutrition (19). In order to monitor progress in these matters it became necessary to have accurate and timely assessment and reporting of trends in infant and child mortality. Such reporting was also necessary if countries were to mobilize resources so as to meet all child health challenges promptly and efficiently.
There have been several attempts to assess mortality levels and trends among children under 5 years of age. They include:
a review of child mortality data for sub-Saharan Africa up to the mid-1980s (20);
a review of estimates and projections of the global under-five mortality rate for the period 1950-2025 (21);
a review of child mortality in the developing world (7);
a compilation of a child mortality database for developing countries (22);
a review of child mortality for the period 1960-90 (23);
comparative studies on infant and child mortality (24, 25);
a review of child mortality for the period 1960-96 (26).
These analyses yielded various estimates of deaths among children under 5 years of age. For instance, as part of the Global Burden of Disease Study, the number of deaths in this age category was estimated to be 12.8 million in 1990, all but 200 000 of them in developing regions (27). The earlier UNICEF review, covering the period 1960-90, yielded an estimate of 11 million childhood deaths in 83 developing countries during 1992 (23). The most recent review by UNICEF, covering the period 1960-96, produced an estimate of 10.9 million childhood deaths in 94 developing countries during 1996 (26). The numbers are not directly comparable but they suggest a substantial reduction in overall mortality among children under 5 years of age.
The continued focus on child survival by the global public health community, the increasing availability of data, and the apparent dramatic declines in child deaths suggest the need for a reappraisal of child mortality trends. The present paper aims to review the evidence and describe current levels and trends on the basis of data for both developing and developed countries. The specific aims are to examine, describe and document country-specific trends in under-five mortality rates in the 1990s and to identify those countries and regions where sustained improvement has occurred and those where there have been setbacks. In many respects our analysis updates previous work by UNICEF, the World Bank and the United Nations (21, 23, 26). It also adds to these studies by providing a consistent series of estimates of under-five mortality rates and by indicating historical trends for the period 1950-2000 in both developed and developing countries.
Data and their quality
The probability of dying between birth and the exact age of 5 years (the under-five mortality rate) is expressed per 1000 live births. It is widely recognized as the most appropriate indicator of the cumulative exposure to the risk of death during the first five years of life. It has a number of advantages over the infant mortality rate as a composite measure of health risks at young ages. In particular, the indirect demographic techniques for measuring the under-five mortality rate are more robust and less sensitive to assumptions about the underlying patterns of mortality than the infant mortality rate (28, 29). Moreover, the risk of death from several of the diseases that are principal causes of infant mortality remains high in the early years of childhood. It is thus also an appropriate outcome measure for assessing the impact of various intervention programmes aimed at improving child survival.
Data obtained from the complete registration of births and deaths provide the best possible basis for the direct estimation of child mortality. Unfortunately, such data are available for only 55-60 countries.d Where necessary, data have been corrected for underreporting. Many countries, especially in Africa but also in parts of Asia and Latin America, lack functioning vital registration systems. Estimates of childhood mortality in these countries are therefore largely based on cross-sectional surveys that collect complete birth histories from respondents. In a few cases, longitudinal demographic surveillance systems provide routine data. Such systems typically track and record all vital events in a well-defined sample of a national population. They are, however, very expensive to establish and maintain and often cover a non-representative sample, which makes their usefulness limited. In the absence of such sources of data, indirect estimates of child mortality levels and trends may be obtained from information on children ever-born and the proportion surviving by age of mother (30). This information has been widely collected in censuses and surveys.e
Given the data limitations, issues of data quality and comparability of estimates can be expected to hinder the international analysis of both trends and differentials in child mortality within and between major regions. In any country there are likely to be variations in estimates at a given time and across time, depending on the number of sources and on differences in methods of estimation. For many developing countries, data for the earlier periods, especially those from censuses, are often of relatively poor quality because of substantial errors of closure.f For these countries, data from the World Fertility Survey and the Demographic and Health Survey programmes are often the best available. Nonetheless, these sources may still be affected by a number of errors, the most important being omissions of births and deaths, misreporting of ages at death, and misreporting of maternal ages. The omission of vital events is probably the most serious error, occurring most frequently for children not living with their natural mothers, and for children who have died, especially those who died many years before the survey date. Since women of reproductive age are the basic sampling units in these surveys, their premature death excludes their children from the studies. Such children often have an elevated mortality risk and their exclusion tends to bias child mortality downward. This has a potentially serious effect in situations where HIV/AIDS accounts for a significant proportion of maternal deaths, and, as a consequence, efforts to assess the impact of AIDS on childhood mortality can be seriously impeded.
Another problem is the misreporting of birth dates, which may affect trends in mortality, especially if the misreporting varies according to whether a child is alive or not at the time of the survey. Age preference, typically for ages ending in 0 or 5, is a less serious but much commoner problem. A much more serious form of age misreporting consists of systematic transfer into adjoining ages.
Typically, the women selected for analysis from these censuses and surveys are currently married and under 49 years of age. Consequently, information about the mortality of children is limited to those born to progressively younger women, at the time of the births, as the reference period extends further into the past. Such age truncation could adversely affect and distort time trends, and is biased towards capturing mortality in the most recent times. Sampling errors also affect the estimates obtained.
The analysis reported here has been facilitated by compilations of data from the six major reviews carried out by UNICEF, the United Nations and the World Bank. It covers 171 of the 191 Member States of WHO. Populations below 100 000 have been excluded from the trend analysis although they have been included in the overall estimate of child mortality for 1999.
Methods
Child mortality estimation
The under-five mortality rate has been chosen as the indicator of child mortality because it provides the best means of capturing mortality risks during the most vulnerable years of childhood. The estimates for developing countries are, in general, drawn from previous studies (7, 20-26), and also from country reports of the World Fertility Survey and the Demographic and Health Survey programmes. The methods used for estimation depend on the type of data available. Where mortality data are derived from complete birth histories, direct techniques based on the life table approach are applied. In this case, specific probabilities of dying are computed from reported deaths and the numbers of children of a particular age exposed to the risk of dying during a specified period, e.g. 5, 10 or 15 years before a survey. Alternatively, the synthetic cohort method may be used to derive risks of child death. The procedure for calculating synthetic cohort probabilities of dying is based on an approach developed by Somoza (31) and modified by Rutstein (32). Probabilities of dying are built up from probabilities calculated for specific age intervals: less than 1 month, 3-5 months, 6-11 months, 12-23 months, 24-35 months, 36-47 months and 48-59 months. This method has been used for almost all the direct estimates based on birth history data which are presented in the present study. For countries lacking appropriate information, estimates of under-five mortality rates were computed from published information (33). For countries with consistent series of vital registration data, mostly developed countries, the estimates of mortality in this age group are taken from the 1999 WHO life tables.
Indirect techniques developed by Brass (30) and modified by Sullivan (34) and Trussell (35) can be used to obtain childhood survivorship probabilities when the only data available are information on children ever born and the proportion of them surviving to respondent mothers of reproductive age. The procedure allows the conversion of the proportion of children who have died, reported by women of each age group, to probabilities of dying between birth and exact ages of childhood (xq0),g after adjusting for the influence of the different age patterns of fertility. The application of this method requires the choice of a family of the Coale & Demeny model life tables (29); hence the estimates of xq0 obtained can be significantly affected by the choice of model. As indicated earlier, estimates of the under-five mortality rate are less sensitive to the model chosen than are estimates of the infant mortality rate. Coale & Trussell have developed a procedure for estimating the time location of each of the standard estimates (36).
In addition to knowing the risk of death among children under 5 years of age it is often useful to know the absolute number of deaths in this age group in individual countries and globally. The number of child deaths for any given calendar year is obtained by multiplying the risk of death (5q0) for that year by the corresponding estimated annual number of live births. This approach, however, tends to overestimate deaths slightly in a growing, non-stationary population, especially if the number of births increases each year (26). In estimating the number of deaths among children under 5 years of age in 1999 we used the WHO life table estimates of the central death rate in conjunction with the estimated population aged 0-4 complete years derived from the UNs 1998 demographic assessment (33).
Trend analysis
When different data sources and methods of estimation are used, estimates of child mortality for a particular reference period tend to differ substantially because of both sampling and non-sampling errors. They may also be dependent on the choice of methods as well as on violations of methodological assumptions. The magnitude of the differences increases with the degree of variation in the data sources and methods of analysis. This makes it somewhat difficult to analyse trends. Various techniques have been reported for obtaining a series of consistent estimates of child mortality over time from such data. They range from hand smoothing to more sophisticated regression techniques. For the present analysis we chose an approach that included both a systematic evaluation of data quality and plausibility followed by simple averaging over all plausible data points for a given year or set of years.
A data point is considered valid for inclusion if based on vital registration, time series from maternity history, or a time series based on children ever born and those surviving. For the latter it was decided to exclude all indirect estimates based on information from mothers in the two youngest age ranges, 15-19 years and 20-24 years. When working with data classified by age it has generally been found that reports of women under 20 years are not a good basis for estimation (28). Furthermore, children born to younger mothers have a higher risk of dying than those born to women aged 25-34 and thus tend to produce an upward bias in estimates of the under-five mortality rate. In cases where two or more sets of estimates are available for an overlapping period, a set is excluded if it diverges completely from other mutually consistent series from more than one independent source. Mortality estimates for each five-year interval were obtained by unweighted averaging of all estimates resulting from the first two steps. This had the disadvantage that trends were based on unequal amounts of information at different points in time. Where data were completely lacking, estimates based on UN data (33) were graphically compared with estimates produced by Hill et al. (26). Where the two estimates differed substantially a compromise was selected so as to maintain consistency with the computed trend line. Table 1 gives estimates of under-five mortality rate for five-year periods from 1955-59 to 1995-99 and estimates for 1999 together with the corresponding uncertainty intervals.
Results
Table 2 shows the distribution of deaths among children aged 0-4 years in 1999 by WHO regions and within each region according to five mortality strata.h These strata were developed to distinguish countries on the basis of their relative levels of child mortality (5q0) and adult mortality (45q15).i Based on this analysis, an estimated 10.5 million children aged 0-4 years died in 1999, about 2.2 million or 17.5% fewer than a decade earlier (27). Of these deaths, 99.3% occurred in countries other than those characterized by very low child mortality and very low adult mortality (stratum A), and even when the countries of Eastern Europe with comparatively high adult mortality were included the share of child deaths globally in what can be termed developed countries barely exceeded 2%. An estimated 3.8 million child deaths occurred in Africa alone, 36% of the world total, and there were 2.5 million in India and 750 000 in China.
Globally, a newborn baby in 1999 had about a 6.7% chance of dying before reaching the age of 5 years (Table 1, Fig. 1); the corresponding values for 1990, 1980 and 1950 were 9.6% (27, 38), 12% (38) and 25% (38). Differences in the risk of child death across broad regions of the world and different mortality strata are summarized in Fig. 2. On average about 15% of children born in Africa are expected to die before reaching their fifth birthday. This compares with 3-8% in many other parts of the developing world and less than 2% in Europe. The disparity is even wider when countries are grouped according to the five mortality strata. Countries in the very high mortality strata (D, E) have approximately 15 times the mortality risk of those in the very low mortality stratum (A).
Global trend
Marked reductions in childhood deaths occurred during the second half of the 20th century. Globally, the median under-five mortality rate declined from around 150 per 1000 live births in the 1950s to around 40 per 1000 in the 1990s (Fig. 3). However, the range of values remained relatively unchanged over the same period (Table 3). For instance, in 1955-59 the range was between 21 per 1000 live births in Sweden and 381 per 1000 in Sierra Leone. The range for 1995-99 was between 5 per 1000 in Finland, Luxembourg, Norway, and Sweden and 334 per 1000 in Niger. When one excludes outliers, however, a marked reduction in the degree of heterogeneity is observed.
The rate of decline in childhood mortality is a key issue of concern to development agencies and the global public health community. With economic reversals, political instability and the advent of the AIDS epidemic, progress in reducing child mortality might well be expected to have slowed down dramatically over the last few decades. The rate of the decline globally, as measured by the trend in the percentage decline between succeeding quinquennia, increased dramatically between 1960 and 1979, reaching a peak of 12% in the second half of the 1970s. It remained at this level throughout the 1980s. A substantial slowing of the rate of decline in child mortality marked the period following the mid-1980s (Table 3, Fig. 3).
The regional variations around this broadly favourable global trend are summarized in Table 4 and Fig. 4. With reference to both broad regional groupings and mortality strata it can be seen that there were very different paths of decline in child mortality over the last four decades. For example, whereas the level of under-five mortality in stratum B was similar in 1995-99, ranging from 28 to 40 deaths per 1000 live births, except for the Western Pacific Region where it was 64 per 1000, the extent of decline since 1955-59 was greatest in the Eastern Mediterranean countries and least in the Americas and the European Region. A similar pattern is observed for the countries classified in stratum D in each region, suggesting that, despite the wide disparity in current levels of child mortality worldwide, the differences were even greater several decades ago. Fig. 4 shows the variation in the mortality rate by region for the periods 1970-74, 1980-84 and 1990-94. The main findings were as follows.
All regions experienced sharp reductions in the median under-five mortality rate between 1970 and 1994.
Throughout the period, Europe had the lowest under-five mortality rate, followed by the Americas and the Western Pacific, the Eastern Mediterranean, South-East Asia and Africa in that order.
By the early 1970s:
- the median under-five mortality rate in Europe was below 50 per 1000 live births;
- the Western Pacific and the Americas had median mortality rates below 100 per 1000 births;
- the estimate for Africa was over 200 per 1000 live births;
- the Eastern Mediterranean and South-East Asia had median mortality levels of approximately 160 per 1000 live births.
The region that experienced the steepest sustained decline was the Eastern Mediterranean, where the median mortality rate dropped from about 170 per 1000 in 1970-74 to below 50 per 1000 in 1990-94.
The region with the poorest performance was Africa.
Three groups of countries were thus distinguished. In the first group, mortality levels were already relatively very low and further dramatic declines were therefore unlikely. This group is made up of countries in the European Region, the Region of the Americas, and the Western Pacific Region. The second group comprises countries in the Eastern Mediterranean Region, where the under-five mortality rates dropped rapidly from very high initial values to very low levels, thereby limiting the possibility of further rapid reductions. The third group consists of the countries in the African Region and the South-East Asia Region, where the initial mortality levels were high, the magnitude of the declines was very small, and there was a major slowing in mortality decline. These are also the regions with severe economic dislocations. Provided that the HIV/AIDS situation does not worsen, it is conceivable that improvements in their political and socioeconomic circumstances could result in further substantial reductions in mortality.
Trends within regions and relative performances
Countries were grouped according to the relative change between their most recent average level of mortality and the level two decades earlier. Countries in each region were categorized as follows according to the decline in under-five mortality rate achieved during the period between 1980 and 1999:
increased mortality;
declines of 0-4%, 5-9%, 10-14% and so on at 5% intervals to the highest category of over 50% decline.
The results are shown in Table 5.
African Region. In Africa the mean mortality rate dropped from about 264 deaths per 1000 live births in 1955-1959 to about 152 per 1000 in 1995-1999, a decline of about 42% (Table 6). This compares with a decline of between 60% and 72% in the other regions over the same period. Table 6 also shows the sluggish nature of the rate of the decline, from 4% to a peak of 9%. Furthermore, the degree of heterogeneity within the region, as indicated by the variance, has remained relatively unchanged over the last four decades (Table 6). Thus for 1955-59 the range of mortality rates was from113 per 1000 in Mauritius to 381 per 1000 in Sierra Leone. For 1995- 99 the range became even wider, from 21 per 1000 in Mauritius to 334 per 1000 in Niger.
Africa had the worst performance, more than half the countries achieved declines of less than 20%. Only Algeria, Gabon, and the Gambia achieved over 40% reduction during the last two decades (Table 5). Countries with only a very modest decline (0-10%) included Burundi, Lesotho, Madagascar, Mauritania, Nigeria, Sierra Leone, and the United Republic of Tanzania. Some countries experienced an increase in child mortality, probably because of the HIV/AIDS epidemic: they included Botswana (16%), Namibia (4%), Niger (1%), Zambia (9%), and Zimbabwe (15%). Increases in child mortality of less than 5% over the 20-year period should probably be interpreted with caution, since sampling errors and data of poor quality might have led to spurious findings.
South-East Asia Region. This region has experienced a moderately sustained reduction in mortality among children under 5 years of age during the last four decades, from about 222 deaths per 1000 live births to about 90 per 1000, and there has been some narrowing of differentials (Table 6). All the countries in the region except the Democratic Republic of Korea and Myanmar experienced reductions of 20% or more over the last two decades (Table 5).
Western Pacific Region. Mean mortality dropped by almost 70% from about 154 per 1000 live births in 1955-59 to about 48 in 1995-99 (Table 6). Only Mongolia and Papua New Guinea failed to achieve major reductions; indeed, Papua New Guinea experienced an increase (Table 5).
Eastern Mediterranean Region. Overall, the mean under-five mortality rate dropped sharply from about 238 per 1000 live births in 1955-59 to about 71 per 1000 in 1995-99 (Table 6). There was, however, substantial variation. In 1955-59 the rate for Cyprus was 57 per 1000 while that for Afghanistan was 368 per 1000. In 1995-99 the corresponding rates were 9 per 1000 (Cyprus) and 271 per 1000 (Afghanistan), and there was relatively little change in the ranking of countries between these extremes. The countries that contributed to this sustained decline included the Islamic Republic of Iran, Jordan, Oman, Qatar, Saudi Arabia, Tunisia and the United Arab Emirates (Table 5). Much more modest reductions, in the range 5-15%, occurred in Afghanistan, Djibouti, and Somalia.
Region of the Americas. Between 1955 and 1999 the under-five mortality rate fell from about 140 to 38 deaths per 1000 live births, i.e by about 70% (Table 6). This was accompanied by a substantial reduction in heterogeneity (Table 6). The rate of decline was initially slow, increased sharply during the 1970s, and slowed down during the 1980s (Fig. 5). This region had the best overall performance, all countries achieved reductions of at least 20% over the last two decades. Chile, Ecuador, El Salvador, Guatemala, Honduras, Nicaragua, Mexico, and Trinidad and Tobago experienced sustained declines of more than 50% over the last two decades.
European Region. The under-five mortality rate had already dropped substantially in Europe by the mid-1950s, largely as a result of improved living conditions and the availability of antibiotics. In 1955- 59 the mean mortality rate was about 63 per 1000 live births. It dropped to 19 per 1000 over the next 40 years. Intercountry differentials narrowed considerably during the same period. In 1955-59 the range was from 21 per 1000 in Sweden to 210 per 1000 in Turkey. By the end of 1999 the lowest value was 5 per 1000 in Finland, Luxembourg, Norway, and Sweden, and the highest was 82 per 1000 in Turkmenistan. Over the last two decades there were substantial reductions in all countries except Albania, Armenia, Kyrgyzstan, and Latvia.
Discussion
Given the substantial focus of international public health programmes on improving child survival, it is important that periodic reviews of progress be undertaken to identify needs and opportunities for priority action. With improvements in vital registration, demographic surveillance and survey programmes, reasonably reliable estimates of child mortality are now available for many countries.
Despite setbacks to international development efforts and the emergence of major new threats to health, such as the HIV/AIDS epidemic, the overall under-five mortality rate declined by almost 30%j during the 1990s, building on a similar decline during the 1980s. Globally, the number of child deaths declined from over 13 million in 1980 to about 10.5 million in 1999. These are, by any standard, remarkable public health achievements. There are, however, strong suggestions that the rate of decline may have slowed down somewhat. This arises partly because of the already low child mortality levels in the European Region, the Region of the Americas, the Western Pacific Region and the Eastern Mediterranean Region, and also because of an apparent lack of substantial progress in the African Region and the South-East Asia Region. In this connection the following questions should be asked. Are the declines real? If so, what has caused them? What are the prospects for further reductions in child mortality?
With regard to the measurement of childhood mortality trends, the wealth of data from the Demographic and Health Survey programme relating to almost 60 countries has greatly increased confidence in the comparative assessment of child mortality over the last two decades or so. In other countries, routine demographic surveillance systems, such as the Sample Registration System in India, now provide useful, if not very reliable, data on child mortality. Vital registration systems have improved in many developing countries, and even where still deficient they can provide valuable information to complement other sources and inform judgements about likely levels and trends in mortality. Thus while many of the estimates reported here, particularly for the less developed regions, are uncertain, we are reasonably confident that they describe recent trends fairly reliably.
The 1999 estimates are based on life tables produced principally for work on the measurement of health systems performance (37). Because a complete life table is a complicated non-linear function of uncertain parameters, we have used Monte Carlo simulation techniques to develop numerical estimates of the ranges of uncertainty around the life tables. A detailed discussion of the sources of uncertainty and methods for uncertainty analysis for life tables may be found in a separate publication (40). The estimated bounds around the 1999 estimates of under-five mortality rate are shown in Table 1. Globally, the number of child deaths in 1999 almost certainly lies between 9 and 12 million; the greatest uncertainty occurs in respect of Africa and parts of Asia. The global and regional estimates presented here are probably a reasonable reflection of true child mortality levels, and are certainly consistent with previous estimates by the UN and UNICEF.
What are the prospects for further declines in child mortality? It is important to recognize that reversals in public health are possible and can substantially alter long-term declines in mortality. Two examples are the dramatic increase in adult mortality in the Russian Federation and several other Eastern European countries in the late 1980s and early 1990s, and the emergence of HIV/AIDS as a leading cause of adult mortality. To the extent that children are likely to be more vulnerable to such developments, continued progress is by no means assured. Certainly, the full impact of the HIV/AIDS epidemic on child mortality has not yet been felt. The sluggish pace of decline and the increases observed in the under-five mortality rate in some African countries suggest the possibility of a major slowdown, if not a reversal, of gains in child survival in countries where the seroprevalence of HIV is high. As the epidemic continues to unfold in Africa, particularly southern Africa, further reductions in child mortality are expected to become increasingly unlikely until substantial progress in controlling the spread of HIV is achieved. Moreover, if rates of infection in Asia continue to increase, progress in reducing child mortality will be hindered, as was apparent in Thailand during the 1990s.
Conclusion
Good vital registration offers the best means of monitoring the progress of countries in reducing premature deaths, particularly child deaths. In many developing countries, however, progress in this field can be expected to remain modest because of the cost of maintaining and improving systems for the registration of births and deaths. Nevertheless, the extensive population survey programmes that were established over the last few decades, combined with sample registration and censuses, yield valuable information on child mortality. These data sources, appropriately checked for implausible observations, are now sufficiently common in developing countries to permit levels and trends in child mortality to be estimated with some confidence.
Using these data we have estimated that some 10.5 million children aged 0-4 years died in 1999, about 2 million fewer than at the start of the decade. Reasonable uncertainty bounds on country-specific estimates suggest that the global total was unlikely to be below 9 million or above 12 million. Of these deaths, 98-99% occurred in developing countries. Globally, the mean risk of death before the age of 5 years for a newborn baby is about 6.7%. In Africa, approximately 15% of newborn babies die before their fifth birthday; this compares with 6-10% of newborn babies in South-East Asia and of 4% or fewer in many parts of Latin America, the Western Pacific Region and the Eastern Mediterranean Region.
Our analysis also suggests that the steady progress in reducing child mortality over the last five decades has continued throughout the 1990s, despite the advent of HIV/AIDS, continued civil strife in some countries and a lack of economic growth in others. Although the rate of decline fell in all regions except the Western Pacific Region during the 1990s, there is no widespread evidence of rising child mortality rates. The notable exceptions are countries in southern Africa, where there is an extremely high prevalence of HIV, and a few in Asia beset by difficult economic circumstances. Nevertheless, the slowdown in Africa and South-East Asia is particularly worrisome because it is occurring at very high levels of mortality and in countries experiencing severe economic dislocation.
The causes of the continued decline need to be better understood if public health programmes are to be more appropriately targeted to ensure that it is maintained. Because of the paucity of reliable data on causes of death in developing countries it is extremely difficult to determine whether the gains can be attributed to the success of specific intervention programmes, changes in childbearing patterns, general social and economic development or, as is likely, to some combination of these and perhaps other factors. Failure to specify the causal pathways linking child mortality and its proximate and distal determinants has led to a further proliferation of child survival initiatives with loosely defined strategies for achieving certain child mortality targets. It is desirable to concentrate more on elucidating the mechanisms through which various interventions operate to lower or raise the risk of child death, and somewhat less on simply identifying the determinants of child survival. It is also desirable to investigate further the reasons for the success or failure of specific interventions.
In many countries the rapid reduction of child mortality remains a priority aim in the field of public health. Investing in research and surveillance systems to improve knowledge about the causes of declines in child mortality can be expected to ensure that intervention programmes are more appropriately targeted and thus more likely to achieve this aim.
Continued vigilance to protect the gains in child survival prospects is clearly essential. Further gains can be expected as knowledge is translated into effective programmes of action. Provided that major threats to health such as HIV/AIDS and malaria do not increase dramatically over the next few decades, particularly in Africa, current trends suggest that that child mortality might be halved by about 2030. Success is likely to be greatest in those countries where political stability and commitment to service delivery and socioeconomic development are strongest. 
Résumé
Baisse de la mortalité infanto-juvénile: nouvelle évaluation
Les auteurs de larticle passent en revue les niveaux actuels et les tendances de la mortalité infanto-juvénile à la lumière de toutes les données disponibles pour les pays développés comme pour les pays en développement. Ils se sont particulièrement attachés à examiner, décrire et documenter dans chaque pays les tendances des taux de mortalité chez les moins de cinq ans au cours des années 90, et à identifier les pays et les Régions ayant enregistré une amélioration durable, ainsi que ceux où les reculs sont manifestes. A de nombreux égards, cette analyse réactualise celles qui ont déjà été faites par lUNICEF, la Banque mondiale et lOrganisation des Nations Unies. Elle les complète par ailleurs en fournissant des séries cohérentes destimations de la mortalité chez les moins de cinq ans et en indiquant les tendances historiques pour une période allant de 1950 à 2000, tant pour les pays développés que pour les pays en développement.
Selon les estimations, 10,5 millions denfants de 0 à 4 ans sont morts en 1999, soit environ 2,2 millions ou 17,5% de moins que dix ans plus tôt. Sur ce nombre, on compte 3,8 millions denfants (36 %) en Afrique, 2,5 millions en Inde et 750 000 en Chine. A léchelle mondiale, le risque quun enfant décède a diminué de manière marquante en un demi-siècle. En 1999, un nouveau-né courait un risque denviron 6,7% de mourir avant lâge de 5 ans, contre 9,6% en 1990, 12% en 1980 et 25% en 1950. Il existe néanmoins des différences importantes au niveau régional: environ 15% des nouveau-nés africains mourront avant leur cinquième anniversaire alors quils ne sont que 3 à 8% dans bien dautres régions du monde en développement et moins de 2% en Europe. Ces disparités sont encore plus grandes lorsquon regroupe les pays selon les cinq strates établies par lOMS pour la mortalité. Dans les pays correspondant à la strate où la mortalité est la plus élevée, le risque de mortalité est environ 15 fois plus grand que dans ceux qui se trouvent dans la strate la plus basse.
La vitesse à laquelle la mortalité infanto-juvénile diminue est un point essentiel pour les organismes de développement et la communauté mondiale soccupant de la santé publique. Lanalyse indique une diminution rapide de la mortalité chez les moins de cinq ans avec un pic de vitesse de 1975 à 1980, puis une baisse constante au cours des années 80. Bien que lon ait observé un ralentissement dans toutes les Régions à lexception de celle du Pacifique occidental au cours des années 90, rien ne permet véritablement de conclure à une augmentation des taux de mortalité infanto-juvénile, à lexception notable de lAfrique australe, qui connaît une prévalence extrêmement élevée du VIH, et de quelques pays asiatiques en proie à des difficultés économiques. Le ralentissement de la baisse est particulièrement pré-occupant en Afrique et en Asie du Sud-Est parce quil survient alors que les taux de mortalité sont relativement élevés et que les pays font face à de graves bouleversements économiques. Avec lépidémie de VIH/SIDA qui continue de sétendre en Afrique, notamment en Afrique australe, et dans certaines régions dAsie, il devient de plus en plus improbable que de nouvelles diminutions de la mortalité infanto-juvénile soient enregistrées tant que des progrès importants nauront pas été accomplis pour maîtriser la propagation du virus.
Resumen
Reevaluación del descenso de la mortalidad infantil
Los autores examinan los actuales niveles y tendencias de la mortalidad de niños pequeños a la luz de todos los datos disponibles tanto para los países en desarrollo como para los países desarrollados. Los objetivos específicos son examinar, describir y documentar las tendencias de países concretos en lo referente a las tasas de mortalidad entre los menores de cinco años en la década de 1990, e identificar los países y Regiones donde ha tenido lugar una mejora sostenida y aquellos donde ha habido claros reveses. En muchos aspectos nuestro análisis actualiza otros anteriores llevados a cabo por el UNICEF, el Banco Mundial y las Naciones Unidas. Además complementa esos estudios proporcionando una serie coherente de estimaciones de la mortalidad entre los menores de cinco años y revelando tendencias históricas del periodo 1950-2000 tanto para los países desarrollados como para los países en desarrollo.
Calculamos que en 1999 murieron 10,5 millones de niños de 0-4 años de edad, lo que representa unos 2,2 millones, o 17,5%, menos que diez años antes. De esa cifra, 3,8 millones (36%) corresponden a África, 2,5 millones a la India, y 750 000 a China. A nivel mundial, el riesgo de defunción de los niños ha descendido marcadamente durante el último medio siglo. En 1999, un recién nacido tenía una probabilidad de un 6,7% de morir antes de alcanzar los 5 años, mientras que las cifras correspondientes a 1990, 1980 y 1950 son del 9,6%, 12% y 25%, respectivamente. Existen sin embargo marcadas diferencias regionales: como promedio, puede estimarse que aproximadamente un 15% de los recién nacidos en África morirán antes de cumplir los cinco años, lo que contrasta con un 3% - 8% en muchas otras partes del mundo en desarrollo y con menos del 2% en Europa. Las disparidades son aún mayores cuando se procede a agrupar los países de acuerdo con los cinco estratos de mortalidad de la OMS. En los países situados en el estrato de mortalidad muy alta el riesgo de defunción es aproximadamente 15 veces mayor que en el de mortalidad muy baja.
Un dato de sumo interés para los organismos que actúan en pro del desarrollo y para la comunidad implicada en la salud pública mundial es la tasa de disminución de la mortalidad infantil. El análisis efectuado muestra una rápida disminución de la mortalidad entre los menores de cinco años, con tasas que alcanzaron su valor máximo en la segunda mitad de los años setenta y permanecieron inalteradas durante los años ochenta. Aunque la tasa de disminución de la mortalidad infantil se aminoró en todas las Regiones, salvo en el Pacífico Occidental, durante los años noventa, no hay muchos datos que apunten a un aumento de las tasas de mortalidad infantil. Excepciones destacables son los países del sur de África, donde la prevalencia del VIH es sumamente alta, y unos pocos países asiáticos acosados por unas circunstancias económicas difíciles. La atenuación de la disminución resulta especialmente preocupante en África y en Asia Sudoriental, porque se está produciendo a niveles relativamentes altos de mortalidad y en países que atraviesan dificultades económicas graves. A medida que avanza la epidemia de VIH/SIDA en África, sobre todo en su parte meridional, y en determinadas zonas de Asia, cualquier reducción adicional de la mortalidad infantil resulta crecientemente improbable, a la espera de que se logren progresos sustanciales en la lucha contra la propagación del VIH.
References
1. Stolnitz GJ. Recent mortality trends in Latin America, Asia and Africa: review and reinterpretation. Population Studies, 1965, 19: 117-138.
2. Caldwell JC. Routes to low mortality in poor countries. Population and Development Review, 1986, 12: 171-220.
3. Gwatkin DR. Indications of change in developing country mortality trends: the end of an era?. Population and Development Review, 1980, 6: 615-644.
4. Ruzicka LT, Hansluwka H. Mortality transition in South and East Asia: technology confronts poverty. Population and Development Review, 1982, 8: 567-588.
5. Palloni A. Mortality in Latin America: emerging patterns. Population and Development Review, 1981, 7: 623-649.
6. Mortality and health policy. Highlights of issues in the context of the World Population Plan of Action. New York, United Nations Population Division, 1983.
7. Hill K, Pebley AR. Child mortality in the developing world. Population and Development Review 1989, 15: 657-687.
8. Foster SO. Immunizable and respiratory diseases and child mortality. In: Mosley WH, Chen L, eds. Child survival: strategies for research. Population and Development Review, 1984, 10 ( suppl.): 119-14 0.
9. Kasongo Project Team. Influence of measles vaccination on survival pattern of 7-35 month old children in Kasongo, Zaire. Lancet, 1981, 4: 764-767.
10. Hendrickse RG. Problems of future measles vaccinations in developing countries. Transactions of the Royal Society of Tropical Medicine and Hygiene, 1975, 69: 31-34.
11. Trussell J, Preston SH. Estimating the covariates of childhood mortality from retrospective reports of mothers. Health Policy and Education (Amsterdam), 1982, 3: 1-36.
12. Socio-economic differentials in child mortality in developing countries. New York, United Nations, 1985 ( Sales No. E.85.XIII.7).
13. Child mortality in developing countries: socio-economic differentials, trends and implications. New York. United Nations, 1991 ( Department of International Economic and Social Affairs, ST/ESA/SER.A/123).
14. Mosley WH, Chen L. An analytic framework for the study of child survival in developing countries. In: Mosley WH, Chen L, eds. Child survival strategies for research. Population and Development Review, 1984, 10 ( suppl.): 24-45.
15. Ahmad OB, Eberstein L, Sly DF. Proximate determinants of child mortality in Liberia. Journal of Biosocial Science, 1991, 23: 313-326.
16. Omorodion FI. The socio-cultural context of health behavior among the Esan communities, Edo State, Nigeria. Health Transition Review, 1993, 3 ( 2): 125-136.
17. Castle SE. Intra-household differentials in womens status: household functionsand focus as determinants of childrens illness management and care in rural Mali. Health Transition Review, 1993, 3 ( 2): 137-158.
18. Hobcraft J. Womens education, childrens welfare and child survival: a review of the evidence. Health Transition Review, 1993, 3 ( 2): 159-176.
19. United Nations Childrens Fund. The state of the worlds children 1991. Oxford. Oxford University Press, 1991.
20. Hill A. Child mortality in sub-Saharan Africa. Yaounde, Cameroon, 1987 ( Paper presented to the IUSSP Seminar on Mortality and Society in sub-Saharan Africa).
21. Mortality of children under age 5: world estimates and projections, 1950-2025. New York, United Nations, 1988 ( Department of International Economic and Social Affairs, Population Studies No. 105).
22. Child mortality since the 1960s: a database for developing countries. New York, United Nations, 1992 (Department of Economic and Social Development, ST/ESA/SER.A/128).
23. Hill K, Yazbeck A. Trends in child mortality, 1960-90: estimates for 84 developing countries. In: World development report 1993: investing in health. New York, Oxford University Press for The World Bank, 1993.
24. Sullivan JM, Rutstein SO, Bicego GT. Infant and child mortality. Calverton, MD, Macro International Inc, 1994. (Demographic and Health Surveys, Comparative Studies No. 15).
25. Bicego G, Ahmad OB. Infant and child mortality. Calverton, MD, Macro International Inc, 1996 ( Demographic and Health Surveys, Comparative Studies No. 20).
26. Hill K et al. Trends in child mortality in the developing world: 1960-1996. New York, UNICEF, 1999.
27. Murray CJL, Lopez AD. Global burden of disease. In: Murray CJL, Lopez AD, eds. The Global Burden of Disease and Injury Series. Cambridge, Harvard University Press, 1996: 340-342.
28. Hill K. Evaluation of indirect methods for estimating mortality. In: Vallin J, Pollard JH, Heligman L, eds. Methodologies for the collection and analysis of mortality data. Liège, Ordina Editions, 1984: 145-176 (proceedings of a seminar of the International Union for the Scientific Study of Population, Dakar, Senegal, 7-10 July 1981).
29. Coale AJ, Demeny P. Regional model life tables and stable populations. Princeton, Princeton University Press, 1966.
30. Brass W. Uses of census and survey data for estimation of vital rates. Addis Ababa, United Nations Economic Commission for Africa, 1964 ( African Seminar on Statistics).
31. Somoza JL. Illustrative analysis: infant and child mortality in Colombia. WFS Scientific Reports, 1980, 10.
32. Rutstein SO. Infant and child mortality levels, trends and demographic differentials. The Hague, International Statistical Institute, 1984 ( Comparative Studies, No. 43).
33. World population prospects: the 1998 revision. Vol. I & II. New York, United Nations, 1999 ( Economic and Social Affairs, Population Division, ST/ESA/SER.A/180).
34. Sullivan JM. Models for the estimation of the probability of dying between birth and exact ages of early childhood. Population Studies; 1972; 26 ( 1): 79-97.
35. Trussell TJ. A re-estimation of the multiplying factors for the Brass techniques for determining childhood survivorship rates. Population Studies, 1975, 29 ( 1): 97-107.
36. Coale AJ, Trussell TJ. Estimating the time to which Brass estimates apply. Population Bulletin of the United Nations, 1977, 10: 87-91 ( United Nations Publication, Sales No. E.78.X111.6).
37. The world health report 2000 Health systems: improving performance. Geneva, World Health Organization, 2000.
38. World development report 1993: investing in health. New York, Oxford University Press for The World Bank, 1993.
39. Salomon J, Murray CJL. Methods for life expectancy and disability-adjusted life expectancy uncertainty analysis. Geneva, World Health Organization, 2000 ( Global Programme on Evidence for Health Policy Discussion Paper No. 10).
1 Demographer, 2 Coordinator, 3 Statistician, Burden of Disease Team, Global Programme on Evidence for Health Policy, World Health Organization, 1211 Geneva 27, Switzerland. Correspondence should be addressed to Dr Ahmad.
a Growth Monitoring, Oral Rehydration Therapy, Breast-feeding and Immunization (UNICEF).
b Expanded Programme on Immunization (WHO).
c Combating Childhood Communicable Diseases (United States Centers for Disease Control and Prevention).
d A list of countries with functioning vital registration systems can be obtained from the authors.
e A list of various sources of non-vital registration data included in the current study can be obtained from the authors.
f Also called residual errors, referring to the difference between the expected and the enumerated population.
g Implies the probability of dying between birth and exact ages of childhood, e.g. 1q0 is the probability of dying between birth and exact age 1, 2q0 is the probability of dying between birth and exact age 2.
h This study was undertaken as part of work on estimating the WHO life tables for 1999, which was needed for the assessment of health systems performance in The World Health Report 2000 (37). The life tables and the estimates of child mortality therefore had to be prepared for WHO regions, and that is how they are reported here. The composition of each WHO region and mortality subregion is given elsewhere (37).
i 45q15 is read as 45 Q 15, corresponding to the adult mortality rate or, more technically, to the probability of dying between the ages of 15 and 60 years.
j WHO estimate of 67 deaths per 1000 live births in 1999; World Bank estimate of 96 deaths per 1000 live births in 1990 (World Bank, 1993).
Ref. No. 00-0792