NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Jamison DT, Feachem RG, Makgoba MW, et al., editors. Disease and Mortality in Sub-Saharan Africa. 2nd edition. Washington (DC): The International Bank for Reconstruction and Development / The World Bank; 2006.

Cover of Disease and Mortality in Sub-Saharan Africa

Disease and Mortality in Sub-Saharan Africa. 2nd edition.

Show details

Chapter 5Causes of Death

, , and .

Consistent estimates of cause-specific mortality are essential for understanding the overall epidemiological profile of disease in a population. The principal data source for these estimates is civil registration systems. Adequately functioning systems that produce statistics on causes of death on a regular basis exist in only about one-third of all countries of the world (Lopez et al. 2002). In Sub-Saharan Africa, very little information has been available on cause-specific mortality, let alone data from civil registration systems, as described in the previous edition of this book (Feachem and Jamison 1991). Estimates at that time were derived largely from independent disease-specific epidemiological studies and were not examined within the context of an overall demographic "envelope" of mortality, as is required to ensure that claims about causes of death are not exaggerated.

Over the past decade, much progress has been made in the collection of mortality statistics from a wide array of sources. These include data from previously existing sources that were uncovered during a systematic search, as well as data from new data collection ventures that were established to fill these data gaps. Although we are still a long way from having satisfactory empirical data that can be directly used to derive national and regional cause-specific mortality estimates, the expansion in available data suggests that estimates of causes of death can now be made with somewhat greater confidence. The absence of complete vital registration data in virtually all countries of the region nonetheless means that we need to rely on epidemiological research and demographic surveillance to generate model-based estimates of deaths by cause.

Such an estimation process is complex and involves two stages. First, a demographic estimate of overall mortality by age and sex is required. Second, a cause-specific mortality structure is fitted to this estimate. Many assumptions are required, and an attempt has been made here to delineate these clearly, so that they can be kept in mind when interpreting the results.

Data Sources

The previous edition of this book provided information on cause-specific mortality in the form of a set of estimates for the entire region, disaggregated over seven broad disease categories, and in six age groups. These estimates were essentially derived from linear models relating mortality levels to broad causes of death (Preston 1976), with very little attempt to incorporate other epidemiological information. The authors stated that the large proportion of deaths lumped under other causes reflected the weakness of the estimates, and suggested great caution in interpreting or using them (Feachem, Jamison, and Bos 1991).

Building on previous experience, researchers have made much more comprehensive attempts over the past decade to collect and analyze information in order to estimate the cause-specific mortality structure for Sub-Saharan Africa as part of the Global Burden of Disease (GBD) Study (Mathers, Stein et al. 2002; Murray and Lopez 1996). Discussed here are the results of two major data collection efforts that were undertaken to estimate the burden of premature mortality in Africa. The first effort involved a search for, and analysis of, epidemiological literature on causes of death in the region. The second effort involved a similar exercise using data from national vital records.

Review of Epidemiological Literature

Much of the earlier attempts to estimate causes of death in the region drew on published and unpublished reports based on epidemiological studies. Adetunji, Murray, and Evans (1996) conducted a major review of epidemiological studies to identify usable cause-specific mortality data. They identified a total of 48 research studies from 16 countries that met specific criteria related to data definition and data quality issues for inclusion in the analysis. The main criteria for inclusion was that the study should have reported the relative magnitude of all causes of death and not focus on a single or a few causes without mention of other causes within the sample. Other criteria refer to methods used to derive the cause of death, age groups reported, site, and period of the study. Reports from South Africa were excluded from this analysis, since the data overlapped with that available from vital records. About 80 percent of the reports were based on information from hospital records, verbal autopsy interviews, or both. A majority of the studies (again 80 percent) focused on causes of maternal and child mortality, whereas causes of adult male deaths were relatively neglected. The authors presented results in the form of cause composition of mortality for three distinct cause groups: communicable diseases and maternal, perinatal, and nutritional conditions (group 1); noncommunicable diseases (group 2); and injuries (group 3). These cause groups were the same as those used in the GBD Study. The results of their review of deaths of children younger than five years are summarized in table 5.1.

Table 5.1. Proportion of Deaths Due to Diseases and Injuries in Children under Five Years, by Source of Data.

Table 5.1

Proportion of Deaths Due to Diseases and Injuries in Children under Five Years, by Source of Data.

The high proportion of deaths for which the cause was unknown raises serious questions about data quality. The largest number of deaths in this review was in the vital records for Abidjan, Côte d'Ivoire (1987–92). However, these too have high proportions of "other" causes (29 percent). A detailed analysis of specific causes based on the data from these studies suggested that respiratory infections, diarrhea, malnutrition, and anemia were the leading causes of death in this age group.

At ages 5 to 14 years (see table 5.2), the largest number of deaths was also in the vital records from the same source, although the total number of deaths registered is very small. In such situations, deriving a meaningful estimate of even the proportions across broad cause groups is difficult, because paucity in numbers is coupled with high proportions of ill-defined causes (44 percent). Nevertheless, the results suggested that malaria, diarrheal diseases, and malnutrition were the leading causes of death among school-age children.

Table 5.2. Proportion of Deaths Due to Diseases and Injuries in Children Age 5 to 14 Years, by Source of Data.

Table 5.2

Proportion of Deaths Due to Diseases and Injuries in Children Age 5 to 14 Years, by Source of Data.

For ages above 15 years, epidemiological studies on causes of death concentrate on causes of maternal mortality. Adetunji, Murray, and Evans (1996) reviewed information from 21 studies and identified 3,818 deaths due to maternal causes. About 80 percent of these studies were based on hospital records. This could have two possible implications. On the one hand, hospital-based studies could overestimate maternal mortality, as high risk and emergency cases are likely to be concentrated in hospitals. On the other hand, better emergency care in hospitals may avert more deaths than would have occurred in the community. Notwithstanding these possible biases, the causes as documented in hospital records are likely to be more reliable than those from community-based studies. The pooled results suggested hemorrhage (19 percent), puerperal sepsis (13 percent), hypertensive disorders of pregnancy (7.8 percent), and ruptured uterus (7 percent) as the leading causes of maternal mortality in Sub-Saharan Africa, at least among women referred to hospitals.

Adetunji, Murray, and Evans (1996) also documented three sources of data on causes of adult mortality; two from vital records in Lagos (1977) and Abidjan (1987–92), and one from the Adult Morbidity and Mortality Project (AMMP) in Tanzania (1992). The data from Lagos predate the HIV/AIDS period and record accidents and violence as the leading cause of adult deaths (26.3 percent). In Abidjan, the data suggest hypertensive disease (31 percent), diarrheal disease (11 percent) and HIV/AIDS (10.5 percent) as the leading causes. Data from the AMMP are available by sex and show HIV/AIDS to be the leading cause in both sexes (higher proportions in women) and injuries in males and pregnancy-related causes to be the other significant conditions. All these results suggest that hypertensive disease, HIV/AIDS, pregnancy-related causes, and injuries are the leading causes of death among adults in Sub-Saharan Africa.

We have described this review in some detail in part because it was the most comprehensive previous attempt to estimate causes of death in Africa, and in part to show that compiling information from various sources, despite the enormous effort involved, still results in substantial uncertainty about the cause structure of mortality owing to biases in the way the data were collected and the high proportions of unspecified causes in the reports. However, these epidemiological studies were designed to focus on one or a few specific causes of mortality, yielding useful information on incidence, duration, and case-fatality rates. These indexes were used by independent researchers to infer mortality specific to that cause at the population level, and such epidemiological estimates for individual causes are useful to build the overall cause-specific mortality picture.

Information from Vital Records

Another data collection effort was centered on the acquisition of vital records data. The World Health Organization (WHO) conducted a comprehensive search for these data over the period 2001–02, as part of the data collection for the GBD 2000 project (Kowal, Rao, and Mathers 2003). Few countries in Africa have vital registration systems that are more than 50 percent complete. Coverage is about this level in Kenya and Zimbabwe (Lopez et al. 2002), and close to 90 percent in South Africa (Dorrington et al. 2001). In Mozambique a major exercise was undertaken to improve mortality registration and cause-of-death attribution in four cities (Cliff et al. 2003). The ministries of health in Botswana, Eritrea, and Zambia collect and collate information on causes of death from health facility data and from vital records, essentially in urban areas. The AMMP in Tanzania has built up a comprehensive district-level mortality surveillance system that operates currently in three districts (Hai, Morogoro, and Dar es Salaam) and has compiled information on causes of death over a 10-year period. Vital records data from the city of Antananarivo have been systematically compiled over a 12-year period to describe urban causes of death in Madagascar (Waltisperger, Cantrelle, and Ralijaona 1998). Data from all these countries have been collated and analyzed at the WHO.

Vital records data, although impressive in overall numbers of deaths for which information on cause is available, have their own limitations. First, coverage is incomplete, and it is not possible to map these data to an underlying population to make an assessment of death rates. Second, there is no indication of the bases used for cause-of-death attribution, whether physician attribution at the time of death, hospital records, verbal autopsy, or lay reported information. The proportionate distribution of causes of death are therefore examined in this chapter according to the GBD groups at different ages, as was done in the earlier analyses using data from epidemiological studies.

Included in these tabulations are the proportionate distributions for two epidemiological subregions in Sub-Saharan Africa (AFR D and AFR E),1 as estimated by the GBD cause-of-death study. These subregions group countries with similar estimated levels of child and adult mortality, as described in detail later in the section on the GBD estimation process. For comparison purposes, and so their plausibility may be judged, these figures are included here with available local data.

Among children under five, the highest proportion of deaths, as expected, is from group 1 causes (perinatal conditions, communicable diseases, and malnutrition), ranging from 75 percent in the South African data set to 94 percent in the data from Zambia. A reassuring feature of these vital records data is the relatively low proportion of deaths not classified in any of the three cause groups. In the GBD estimates, unclassifiable deaths at ages below five years are reallocated to group 1 (Murray and Lopez 1996). Hence, we estimate that about 95 percent of deaths at these ages are due to group 1 causes, which approximates the upper end of the range of observed proportions from this cause category from the vital records data sets (table 5.3).

Table 5.3. National Vital Records Data: Proportionate Distribution of Cause of Death of Children under Five.

Table 5.3

National Vital Records Data: Proportionate Distribution of Cause of Death of Children under Five.

As age increases, the proportion of deaths due to noncommunicable diseases and injuries rises. In global comparisons, this shift in cause composition across ages is less evident in Sub-Saharan Africa than in other regions, due to the catastrophic HIV/AIDS epidemic. At ages 5 to 14 years, the South African data set suggests a low proportion of deaths due to group 1 causes, which may be because the population covered by registration is urban, with higher socioeconomic status (expressed as GDP per capita) than the national average (table 5.4). Another possible cause for low group 1 proportions is that the coverage of vital registration in South Africa is only 90 percent, and the missed deaths could be group 1 deaths. However, such corrections would result only in marginal changes in the specific proportions of the groups. Again, from all the data sets, the proportion of deaths due to ill-defined causes is low. In the GBD estimation process, these deaths due to ill-defined causes at ages above five years are reallocated proportionately across group 2 causes. In comparison, group 1 proportions in the estimates again approximate the higher end of the range of observations in the local data.

Table 5.4. National Vital Records Data: Proportionate Distribution of Cause of Death at Age 5 to 14 Years.

Table 5.4

National Vital Records Data: Proportionate Distribution of Cause of Death at Age 5 to 14 Years.

Much more complex is to estimate cause-specific mortality for the young adult (15 to 44 years) age group. At these ages in Sub-Saharan Africa, a triple burden of cause-specific factors exists, namely, the HIV/AIDS epidemic; high incidence rates of injury and violence, especially among males; and the high burden of maternal mortality among females. These aspects of cause-specific mortality are evident from the results of the epidemiological literature review discussed earlier. Table 5.5 shows the group-specific proportions from vital records at these ages. The proportion of deaths in vital records due to injuries at these ages appears to be low in most countries, given the wars and violence in the region. The GBD estimates of war-related deaths are included in deaths due to violence and marginally increase the proportions of deaths due to injuries. Also, the relatively low proportion of deaths due to group 1 causes in the Madagascar data set could be because the population covered was urban, or that HIV prevalence, transmission, and mortality is low in Madagascar (Ravaoarimalala et al. 1998). Group 1 conditions appear to cause a major proportion of mortality at these ages.

Table 5.5. National Vital Records Data: Proportionate Distribution of Cause of Death at Age 15 to 44 Years.

Table 5.5

National Vital Records Data: Proportionate Distribution of Cause of Death at Age 15 to 44 Years.

Local data for deaths at older ages (45 years and older) from all sites have significantly higher proportions of unclassifiable deaths, approaching 31 percent in Zimbabwe (table 5.6). At these ages, one would normally expect higher proportions of deaths from noncommunicable diseases, which is reflected in the data. There is a possibility of misclassification between group 2 and group 1, since in some instances individuals with long-standing noncommunicable diseases develop infectious complications during terminal stages, which tend to be more readily identified and classified as the underlying cause of death. The estimates are close to the upper end of the range of proportions of group 2 conditions from local data.

Table 5.6. National Vital Records Data: Proportionate Distribution of Cause of Death at 45 Years of Age and Older.

Table 5.6

National Vital Records Data: Proportionate Distribution of Cause of Death at 45 Years of Age and Older.

From the above analyses, vital registration data would seem to be useful in understanding general patterns of cause-specific mortality at broad-cause-group level. However, biases introduced by incomplete or selective coverage and difficulty in identifying specific causes beyond the broad-cause-group level preclude their usage directly in the estimation process. In situations of incomplete data, the process of estimating mortality from specific causes entails the synthesis of information from different data sources or, if necessary, by drawing on other regional or international cause-specific mortality patterns.

GBD Process for Estimating Cause-Specific Mortality

The GBD Study method of estimating cause-specific mortality in populations without detailed information on the levels or cause structure of mortality is essentially based on three sequential steps:

  1. The first step is to derive an overall envelope of mortality, in terms of estimated numbers of deaths, for each age-sex group within the population. In practical terms, a model life table–based age-specific risk of mortality is applied to a national age-sex population estimate. The purpose of this step is to provide an upper limit or "demographic envelope," to constrain cause-of-death estimates within the bounds of demographic plausibility.
  2. The second step is to use cause-of-death models to predict a GBD group-specific (broad causes) proportionate composition of mortality for each age-sex group and to apply the proportions to the mortality envelope derived in step 1 to derive an age-sex GBD group-specific envelope of mortality.
  3. The third step is to fit a condition-specific cause structure of mortality for each broad-cause group onto their respective envelopes as derived in step 2 to derive estimates of deaths from specific conditions by age and sex for the population.

In the current analysis, separate estimates were developed for each country in Sub-Saharan Africa, and they were summed up to generate two subregional population aggregations. The classification of WHO member states into the mortality strata was carried out using population estimates for 1999 (United Nations Population Division 1998) and estimates of child mortality (defined as 5q0; the risk of dying between birth and age 5) and adult mortality (defined as 45q15; the risk of dying between ages 15 and 60) based on WHO analyses of mortality rates for 1999 (Mathers, Stein et al. 2002). Five mortality strata were defined in terms of quintiles of the distribution of 5q0 and 45q15 (both sexes combined). Adult mortality 45q15 was regressed on 5q0 and the regression line used to divide countries with high child mortality into high adult mortality (stratum D) and very high adult mortality (stratum E), as in figure 5.1. Stratum E includes the countries in Sub-Saharan Africa where HIV/AIDS has had a substantial impact.

Figure 5.1

Figure 5.1

Global Mortality Strata for GBD 2000 Regions Source: Mathers, Stein et al. 2002.

A fundamental principle in the estimation process is to maintain internal consistency between the overall demographic envelope of mortality and the cause composition within the envelope. In other words, the mortality due to specific causes within an age group should exactly sum up to the life table–derived envelope of mortality. Such constraints call for significant assumptions and judgments; hence, the final estimates should be interpreted based on these choices. With the high estimated levels of HIV/AIDS mortality in the region, judgments regarding burden due to competing causes of death are based on evidence from vital records and epidemiological estimates for specific diseases. However, preceding these issues is the process of deriving the life tables for estimating the mortality envelope by age and sex and the use of cause-of-death models for allocating deaths to broad-cause groups.

Life Tables

Detailed age-specific mortality rates are required to construct life tables for populations. Such rates are not available for any country within Sub-Saharan Africa. Attempts at estimating death rates based on recall of deaths in censuses and surveys have generally led to implausibly low death rates (Brass 1968). The best available data on mortality risk are levels of child mortality derived from Demographic and Health Surveys (DHS) conducted in most countries within the region during the 1990s. These estimates are derived using data from birth histories collected from women respondents age 15 to 49 years. In a few countries, information on sibling survival, subsequently analyzed to derive estimates of adult mortality, has also been collected and compiled.

A new modified logit model life table system (Murray et al. 2003) allows for predicting an abridged life table using as inputs true or estimated levels of child and adult mortality or, if necessary, using only levels of child mortality. Recent work by Murray and colleagues has resulted in development of a validated life table system that uses a single global standard that represents the full range of mortality patterns seen in contemporary human populations; the system generates better predictions of age-specific mortality rates than the Coale-Demeny or original Brass systems.2

A set of 1,802 life tables, based on empirical data from national mortality registration systems between 1901 and 1999, were used to develop and test this model. These life tables were based on observed mortality risks in HIV-free populations, and hence, predicted life tables from the model can be used to estimate HIV-free age-specific mortality. In this context, estimates of child mortality from DHS surveys for the 52 countries in Sub-Saharan Africa were used to develop HIV-free envelopes of mortality at the national level. Steps 2 and 3 as described earlier were carried out for each country to construct a cause-specific mortality estimate without accounting for HIV, and as a final step, epidemiological estimates of HIV mortality were added on to derive the overall national age-, sex-, and cause-specific mortality estimate (Lopez et al. 2002). Figure 5.2 shows, by way of example, the final age-specific death rates for Zambia, for the year 2000. Note the markedly high death rates at young adult ages, including the deaths due to AIDS.

Figure 5.2

Figure 5.2

Estimated Age-Specific Death Rates for Zambia, 2000 Source: Authors.

Compositional cause-of-death models have been developed around the relationship between cause-specific mortality and total mortality by age, as observed from patterns of causes of death recorded in nations with good vital registration systems (Salomon and Murray 2002). For each age-sex group an income variable for each observed country-year of data has been added to the model to strengthen the predicted proportionate composition by the three GBD cause groups. The models have been built using an empirical data set of 1,576 country-years of observations from 58 countries for the years 1950 to 1998. A set of regression equations have been developed that predict the cause composition of mortality by age, for a given input of total mortality by age and an estimate of national per capita gross domestic product (GDP). For each country, the estimate of total mortality by age and sex derived from the life tables (from step 1) and the national GDP are used as inputs, and the model produces an implied cause composition of mortality by age and sex. Since the empirical data set for developing the regression equations does not include any information on mortality in countries experiencing AIDS epidemics, it is difficult to predict with confidence the proportions of deaths due to group 1 causes at different ages. Wherever possible, the outputs of the model were validated against compositional structures from available vital records data from the country or region and from small-scale longitudinal population surveillance systems. An example of the model outputs is shown in figure 5.3, depicting the predicted proportionate composition by age for males and females, respectively.

Figure 5.3

Figure 5.3

Model-Based Predictions of GBD Cause–Group Composition of Mortality by Age and Sex, Zambia, 2000 Source: Authors.

Communicable diseases (group 1) account for a large proportion of deaths at all ages up to about age 60. The proportions derived from the model are used to derive GBD group-specific numbers of deaths for each age group from the life table–derived total numbers of deaths in each age group. After this, information from vital records or specific disease epidemiological studies are used to construct the detailed cause-of-death estimates.

Epidemiological Estimates

Cause-of-death models in combination with life table–derived mortality risks and national population estimates provide an estimate of the numbers of deaths by age, sex, and broad-cause group. Estimating the numbers of deaths due to specific causes within each broad group—for example, the composition within group 1 due to different infectious diseases and perinatal disorders at infant ages or the composition within group 2 due to neoplasms, cardiovascular diseases, and other noncommunicable diseases at adult ages—is the next stage. The mass of epidemiological evidence on mortality due to different infectious diseases and maternal causes over the past two decades has been invaluable in this estimation process. Researchers have critically examined data from different studies and used them to derive independent epidemiological estimates of cause-specific mortality for the region. It is useful to examine in brief the results of these independent estimation exercises and to assess their plausibility with reference to an overall demographic envelope of mortality. We have included in this discussion results from such estimation exercises available for Africa in 2003. There is a need to conduct similar estimations for other important conditions, such as tuberculosis and road traffic accidents.


Malaria is one disease that has been extensively researched in Africa. Snow and his colleagues initially estimated malaria mortality in the region for 1990 and later refined their approach to produce another set of estimates for 1995 (Snow et al. 1999). They integrated evidence from research on several aspects of the epidemiology of malaria. First, they used a detailed analysis of environmental factors that affect the distribution, seasonality, and transmission intensity to develop an epidemiological stratification of the continent into five regions based on climatic suitability for the existence and stability of malarial transmission. In the next step, they used geographical information systems to define at-risk populations within regions suitable for transmission based on different transmission scenarios as described. A specially constructed population database for more than 4,000 administrative units in Africa (Deichmann 1996) was used in this step. Finally, direct estimates of fatal risks for malarial mortality were selected from over 200 empirical data sources on the health impact of malaria, and these were combined with estimates of at-risk populations to derive age-specific estimates of deaths due to malaria in the region, as summarized in table 5.7. The median estimates were taken as the guideline for the GBD estimates for the entire region.

Table 5.7. Epidemiological Estimates of Malaria Mortality in Nonpregnant African Population, 1995.

Table 5.7

Epidemiological Estimates of Malaria Mortality in Nonpregnant African Population, 1995.

Diarrheal Diseases

Estimates of death rates due to diarrheal disease from epidemiological studies vary widely across populations, and within populations over time. Also, complex interactions between intrinsic factors, such as nutritional status, and extrinsic factors, such as exposure to multiple infections, make the attribution of death to a single underlying cause extremely difficult. Kosek, Bern, and Guerrant (2003) reviewed 34 studies in 21 countries to derive an estimate of diarrheal mortality rates. Studies were included only if diarrheal deaths were ascertained through active surveillance, and in the selected studies only a primary cause listed as diarrhea was considered as a diarrheal death. As a result, they estimated a mortality rate of 4.9 per 1,000 children under five per year (95 percent confidence interval (CI), 1.0–9.1) in countries with high levels of overall child mortality. On a global scale, this appears as a decline from an estimated rate of 13.6 per 1,000 in 1982 (Snyder and Merson 1982).

In view of different cause-of-death attribution strategies in different settings, proportionate mortality due to diarrhea is considered as an alternate approach to estimating the number of deaths due to this cause. Morris, Black, and Tomaskovic (2003) developed a prediction model to estimate the distribution of deaths among children under five by cause. The model estimated that about 20 percent of all under-five deaths in Sub-Saharan Africa would be caused by diarrhea.

Table 5.8 shows the total population of children under age five in the region, and the estimate of diarrheal deaths on applying the median observed death rate of 4.9 per 1,000 (Kosek, Bern, and Guerrant 2003), and a proportionate mortality of 20 percent (Black, Morris, and Bryce 2003).

Table 5.8. Estimates of Diarrheal Deaths of Children under Five, 2000.

Table 5.8

Estimates of Diarrheal Deaths of Children under Five, 2000.

Vaccine-Preventable Diseases

Despite expansion in immunization services in the developing world, vaccine-preventable diseases remain significant causes of mortality in Africa. Stein and colleagues (2003) conducted an epidemiological exercise to estimate measles mortality. They used national-level information on vaccine coverage from health surveys and estimates of disease incidence, vaccine efficacy, and case fatality from specific epidemiological studies to develop a static model by which to estimate regional and global measles mortality. They estimated a total of 452,000 deaths due to measles in Sub-Saharan Africa in the year 2000, of which approximately three-fourths would occur among children under five.

Crowcroft and colleagues (2003) developed a model by which to estimate cases and deaths due to pertussis for the year 1999. Parameters used in the model included vaccine coverage and efficacy data and estimates of case-fatality ratios from epidemiological studies. Two coverage scenarios were used in the model, at levels below and above 70 percent, based on information from WHO reports adjusted by survey data where available. For each scenario, a different age structure was used to represent a range of possible patterns of infection in susceptible children. The model assumed a vaccine efficacy of 95 percent for preventing death. They estimated 170,000 deaths due to whooping cough in the African region in 1999.

Lower Respiratory Infections

Acute respiratory infections (ARI) are the third leading cause of death globally among children under five years of age. Much uncertainty surrounds the ascertainment of pneumonia as the underlying cause of death, principally from verbal autopsy–based studies in developing countries. In these settings, significant comorbidity exists in the form of measles, whooping cough, diarrheal diseases, or malaria. Williams and colleagues (2002) developed a model to predict the proportion of ARI mortality for a given level of under-five mortality. A review of 49 studies provided data on the level of child mortality and the proportion of deaths due to ARI. The researchers fitted a log linear curve to the data; from the resultant equation they estimated the number of deaths due to ARI for all countries, using WHO estimates of country-specific under-five mortality in the year 2000. Through this model they estimated a total of 794,000 deaths due to ARI in the region in the year 2000. The authors demonstrated that differences between predicted proportions from the model and observed proportions from verbal autopsy studies could be explained by the variability inherent with the use of verbal autopsy methods, induced by such comorbidity.

Maternal Mortality

Hill, AbouZahr, and Wardlaw (2001) in collaboration with researchers at the WHO collated available evidence on country-specific maternal mortality in the form of vital registration records, DHS-type sibling survival surveys, and special Reproductive Age Mortality Studies to develop a statistical model to predict the proportion of deaths of women of reproductive ages due to maternal causes (PMDF). The PMDF was found superior to the maternal mortality ratio (MMR) as the dependent variable for estimating the number of deaths from maternal causes, mainly because of its applicability to a demographic "envelope" of deaths at maternal ages. Also, information from national-level sisterhood surveys were found to yield more robust measures of PMDF than MMR to use as inputs in the construction of the model. Independent variables chosen to predict national PMDF were general fertility rate, female literacy, per capita income, percentage of deliveries attended by skilled attendants, country-specific estimates of HIV prevalence, and variables for region and level of vital registration in individual countries. The predicted PMDF for each country was then applied to age- and sex-specific mortality estimates from World Population Prospects (United Nations Population Division 1998), to derive the absolute numbers of deaths due to maternal causes. Based on the model, Hill, AbouZahr, and Wardlaw (2001) predicted a total of 272,500 maternal deaths in Sub-Saharan Africa for 1995, with a mean MMR of 1,006 per 100,000 live births.


The Joint United Nations Programme on HIV/AIDS (UNAIDS) and the WHO established an Epidemiology Reference Group to work toward making estimates and projections of mortality on a biennial basis. Country-level information on prevalence of infection among attendants at antenatal clinics and sexually transmitted disease clinics and prevalence among other high-risk groups, such as intravenous drug users, homosexual males, and commercial sex workers, has been used to monitor epidemics at the country level. Using these data, an epidemiological model was developed that included the following parameters:

  1. the initial rate (r) of spread of HIV as determined by the reproductive potential
  2. the peak prevalence (f) as determined by the fraction of population at risk of infection
  3. the final epidemic prevalence (ϕ) as determined by the behavioral response of the population
  4. the start date of epidemics in individual countries.

A negative value of ϕ indicates that people become less likely to adopt risky behavior in response to observed AIDS mortality or prevention programs. Hence, apart from prevalence data from sentinel sites, behavioral surveys are essential to assess the potential of HIV epidemics and to estimate the size of at-risk populations.

An assessment of factors that affect survival of adults suggested that an overall median survival time of 9 years, with a range of uncertainty of 8 to 11 years, and a Weibull distribution of the survival function were used in the modeling process. For children, the survival curve was built to account for two periods of high mortality, which are infancy, when HIV frequently overwhelms the immature immune system, and after age nine years, when the response to HIV infection resembles that in adults. Overall, the survival curve for children predicts 40 percent survival from HIV-related mortality at five years of age.

The UNAIDS/WHO model was used to develop country-specific point estimates of HIV/AIDS mortality for 2001 (WHO and UNAIDS 2002). The estimated regional death toll from this disease for Sub-Saharan Africa stands at a total of 2.2 million deaths. Policy decisions aimed at addressing this epidemic should include activities aimed at improving such measurements.


Ferlay and colleagues (2001) at the International Agency for Research in Cancer developed a data set of worldwide estimates of cancer incidence, mortality, and prevalence for the year 2000, which they called the Globocan 2000. Their mortality estimates were based on vital registration data where available; for other regions they used information from survival models derived from available cancer registry data (Sankaranarayan, Black, and Parkin 1998). These mortality estimates did not correct for under-reporting in vital registration or for possible misclassification of causes of death.

As part of the GBD 2000 Study, Mathers, Shibuya et al. (2002) used relative interval survival data from the Surveillance Epidemiology and End Results (SEER) Program at the National Cancer Institute in the United States to develop an age-period-cohort survival model, which was further adjusted for levels of economic development. The model was then applied to age-, sex-, and site-specific incidence estimates from cancer registries that were compiled for the Globocan 2000 and from other specific incidence studies. The model was useful to smooth out random variations in observed incidence and survival rates that resulted from small numbers of cases or cases lost to follow up. The estimated cancer survival rates by site, age, and sex for different regions in the world were used as key inputs to estimate the distribution of cancer deaths by site. Estimates of cancer mortality by region were then developed (Shibuya et al. 2002), correcting for levels of overall mortality in regions with incomplete coverage of registration, and also correcting for the likely differences in cause-of-death patterns that would be expected in uncovered and often poorer subpopulations. As a result, the GBD 2000 Study estimated a total of 572,000 deaths due to cancer within the region.


Deaths due to war within Sub-Saharan Africa merit attention as an avoidable burden. Murray and colleagues (2002) used a comprehensive analysis of media reports, vital registration records, and adjustments based on observed relationships between direct and indirect mortality to estimate the global burden of mortality due to armed conflict. Keeping in mind the limitations of estimates based largely on qualitative analysis of media reports, conservative estimates were used for some of the major conflicts in the world. According to these estimates, armed conflicts had resulted in an overall death toll of about 77,000 for Sub-Saharan Africa for the year 2001.

Estimating Cause-Specific Mortality

Using all the various dimensions of cause-specific mortality estimation for Sub-Saharan Africa—available data from vital records, specific studies, and global disease epidemiological extrapolations for the region—we have prepared estimates of causes of death in Sub-Saharan Africa based on the GBD approach. Figure 5.4 summarizes the estimation process for countries within the region.

Figure 5.4

Figure 5.4

Summary of GBD Process for Estimating Cause-Specific Mortality in African Countries Source: Authors.

As mentioned earlier, predicted proportions for broad-cause groups by age and sex were validated against information from country-specific vital records where available and for neighboring or epidemiologically similar countries where possible. Also, wherever applicable, cause-specific proportions from vital records for specific causes were used in deriving population-level estimates for these causes. The resultant cause-specific structure was finally adjusted with disease-specific mortality estimates to produce the overall numbers of deaths by age, sex, and cause. National estimates were summed up to totals for two subregions and for Sub-Saharan Africa as a whole. During this process of synthesizing epidemiological estimates into the overall demographic and cause-of-death model based on the broad-cause group envelope of mortality, some of the disease-specific estimates described earlier were reduced to meet these envelope constraints.

Results and Discussion

Distinct differences exist between countries belonging to the two epidemiological subregions, essentially divided by the level of HIV/AIDS-related mortality within them.

Overall, we estimate that about 10.8 million deaths (table 5.9) occurred in the year 2002 in the region, or just under 20 percent of global mortality. The age structure of mortality roughly divides into 46 percent of deaths occurring before the age of 15 years, another 36 percent between the ages of 15 and 59 years, and the remaining 18 percent at age 60 and above.

Table 5.9. GBD Estimates of Leading Causes of Death, by Sex, 2000.

Table 5.9

GBD Estimates of Leading Causes of Death, by Sex, 2000.

It is not surprising, therefore, that five of the six estimated leading causes of mortality in Sub-Saharan Africa are those that cause deaths at childhood ages, namely, infectious diseases and conditions originating in the perinatal period (also see table 5.10). About 20 percent of the estimated 2.2 million deaths due to HIV/AIDS are also predicted to occur in childhood. Although the rank order of tuberculosis is almost similar in males (7) and females (9), the estimated number of deaths in males (about 210,000) is more than double that predicted in females (95,000). Similarly, road traffic accidents, chronic obstructive pulmonary disease, and war were estimated to cause more than double the number of deaths in males than in females. In the entire region, maternal conditions are estimated to cause 4.4 percent of deaths in females.

Table 5.10. GBD Estimates of Leading Causes of Death in AFR D and AFR E, 2000.

Table 5.10

GBD Estimates of Leading Causes of Death in AFR D and AFR E, 2000.

A subregional disaggregation of the magnitude of cause-specific mortality sheds more light on the epidemiological variation possible in Sub-Saharan Africa. Two-thirds of the total mortality occurs in countries within the AFR E region, which contains just over half (54 percent) of the regional population. A comparison of the rank structure of leading causes of death between the two epidemiological subregions, AFR D and AFR E, is shown in table 5.10.

The threefold difference in rank and percentage of deaths caused by HIV/AIDS between the two regions translates into a nearly fourfold difference in absolute numbers of deaths. There is reasonable similarity in the rank order and magnitude of the other leading causes, when deaths due to all ages are combined. The importance of war as a cause of death at all ages in AFR E stands out, being ranked as the 12th leading cause.

At childhood ages, the rank structure and proportion of deaths from individual leading causes is almost similar in the two regions, and between males and females, except for the HIV/AIDS subregional difference mentioned earlier. Overall, the subregional proportions of child mortality are nearly equal; 48 percent of deaths occur in AFR D, and 52 percent occur in AFR E. Sex differentials too are slight, with 52 percent of deaths in males and 48 percent in females.

Table 5.11 shows the leading causes of child deaths for the two subregions together. As expected, estimated leading causes in Sub-Saharan Africa are infectious diseases, perinatal conditions, and malnutrition. Two important implications of these observations are that childhood mortality remains a major cause of burden in Sub-Saharan Africa and that there is no difference in the magnitude of the burden between different populations within the region.

Table 5.11. GBD Estimates of Leading Causes of Death at Age 0 to 14 Years, 2000.

Table 5.11

GBD Estimates of Leading Causes of Death at Age 0 to 14 Years, 2000.

When comparing mortality during adulthood, however, there are sizable differences, both between the two subregions and, within each subregion, between males and females. The mortality differences between the two subregions, as seen from the leading causes for both sexes combined, are shown in table 5.12.

Table 5.12. GBD Estimates of Leading Causes of Death at Age 15 to 59 Years, Subregional Comparison, 2000.

Table 5.12

GBD Estimates of Leading Causes of Death at Age 15 to 59 Years, Subregional Comparison, 2000.

The most striking feature is that the estimated number of deaths at these ages in AFR E is nearly double that estimated for AFR D. Readers will recall that the number of childhood deaths in the two regions is nearly equal. From table 5.12 it is clear that in AFR E, the HIV/AIDS epidemic and in specific countries, armed conflicts are the causes of the excess mortality. Clearly, the number of deaths from each of the leading causes is significantly higher in AFR E, despite a somewhat similar rank structure of causes apart from HIV/AIDS and war.

As observed at childhood ages, the difference between mortality in males and that in females is minimal (52 to 48). The leading causes of mortality in each sex in AFR E are shown in table 5.13.

Table 5.13. GBD Estimates of Leading Causes of Death in AFR E at Age 15 to 59 Years: Comparison between Males and Females, 2000.

Table 5.13

GBD Estimates of Leading Causes of Death in AFR E at Age 15 to 59 Years: Comparison between Males and Females, 2000.

Apart from the high burden of mortality due to infectious diseases, it is evident that injury-related causes of death among males, pregnancy-related causes among females, and cardiovascular diseases in both sexes are major issues to be dealt with. The high mortality due to HIV/AIDS in both sexes has a major bearing on the occurrence of orphanhood within these countries.

The subregional differences in mortality among the elderly is similar to that observed in children: 47 percent in AFR D and 53 percent in AFR E. The leading causes of death at these ages are similar for the two mortality strata. For convenience, the discussion here will focus on the difference in the cause-of-death structure between the two sexes, aggregated for both regions together or, in other words, for Sub-Saharan Africa.

As expected, in both sexes, cardiovascular diseases are estimated as the leading causes of mortality among the elderly (table 5.14). Among males, chronic obstructive pulmonary disease and prostate cancer are other leading causes, whereas in women, kidney disorders and diabetes mellitus are major causes of mortality. Cirrhosis of the liver is estimated as the 11th leading cause in males; the corresponding rank for women is taken by breast cancer. The presence of lower respiratory infections, tuberculosis, diarrheal diseases, and HIV/AIDS among the leading causes of death in both sexes, even among the elderly, clearly defines the importance of communicable disease control in African countries.

Table 5.14. GBD Estimates of Leading Causes of Death at Age 60 Years and Older: Comparison between Males and Females, 2000.

Table 5.14

GBD Estimates of Leading Causes of Death at Age 60 Years and Older: Comparison between Males and Females, 2000.


Estimating cause-specific mortality in Sub-Saharan Africa poses a major epidemiological challenge. The availability of data and information on causes of death has increased within the region, and the results lead us to a few important observations. First, childhood mortality is a major cause of the high premature mortality rates in Africa, accounting for nearly half the total mortality in the region. As a corollary, the observed leading causes of childhood mortality—malaria, diarrheal diseases, measles, lower respiratory tract infections, and conditions originating in the perinatal period—require immediate attention.

Second, the young adult population in countries within the AFR E mortality subregion is at significant risk of possible premature death from HIV/AIDS, armed conflict and other forms of violence, road traffic accidents, tuberculosis, and, among women, causes related to pregnancy and childbearing.

Third, nearly a fifth of the mortality in the region occurs among individuals age 60 years and older. The proportion of deaths at these ages is much less than that observed in developed countries. For instance, in Australia, over 80 percent of deaths occur above the age of 60 years (ABS 2003). However, in Sub-Saharan Africa, the absolute number of deaths itself—1.92 million—merits attention to the health needs of the elderly. At these ages, cardiovascular disease, chronic obstructive pulmonary disease, cancers (prostate cancer in males, cervix and breast cancer in females), and, notably, infectious diseases are the major causes of mortality.

It is generally accepted that statistics from complete vital registration systems are the "gold standard" for national mortality statistics. Although cause-of-death information from vital records is subject to some biases on account of quality of cause-of-death attribution within the system, there is ample global evidence to justify a reliance on such data for national cause-specific mortality analysis and estimation.

It is heartening to note that countries within the region are making special efforts to improve vital registration systems. The process, however, involves huge resources, and can be expected to take decades before data of reasonable quality from national vital registration systems will be available for such estimation exercises. It has been observed in Kenya, South Africa, and Zimbabwe that although the coverage of registration of vital events by age and sex can be improved rapidly by instituting certain administrative reforms, obtaining information on the cause of death remains elusive, on account of deaths occurring at home in the absence of medical attention in remote areas.

An alternative system that has been tested and found effective has been sample registration, as has been employed in India and China. A representative sample of villages or population clusters is routinely monitored for vital events, and upon death, a formal verbal autopsy procedure is employed to ascertain its probable cause. Initiatives such as the Health Metrics Network (established by the WHO) are currently devising a framework for testing and establishing the Sample Vital Registration and Verbal Autopsy method of obtaining information on causes of death for national mortality estimation purposes. This is the most viable interim solution to meet requirements of data for both health policy as well as for monitoring the impact of health programs and interventions. The GBD method of estimation, with the extensive use of models and extrapolations from specific epidemiological studies, may not be useful for the monitoring function.

Another area of data collection that warrants attention at this stage is the function of physician certification of cause of death and the implementation of guidelines, based on the International Classification of Diseases, for coding and classification of causes of death. Capacity building in these areas will assist national health information systems in providing useful information on causes of death, at least in urban areas, where a significant number of deaths would occur in hospitals or where the deceased may have been attended by a physician in the time immediately preceding death.

Undoubtedly, there is vast uncertainty about causes of death in Africa, but enough is currently known to prepare preliminary estimates, adhering to a rigorous scientific framework for evaluation of data quality and ensuring substantial prudence in interpreting the findings. What data are available suggest that further, massive responses to the HIV epidemic are needed and that major communicable diseases and maternal health require scaled-up investments. Prevention of injuries, in part from war, would contribute much to the improvement of health and survival of young adult males. But what is most urgently needed is investment in cost-effective methods to monitor mortality if we are not to be similarly ignorant about health conditions in Africa 10 years hence.


  1. ABS (Australian Bureau of Statistics). 2003. Deaths Australia. Vol. 3302. Annual report. Canberra: ABS.
  2. Adetunji, J. A., C. J. Murray, and T. Evans. 1996. "Causes of Death in Africa: A Review." Paper presented at a meeting of the Population Association of America, New Orleans, May.
  3. Black R. E., Morris S. S., Bryce J. Where and Why Are 10 Million Children Dying Every Year? Lancet. 2003;361:2226–34. [PubMed: 12842379]
  4. Brass, W. 1968. Demography of Tropical Africa. Princeton, NJ: Princeton University Press.
  5. Cliff, J., J. Sacarlal, O. Augusto, A. Nóvoa, M. Dgedge, G. Machatine, and H. Cossa. 2003. Estudos das principais causas de morte registadas, nas citades des Maputo, Beira, Chimoio e Nampula, em 2001. Maputo City, Mozambique: Ministerio de Saude.
  6. Crowcroft N. S., Stein C., Duclos P., Birmingham M. How Best to Estimate the Global Burden of Pertussis? Lancet Infectious Diseases. 2003;3:413–18. [PubMed: 12837346]
  7. Deichmann, U. 1996. African Population Database. Digital Database and Documentation. National Center for Geographical Information and Analysis, University of California, Santa Barbara. http://www​.glowa-volta​.de/cd_v3.1/landuse/populat/africa.htm.
  8. Dorrington, R., D. Bourne, D. Bradshaw, R. Laubscher, and I. Timaeus. 2001. The Impact of HIV/AIDS on Adult Mortality in South Africa. Cape Town: Medical Research Council.
  9. Feachem, R. G., and D. T. Jamison, eds. 1991. Disease and Mortality in Sub-Saharan Africa. New York: Oxford University Press.
  10. Feachem, R. G., D. T. Jamison, and E. R. Bos. 1993. "Changing Patterns of Disease and Mortality in Sub-Saharan Africa." In Disease and Mortality in Sub-Saharan Africa, ed. R. G. Feachem and D. T. Jamison. New York: New York University Press.
  11. Ferlay, J., F. Bray, P. Pisani, and D. M. Parkin. 2001. Globocan 2000: Cancer Incidence, Mortality and Prevalence Worldwide. Version 1. IARC CancerBase 5. Lyons: IARC Press.
  12. Hill K., AbouZahr C., Wardlaw T. Estimates of Maternal Mortality for 1995. Bulletin of the World Health Organization. 2001;79(3):182–98. [PMC free article: PMC2566373] [PubMed: 11285661]
  13. Kosek M., Bern C., Guerrant R. The Global Burden of Diarrhoeal Disease, as Estimated from Studies Published between 1992 and 2000. Bulletin of the World Health Organization. 2003;81(3):197–204. [PMC free article: PMC2572419] [PubMed: 12764516]
  14. Kowal, P., C. Rao, and C. D. Mathers. 2003. Report on a WHO Workshop: Minimum Dataset on Ageing and Adult Mortality in Sub Saharan Africa. Geneva: WHO.
  15. Lopez, A. D., O. B. Ahmad, M. Guillot, B. D. Ferguson, J. A. Salomon, C. J. L. Murray, and K. H. Hill. 2002. World Mortality in 2000: Life Tables for 191 Countries. Geneva: WHO.
  16. Mathers C. D., Shibuya K., Boschi-Pinto C., Lopez A. D., Murray C. J. Global and Regional Estimates of Cancer Incidence and Mortality by Site. I. Application of Regional Cancer Survival Model to Estimate Cancer Mortality Distribution by Site. BMC Cancer. 2002;2(1):36. [PMC free article: PMC140322] [PubMed: 12502433]
  17. Mathers, C. D., C. Stein, D. Mafat, C. Rao, M. Inoue, N. Tomijima, C. Bernard, A. D. Lopez, and C. J. L. Murray. 2002. "Global Burden of Disease 2000. Version 2: Methods and Results." GPE discussion paper 50, WHO, Geneva.
  18. Morris S. S., Black R. E., Tomaskovic L. Predicting the Distribution of Under-Five Deaths by Cause in Countries without Adequate Vital Registration Systems. International Journal of Epidemiology. 2003;32(6):1041–51. [PubMed: 14681271]
  19. Murray C. J. L., Ferguson B. D., Lopez A. D., Guillot M., Salomon J. A., Ahmad O. Modified Logit Life Table System: Principles, Empirical Validation, and Application. Population Studies. 2003;57(2):165–82.
  20. Murray C. J. L., King G., Lopez A. D., Tomijima N., Krug E. G. Armed Conflict As a Public Health Problem. British Medical Journal. 2002;324(7333):346–49. [PMC free article: PMC1122272] [PubMed: 11834565]
  21. Murray, C. J. L., and A. D. Lopez, eds. 1996. The Global Burden of Disease. Cambridge, MA.: Harvard University Press.
  22. Preston, S. H. 1976. Mortality Patterns in National Populations: With Special Reference to Recorded Causes of Death. New York: Academic Press.
  23. Ravaoarimalala C., Andriamahenina R., Ravelojaona B., Rabeson D., Andriamiadana J., May J. F., Behets F., Rasamindrakotroka A. AIDS in Madagascar. II. Intervention Policy for Maintaining Low HIV Infection Prevalence. Bulletin de la Société de Pathologie Exotique. 1998;91(1):71–73. [PubMed: 9559169]
  24. Salomon J. A., Murray C. J. L. The Epidemiologic Transition Revisited: Compositional Models for Causes of Death by Age and Sex. Population and Development Review. 2002;28(2):205–28.
  25. Sankaranarayan, R., R. J. Black, and D. M. Parkin. 1998. Cancer Survival in Developing Countries. IARC Scientific Publications 145. Lyons: IARC Press.
  26. Shibuya K., Mathers C. D., Boschi-Pinto C., Lopez A. D., Murray C. J. Global and Regional Estimates of Cancer Incidence and Mortality by Site. II. Results of the Global Burden of Disease 2000. BMC Cancer. 2002;2(1):37. [PMC free article: PMC149364] [PubMed: 12502432]
  27. Snow R. W., Craig M., Deichmann U., Marsh K. Estimating Mortality, Morbidity, and Disability Due to Malaria among Africa's Non-Pregnant Population. Bulletin of the World Health Organization. 1999;77(8):624–40. [PMC free article: PMC2557714] [PubMed: 10516785]
  28. Snyder J. D., Merson M. H. The Magnitude of the Global Problem of Acute Diarrhoeal Disease: A Review of Active Surveillance Data. Bulletin of the World Health Organization. 1982;60:604–13. [PMC free article: PMC2536091] [PubMed: 6982783]
  29. Stein C. E., Birmingham M., Kurian M., Duclos P., Strebel P. The Global Burden of Measles in the Year 2000—A Model That Uses Country Specific Indicators. Journal of Infectious Diseases. 2003;187(Suppl. 1):S8–14. [PubMed: 12721886]
  30. United Nations Population Division. 1998. World Population Prospects: The 1998 Revision. New York: United Nations.
  31. Waltisperger, D. M., P. Cantrelle, and O. Ralijaona. 1998. La mortalite a Antananarivo de 1984 a 1995. Paris: Centre Population et Développement.
  32. WHO (World Health Organization). 2003. World Health Report 2003—Shaping the Future. Geneva: WHO.
  33. WHO (World Health Organization) UNAIDS (Joint United Nations Programme on HIV/AIDS). Improved Methods and Assumptions for Estimation of the HIV/AIDS Epidemic and Its Impact: Recommendations of the UNAIDS Reference Group on Estimations, Modeling and Projections. AIDS. 2002;16:W1–W14. [PubMed: 12045507]
  34. Williams B. G., Gouws E., Boschi-Pinto C., Bryce J., Dye C. Estimates of World-Wide Distribution of Child Deaths from Acute Respiratory Infections. Lancet. 2002;2:25–32. [PubMed: 11892493]



AFR D (high child and high adult mortality): Algeria, Angola, Benin, Burkina Faso, Cameroon, Cape Verde, Chad, Comoros, Equatorial Guinea, Gabon, The Gambia, Ghana, Guinea-Bissau, Liberia, Madagascar, Mali, Mauritania, Mauritius, Niger, Nigeria, São Tomé and Principe, Senegal, Seychelles, Sierra Leone, Togo. AFR E (high child and very high adult mortality): Botswana, Burundi, Central African Republic, Côte d'Ivoire, Democratic Republic of Congo, Eritrea, Ethiopia, Kenya, Lesotho, Malawi, Mozambique, Namibia, the Republic of Congo, Rwanda, South Africa, Swaziland, Uganda, Tanzania, Zambia, Zimbabwe.


The Coale-Demeny system is a system of model life tables that predicts age-specific mortality rates based on two parameters only, the level of (usually child) mortality in any age group and some idea of the relationship between infant and child mortality, which define the family (N, S, E, W). The Brass logit system is a relational model life table system that predicts the set of age-specific death rates from knowledge of any two rates (usually child and adult ages 15–59) and the choice of a survival curve as standard.

Copyright © 2006, The International Bank for Reconstruction and Development/The World Bank.
Bookshelf ID: NBK2298PMID: 21290659


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (5.0M)

Related information

  • PMC
    PubMed Central citations
  • PubMed
    Links to PubMed

Similar articles in PubMed

See reviews...See all...

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...