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 7Levels and Patterns of Mortality at INDEPTH Demographic Surveillance Systems

, , , , and .

Empirical, longitudinal, population-based data on mortality in Africa have, until recently, been unavailable. This critical information gap for Sub-Saharan Africa became even more evident as an impediment to our understanding of health and disease in Africa with the arrival of a major new cause of rapidly increasing mortality in the form of HIV/AIDS. At about the same time, during the 1990s, many large-scale mortality intervention trials were conducted at the community level, principally for understanding the efficacy of new interventions such as vitamin A supplementation, and the use of insecticide-treated mosquito nets for malaria control (Alonso et al. 1991; Binka et al. 1996; Nevill et al. 1996; Ross et al. 1995). All these trials used demographic surveillance systems (DSSs) to measure the impact of mortality. These successful mortality intervention trial efforts focused renewed attention on the usefulness of DSSs for illuminating the fundamental age, sex, and cause structure of mortality in resource-constrained settings in Africa. The malaria intervention trials themselves forced a complete reconsideration of the previously underestimated role of malaria as a direct and indirect cause of mortality in Africa (Breman, Egan, and Keusch 2002). More and more DSS sites established for these trials have continued to operate long after the end of the original trials and have proved increasingly useful for both researchers and policy makers (Armstrong et al. 1999; de Savigny et al. 1999; Tanzania Ministry of Health 1997).

Accurate data on mortality conditions in Africa are still scarce. Until recently, the main tool for bridging this gap was the use of indirect demographic estimation techniques and model age-specific mortality schedules produced by Brass and colleagues (1973), Coale and Demeny (1966), and the United Nations Department of International Economic and Social Affairs (1982). The Brass relational system is based on empirical data collected in West Africa during the middle of the twentieth century. In contrast, neither the Coale and Demeny nor the UN model life table systems use significant amounts of data collected from Africa. Moreover, all three of the systems are based on 30- to 50-year-old data. Given the dramatic demographic changes that have affected Africa in the past 20 to 30 years and the fact that two of the systems are based largely on data collected from other regions of the world, whereas the third is based on data from only one region of Africa, it may be problematic to use them in the current African context. No doubt, the World Fertility Survey (WFS) and the Demographic and Health Surveys (DHSs) have remedied the above situation in part by increasing our knowledge of the level, trends, and differentials in infant and child mortality in the developing world. Also, since independence, several African countries have undertaken national censuses, but mortality data from these sources are often plagued with underreporting and need, therefore, to be adjusted using hypotheses that are not always realistic.

Much of the developing world is not adequately covered by accurate vital registration systems, leading to a substantial lack of direct empirical data describing mortality. Because planning must still be undertaken, information on mortality is inferred through interpolation or extrapolation from existing commonly observed age patterns of mortality—the so-called model life tables. These model age patterns of mortality are used to substitute, extend, or fill in where observed mortality data are missing and are often key ingredients in methods used to estimate demographic indicators from sparse data. Underlying epidemiological profiles are inferred based on those that are known to underlie the existing model of mortality patterns that most closely match the observed data. The closest-fitting model patterns are then used, among other things, to estimate levels of child mortality and create population projections.

DSSs bridge the data gap that exists in resource-constrained countries. A DSS refers to a methodological approach for monitoring a registered, dynamic cohort of the total population of a geographically confined area. Typical DSS populations monitored include at least 60,000 individuals per site. This is usually sufficient to provide adequate sample sizes to monitor trends in all-cause mortality and the most common cause-specific mortality, by sex and age group. Members of the dynamic DSS cohort are registered as such during an initial census. New members enter the cohort either by system-registered births or in-migrations, and members exit either by system-registered deaths or by out-migration. Causes of death are determined for all deaths by verbal autopsy. Continuous cycles of reenumeration maintain an accurate person-time denominator and help identify numerator events such as pregnancies, births, deaths, and migration. These cycles of reenumeration typically take place three to four times per year in order to optimize the chance to identify pregnancies and thus determine outcomes of pregnancy. In addition to the routine data capture performed by specialist teams of enumerators and supervisors, larger numbers of community key-informants continuously notify the system of birth or death events for immediate follow-up. Only events occurring to registered households and members are linked to the database for analysis. Sophisticated field operational logistics and linked data quality control and computing systems support this routine DSS surveillance (MacLeod, Phillips, and Binka 1995; Phillips, MacLeod, and Pence 2000). DSS sites can generate accurate population structures, person-years at risk, mortality rates, cause-specific mortality rates, proportional mortality, fertility rates, and rates of internal and external migrations. In addition, because of the intensity of household survey frequency, DSS sites commonly document a large array of contextual information on household structures, dependency, occupation, education, access to and use of services, as well as socioeconomic status. A typical DSS produces more than 100 health- and poverty-monitoring indicators annually on each household. Details of DSS concepts and methods are provided by INDEPTH Network (2002).

On the completeness and reliability of data produced by DSS, Korenromp and colleagues (2003) wrote in their review paper, "For Africa, with an increasing number of sites that monitor cause-specific mortality at a population level by means of standard methods under the INDEPTH network, DSS may at present form one of the more complete data sources." And WHO/UNICEF (2003, 20) conclude in their Africa Malaria Report2003 that "[a]t present, the most reliable data on trends in malaria death in children under 5 years of age is obtained from demographic surveillance systems (DSS)."

In 1998 a large number of DSS sites formed an international network called the INDEPTH Network. Details of the sites, their locations, the populations they follow, and basic demographic outputs of each site are available on the Internet ( and in INDEPTH Network (2002). The central purpose of the INDEPTH Network is to facilitate cross-site analyses of comparable data across broader geographic areas with more diverse circumstances in order to answer questions that cannot be answered within individual sites. Comparative mortality and mortality patterns are an obvious first endeavor of such a network, and these are provided in this chapter. By the efforts of the members of the INDEPTH Network, for the first time schedules of mortality in Africa and empirical data on age standardization and on age and sex structure are available. This network is the source of the analyses presented in this chapter.

Standardized mortality rates were computed based on the new INDEPTH Network standard population that for the first time is derived from empirical data. The INDEPTH standard population typifies the true structure of the young population in Sub-Saharan Africa. The chapter then presents basic life table indicators for INDEPTH sites, based on their age-specific mortality rates over the 1995–99 period. Seven new mortality patterns are developed from more than 4.2 million person-years of observation at the African INDEPTH sites. These patterns are demonstrated to be substantially different from conventionally used model mortality patterns applied to Africa.

Levels of Mortality at INDEPTH Network Sites

The data used in this chapter come from 17 DSS sites in Sub-Saharan Africa for which information on mortality was available for at least a full year from 1995 to 1999 (table 7.1).

Table 7.1. Summary of Mortality Data from INDEPTH Sites, 1995–99.

Table 7.1

Summary of Mortality Data from INDEPTH Sites, 1995–99.

The overall average length of the observation period for the contributing sites is four years. The data yield a total of 3,979,155 person-years of exposure, during which 55,356 deaths occurred. An average of about 17 percent of the person-years exposed were lived at ages younger than five years, and an average of 39 percent of the deaths also occurred between birth and age five. The crude death rate for both sexes combined ranges from a low of 7 per 1,000 in Agincourt, South Africa, to 39 per 1,000 in Bandim, Guinea-Bissau.

Overall Mortality

The crude death rate and the expectation of life at birth are the two indicators used in this section to examine overall levels of mortality at the INDEPTH sites. In order to remove the influence of each site's age structure and to make the comparison of the crude death rates more reasonable, it is necessary to standardize such rates. To do this, there are several widely used standard age distributions, including Segi's world standard and World Health Organization (WHO) standard age distributions (Estève, Benhamou, and Raymond 1994). Both of these standards reflect populations with relatively low fertility and mortality. Consequently, they give significant weight to the middle years of life. All the INDEPTH sites record information from relatively young populations with comparatively high fertility and high mortality. Under those conditions, there are proportionally more young people in the population, giving it a "younger" age distribution. When the Segi or WHO standard age distributions are applied to the INDEPTH data, they give too much weight to the relatively high mortality rates prevailing at middle and older ages and too little weight to mortality at younger ages. Consequently, the absolute level of the age-standardized crude death rates produced with those standards significantly overestimates the true level of mortality at the INDEPTH Network sites.

To address this problem and create age-standardized crude death rates that more accurately reflect the true level of mortality at the INDEPTH sites, we have constructed the INDEPTH standard age distribution. An average age distribution is constructed for each site over the period 1995–99 by taking the weighted average of the person-years of exposure in each age group across all the years for which data are reported. The weight for each year is the total number of person-years reported for all ages during that year. The INDEPTH standard age distribution is calculated by taking the weighted average of the individual site average age distributions in each age group. In this case the weights are the total number of person-years in each of the individual site average age distributions. In figure 7.1, the younger age distribution of the INDEPTH standard, which is typical of developing countries, is contrasted with the much older population structures of the Segi and WHO standards.

Figure 7.1

Figure 7.1

Standard Population Age Structure from INDEPTH, Segi, and WHO Source: INDEPTH Network 2002.

Table 7.2 displays the crude death rate for each site and the age-standardized crude death rates calculated using both the INDEPTH and Segi standard age distributions along with the values for the expectation of life at birth.

Table 7.2. Crude Death Rates and Expectation of Life at Birth(per thousand).

Table 7.2

Crude Death Rates and Expectation of Life at Birth(per thousand).

The INDEPTH Network standardized crude death rates range from 7 to 33 per 1,000 for males and 5 to 27 per 1,000 for females, revealing a wide range of mortality at the INDEPTH sites. The figures for the expectation of life at birth vary in a relationship that is loosely inverse to the values of the crude death rate (figure 7.2), and they cover a similarly wide range of from 66 to 36 years for males and 74 to 40 years for females. The data from Bandim are anomalous and reflect some unresolved questions concerning the manner in which they were collected and reported.

Figure 7.2

Figure 7.2

Crude Death Rate and Expectation of Life at Birth Source: INDEPTH Network 2002.

There is some geographic clustering. Together at the low end of the spectrum are three rural sites from Tanzania (Hai, Ifakara, and Rufiji) and one site in Senegal (Mlomp). In the middle of the pack are three sites in West Africa: Farefenni, Nouna, and Oubritenga. At the high end there is a mixture of sites from West, East, and southern Africa. The absolute level of mortality varies considerably over space; sites located close to each other have similar levels of mortality, but all major regions of Africa show a wide range of mortality levels.

For the most part the sex differentials are relatively small but are generally in favor of females, as expected. Two of the sites in southern Africa—Agincourt, in South Africa, and Manhica, in Mozambique—have significant male migration and register more substantial sex differentials, which stand out in contrast to the rest of the sites. Bandim, West Africa, also records a substantial sex differential, but as noted above there may be a methodological explanation for this.

Infant and Child Mortality

The measures of child mortality displayed in table 7.3 are the life table probabilities of dying in a specified age group: 1q0 for ages zero to one year, 4q1 for ages one to five years, and 5q0 for ages zero to five years. The conventional infant mortality rate—the ratio of the number of deaths below one year of age and the number of births for a given period—is also included.

Table 7.3. Infant and Child Mortality(per thousand).

Table 7.3

Infant and Child Mortality(per thousand).

As shown in figure 7.3, there is a wide range in the level of child mortality. The probability that a newborn dies before reaching its fifth birthday ranges from 32 per 1,000 to 255 per 1,000 for males and 34 per 1,000 to 217 per 1,000 for females. The Agincourt site in South Africa records a comparatively low level of child mortality. Another cluster, composed of Mlomp in Senegal and Hai in Tanzania, reports low levels of child mortality but not nearly as low as that of the South Africa site. The next higher cluster is composed of sites from various regions of Africa including Dar es Salaam, Tanzania; Butajira, Ethiopia; Ifakara, Tanzania; Nouna, Burkina Faso; and Manhica, Mozambique. Following after that with 5q0 close to 175 per 1,000 for males and females are Farefenni, The Gambia; Rufiji, Tanzania; Navrongo, Ghana; Gwembe, Zambia; Morogoro, Tanzania; and Oubritenga, Burkina Faso. The three remaining sites—Niakhar, Senegal; Bandim, Guinea-Bissau; and Bandafassi, Senegal—all have substantially higher values of 5q0 closer to 225 per 1,000. There is a wide range in the level of child mortality, but except at the lowest and highest levels, there does not appear to be any geographical clustering. The lowest levels are definitely found in South Africa, whereas the highest levels are reported from West Africa.

Figure 7.3

Figure 7.3

Child Mortality: Probability of Dying between Birth and Age Five (5q0) Source: INDEPTH Network 2002.

Table 7.3 also displays the ratio of 1q0 to 4q1 in order to elucidate the changing risk of death faced by children before and after their first birthday.1 This ratio reveals that children in Rufiji who survive to age one face a probability of death that is improved by nearly a factor of four, whereas children in Bandafassi face a nearly constant probability of dying throughout the first five years of life.

Differences in child mortality according to sex are relatively small and do not appear to consistently favor one sex over the other. Interestingly this pattern is broken by four sites—Manhica, Mozambique; Rufiji, Tanzania; Niakhar, Senegal; and Bandafassi, Senegal—where there is a clear difference favoring females, except in Rufiji, where males are favored.

Adult Mortality

Values for the probability that persons who survive to their twentieth birthday will survive to their fiftieth birthday, that is, 30q20, are displayed in table 7.4 along with values of 5q0 and the ratio of 5q0 to 30q20.2 There are wide variations in the levels of adult mortality from 159 per 1,000 to 501 per 1,000 for males and 112 per 1,000 to 421 per 1,000 for females. A number of sites record substantial differences in adult mortality between the sexes—Agincourt, South Africa; Hai, Tanzania; Manhica, Mozambique; Mlomp, Senegal; and Navrongo, Ghana, in particular. There also exists the opposite differential, wherein female mortality rates exceed male rates, at two sites: Rufiji, Tanzania; and Dar es Salaam, Tanzania. The human immunodeficiency virus and acquired immune deficiency syndrome (HIV/AIDS) and maternal mortality may explain these patterns.

Table 7.4. Adult Mortality and Child and Adult Mortality Ratio(per thousand).

Table 7.4

Adult Mortality and Child and Adult Mortality Ratio(per thousand).

For the first time, Agincourt in South Africa does not define the low end of the range. There also does not appear to be any substantial geographical clustering of similar risk of adult mortality within the region. The cluster of moderate risk includes sites from all major regions of Sub-Saharan Africa as does the high risk cluster.

The relationship between child and adult mortality reveals two distinct groups: sites in West Africa and those in the rest of Africa. Some of the West African sites clearly record levels of child mortality that are high relative to the corresponding levels of adult mortality. In the West African sites of Bandafassi, Senegal; Farefenni, The Gambia; Oubritenga, Burkina Faso; and Niakhar, Senegal, this is the result of unusually high child mortality coupled with substantial adult mortality.

INDEPTH Network Mortality Patterns

This section summarizes the method used to identify the seven INDEPTH patterns of mortality. For a full description of the process, see INDEPTH Network 2002, chap. 7. Key considerations in identifying common underlying patterns (empirical regularities) are the filtering out of small, unimportant, and potentially random variation; the reduction of the dimensionality of the data to as few dimensions as possible; and the provision of a common reference to which all of the observed patterns can be compared.

Principal Components

The principal components technique is used to identify 15 age components that together are able to represent all but a negligible amount of the variation in age patterns of mortality among the 70 male and female site periods. In fact, the first four of these components represent the vast majority of the variation in the original data. When recombined with the appropriate weights the components are able to represent all the observed age patterns of mortality in the INDEPTH database, and furthermore, by choosing different weights any arbitrary age pattern of mortality can be represented to within a negligible error.

The first four principal components address all three of the key considerations important to identifying common underlying patterns: by dropping the other components much of the small-magnitude, random noise is eliminated; the dimensionality of the data is reduced to four weights instead of 18 age groups; and the four principal components provide a common reference against which all the observed patterns can be compared. The task then becomes how to identify similar observed patterns.


The aim of cluster analysis is to identify groups of similar objects. The objects are usually described by a vector of numbers representing measurements of some attributes of the objects. In this case, the vectors each contain four coefficients representing the weights on the first four principal components that are necessary to reproduce a given observed mortality pattern.3 The hierarchical clustering algorithm is used to identify five clusters of similar mortality patterns for males and seven for females. Because the seven female patterns are quite different from one another and because the male patterns corresponding to the two additional female patterns are grouped together in the male clusters, the seven female clusters are chosen as the final clusters of observed mortality patterns.

Once the clusters are identified the person-year weighted average of the individual age patterns in each cluster is calculated to yield the characteristic age pattern of mortality for each cluster.

The Seven INDEPTH Mortality Patterns

The seven commonly observed age patterns of mortality emerging from the INDEPTH data are sufficiently different from the existing model life tables (INDEPTH Network 2002) to qualify as new mortality patterns; and because almost all the data on which they are based come from Africa, they are African patterns appropriate for use in Africa. The natural log–transformed, ln(nqx), values are displayed in figure 7.4.

Figure 7.4

Figure 7.4

INDEPTH Mortality Patterns 1–7, ln(nqx) Source: INDEPTH Network 2002.

Underlying epidemiological profiles underpinning each pattern are still being identified. The INDEPTH Network is actively engaged in gathering and analyzing information on cause of death corresponding to the raw mortality rates presented here. Once that work is complete, it will be possible to identify the specific causes of death contributing to each pattern. However, until that time we must speculate and infer from what is known about the regional distribution of major causes of death in Africa.

The following discussion of the patterns has largely been excerpted from chapter 7 of Population, Health, and Survival at INDEPTH Sites (INDEPTH Network 2002).

Pattern 1

The first pattern is similar to the Coale-Demeny North and UN Latin American model life table age patterns of mortality (INDEPTH Network 2002). There is no indication that HIV/AIDS affects pattern 1, and the male and female age patterns are similar with the exception of a bulge in the female pattern during the reproductive years, presumably caused by maternal mortality. Pattern 1 is primarily derived from sites in West Africa over the entire period covered by the INDEPTH Network data set. HIV/AIDS has not yet become as significant a problem in West Africa as it is in central and southern Africa, so a large impact of AIDS in the data from West Africa is not expected. It is worth noting that child mortality between the ages of one and nine is significant and substantially elevated above the most similar existing models. This is in keeping with the fact that malaria is a significant cause of death in West Africa, and it has a large impact on those ages.

Pattern 2

The INDEPTH Network pattern 2 is derived largely from Asian data and for that reason is not discussed here; see INDEPTH Network 2002 for a discussion of pattern 2.

Pattern 3

The sites contributing to pattern 3 are almost exclusively located in southern Africa and East Africa, South Africa and Tanzania in particular. This pattern obviously contains some influence of HIV/AIDS but not nearly to the degree observed in pattern 5. The South African data come from the Agincourt site (INDEPTH Network 2002), where mortality is extraordinarily low compared with the other INDEPTH sites in Africa, and where HIV/AIDS is recognized but does not yet affect the population in the catastrophic sense that it does in other parts of southern and East Africa. The remainder of the data come from the Dar es Salaam site, where the impact of HIV/AIDS appears to be greater. This pattern is most similar to the UN Far East pattern of mortality, corresponding to the low infant and child mortality compared with mortality at older ages (INDEPTH Network 2002). A noteworthy feature of this pattern is that infant and child mortality does not appear to be substantially elevated, as might be expected when HIV/AIDS is an important contributor to mortality. Timaeus (1998) suggests that the strong secular decrease in child mortality that has been observed in many settings in Africa during the latter half of the twentieth century may largely cancel out the increase in child mortality brought about by HIV to yield approximately unchanging child mortality, rather than the increase that might be expected when HIV is having such an easily observed impact on adult mortality.

Pattern 4

Pattern 4 is a variation on pattern 1 with the important difference manifested in the 35 to 69 age range. At all other ages, patterns 1 and 4 are negligibly different except that infant and child mortality in pattern 4 is consistently slightly lower than pattern 1. But between ages 35 and, roughly, 69 pattern 4 reveals significantly higher mortality than pattern 1. This pattern is most similar to the UN general pattern for females and UN Latin American pattern for males (INDEPTH Network 2002). As was the case with pattern 1, most of the data contributing to pattern 4 comes from West Africa.

Pattern 5

The HIV/AIDS pattern of mortality is most clearly visible in pattern 5. The data contributing to pattern 5 are derived from the three Tanzanian sites run by the Adult Morbidity and Mortality Project (AMMP) in Dar es Salaam, Hai, and Morogoro (INDEPTH Network 2002). There is a striking bulge in the mortality of males between the ages of 20 and 54 and for females between the ages of 15 and 49. The female bulge is significantly narrower and more pronounced, corresponding to the earlier infection of the female population and within a tighter age range than the male population. This pattern is not particularly similar to any of the existing model patterns, but it is most closely matched with the UN General (female) and Latin American (male) model patterns (INDEPTH Network 2002). Pattern 5 differs from pattern 3 mainly in the shape of the HIV/AIDS impact. The effect is more diffuse with age in pattern 3, meaning that mortality is elevated through a broader age range, the magnitude of the elevation is more consistent, and the differences between the sexes are less apparent. Pattern 3 is derived largely from the Dar es Salaam data, and this may reflect the fact that the epidemic is more mature in Dar es Salaam and has consequently had enough time to infect a wider age range of people of both sexes. As with pattern 3, it is worth noting that infant and child mortality do not appear to be substantially affected in a manner comparable to adult mortality.

Pattern 6

Pattern 6 is one of the two additional patterns identified in the female data. It is an interesting pattern that reveals high mortality of children and teenagers and comparatively low mortality of infants and adults of all ages. This pattern is exhibited by sites in northeastern Africa and West Africa, with most of the data coming from Ethiopia. Without additional information, it is not possible to speculate on what may be producing this unique pattern. The male pattern is most similar to the Coale-Demeny North model pattern, and the female pattern is closest to the Coale-Demeny West model, both of which embody high mortality in the same age ranges (INDEPTH Network 2002). They deviate from those patterns in that infant mortality is substantially less than would be found in either model pattern, and child and adolescent mortality is significantly higher: this might be called the "Super North" pattern.

Pattern 7

Pattern 7 is the other additional pattern identified in the female data, and it, too, is interesting. It is derived from two sites located in central and West Africa. The reason it was identified in the female data is obvious; there is a substantial bulge in the female age pattern between ages 25 and 44. This pattern comes mainly from one small site in Zambia, so it is probably not worth extensive interpretation. The corresponding male pattern is similar to pattern 6, and both are similar to the Coale-Demeny North model pattern (INDEPTH Network 2002). The North model pattern contains relatively high child and teenage mortality coupled with comparatively low mortality at older ages. This is consistent with the fact that malaria is an important contributor to mortality in both sites.


In their 1996 review of demographic models, Coale and Trussell (1996) describe the identification of "empirical regularities" as one of the central motifs of demographic modeling. Their term accurately describes the formulation of model mortality schedules and hints strongly at the limits of their usefulness. Because there are no general, reliable, theory-driven models of the individual risk of death, we must rely on commonly observed patterns to guide us—empirical regularities. Because these are simply commonly observed patterns, however, we must be careful not to inappropriately overinterpret them or substitute them for a thorough understanding of the mechanisms underlying the risks of death that lead to each pattern. Beyond being the first step in understanding these mechanisms, identifying underlying empirical regularities in age patterns of mortality is still a useful exercise in its own right.

With fertility and to a lesser degree migration, mortality shapes the size and age composition of a population and is consequently of great interest to policy makers who must adequately plan for the future of a population. Mortality is one of the most valuable indicators of the health of a population and, as such, is of critical importance to those who manage health care infrastructures. As a sensitive indicator of where a population is and where it will be in the near future, knowledge of mortality is a vital component of population planning and management.

Existing systems of model life tables include systems developed by the Population Branch of the United Nations Department of Social Affairs (1955) in the early 1950s, by Coale and Demeny (1966) in the early 1960s, by Ledermann and Breas (1959) in the late 1950s, by Brass (1971) in the late 1960s, and finally by the UN again in the early 1980s (United Nations Department of International Economic and Social Affairs 1982). Only two are in common use today; the Coale and Demeny system and the 1982 UN set. However, neither of these systems was formulated with anything more than trivial data from Africa: apart from the West pattern that appears to contain a small amount of data from the white population of South Africa, none of the Coale and Demeny patterns is based on data from Africa; and apart from a small amount of data from Tunisia, neither are the recent UN patterns.

The lack of empirical data from Africa underlying either of the commonly used model mortality systems is a significant problem for demographers working in Africa, as Preston, Heuveline, and Guillot (2001) eloquently state: "A limitation, common to all the systems discussed so far [excluding the 1982 UN models], is that their empirical basis consists almost exclusively of the experience of developed countries. Most applications of model life tables, however, are addressed to developing countries with incomplete data."

Making this problem worse is the dissimilarity of the epidemiological profile of large parts of Africa to that of the developed world; it is clear that malaria and HIV are primarily to blame. Both increase overall levels of mortality and may affect certain age groups disproportionately. The significant endemic prevalence of these two diseases is responsible for vast excesses in mortality in some of the harder hit age groups; malaria for young children and HIV for children and young-to-middle-aged adults.

It is clear that the existing model mortality patterns are not appropriate for use in Africa, given that they were not built using data from Africa and that there is good reason to believe that the mortality profile of Africa is different from other parts of the world. Until now, however, there has been no alternative. It is to begin addressing this need that the INDEPTH Network has compiled a database of primarily African empirical mortality data and has begun to identify common age patterns of mortality described by those data (INDEPTH Network 2002).

DSS site data are sometimes challenged as being unrepresentative (Lopez et al. 2000) of surrounding constituencies due to the history of clinical and community-based intervention trials that are conducted in such populations from time to time. However, there is little evidence that the mortality patterns in trial sites are markedly different from those prevailing in similar settings nearby, where trials have not been conducted. The reason for the paucity of evidence is that most trials are conducted on a subset of the population, and they target a specific cause of mortality. Usually the interventions do not continue after the trial, unless they are adopted as broad implementation policy (for example, insecticide-treated nets), whereby the nontrial areas are, within a few years, at the same level of coverage for the intervention. We must also appreciate that our health system interventions have relatively weak modulating effects on population mortality compared with secular trends in socioeconomic circumstances and environmental forces. We should recognize that the modest population size of a DSS site does not constitute a major flaw, however, as even sites monitoring small-sized populations can produce robust measures of age-specific mortality when data are aggregated over a period of several years. Moreover, data collected over long periods of time from the same population living in the same area can reveal important age-specific trends in the risk of death. Furthermore, when data from a number of widely dispersed sites are brought together, they provide both a geographically and temporally representative measure of mortality conditions. At present, only DSS sites provide data that can be used to depict the temporal and geographical contours of mortality patterns in Africa.


  1. Alonso P. L., Lindsay S. W., Armstrong J. R., Conteh M., Hill A. G., David P. H., Fegan G. et al. The Effect of Insecticide-Treated Bed Nets on Mortality of Gambian Children. Lancet. 1991;337:1499–502. [PubMed: 1675368]
  2. Armstrong Schellenberg J. R., Abdulla S., Minja H., Nathan R., Mukasa O., Marchant T., Mponda H. et al. KINET: A Social Marketing Programme of Treated Nets and Net Treatment for Malaria Control in Tanzania, with Evaluation of Child Health and Long-Term Survival. Transactions of the Royal Society of Tropical Medicine and Hygiene. 1999;93:225–31. [PubMed: 10492745]
  3. Binka F. N., Kubaje A., Adjuik M., Williams L. A., Lengeler C., Maude G. H., Armah G. E. et al. Impact of Permethrin Impregnated Bednets on Child Mortality in Kassena-Nankana District, Ghana: A Randomized Controlled Trial. Tropical Medicine and International Health. 1996;1(2):147–54. [PubMed: 8665378]
  4. Brass, W. 1971. "On the Scale of Morality." In Biological Aspects of Demography, ed. W. Brass. London: Taylor and Francis.
  5. Brass, W., A. J. Coale, P. Demeny, D. F. Heisel, F. Lorimer, A. Romaniuk, and E. van de Walle. 1973. The Demography of Tropical Africa. Princeton, NJ: Princeton University Press.
  6. Breman J. G., Egan A., Keusch G. T. The Intolerable Burden of Malaria: A New Look at the Numbers. American Journal of Tropical Medicine and Hygiene. 2002;64(1, 2):iv–vii. [PubMed: 11425185]
  7. Coale, A., and P. Demeny. 1966. Regional Model Life Tables and Stable Populations. Princeton, NJ: Princeton University Press.
  8. Coale A., Trussell J. The Development and Use of Demographic Models. Population Studies. 1996;50(3):469–84. [PubMed: 11618377]
  9. de Savigny, D., P. Setel, H. Kasale, D. Whiting, G. Reid, H. M. Kitange, C. Mbuya, L. Mgalula, H. Machibya, and P. Kilima. 1999. Linking Demographic Surveillance and Health Service Needs—the AMMP/TEHIP Experience in Tanzania. Dar es Salaam.
  10. Estève, J., E. Benhamou, and L. Raymond. 1994. Statistical Methods in Cancer Research. Vol. 4, Descriptive Epidemiology. Lyons: IARC.
  11. INDEPTH Network. 2002. Population, Health and Survival at INDEPTH Sites. Vol. 1 of Population and Health in Developing Countries. Ottawa: International Development Research Centre.
  12. Korenromp E. L., Williams B. G., Gouws E., Dye C., Snow R. W. Measurement of Trends in Childhood Malaria Mortality in Africa: An Assessment of Progress Toward Targets Based on Verbal Autopsy. Lancet. 2003;3:349–58. [PubMed: 12781507]
  13. Ledermann S., Breas J. Les dimension de la mortalité Population. 1959;14(4):637–82.
  14. Lopez, A., J. H. Salomon, O. B. Ahmad, C. J. L. Murray, and D. Mafat. 2000. "Life Tables for 191 Countries: Data, Methods, Results." GPE Discussion Paper 9, WHO, Geneva.
  15. MacLeod, B., J. F. Phillips, and F. Binka. 1995. Analysis of Longitudinal Data from the HRS. New York: Population Council.
  16. Nevill C. G., Some E. S., Mung'ala V. O., Mutemi W., New L., Marsh K., Lengeler C., Snow R. W. Insecticide-Treated Bednets Reduce Mortality and Severe Morbidity from Malaria among Children on the Kenyan Coast. Tropical Medicine and International Health. 1996;1(2):139–280. [PubMed: 8665377]
  17. Phillips J. F., MacLeod B., Pence B. The Household Registration System: Computer Software for Rapid Dissemination of Demographic Surveillance Systems. Demographic Research. 2000;2(6) [PubMed: 12178153]
  18. Preston, S. H., P. Heuveline, and M. Guillot. 2001. "Modeling Age Patterns of Vital Events." In Demography: Measuring and Modeling Population Processes, 191–210. Oxford: Blackwell.
  19. Ross D. A., Kirkwood B. R., Binka F. N., Arthur P., Dollimore N., Morris S. S., Shier R. P., Gyapong J. O., Smith P. G. Child Morbidity and Mortality Following Vitamin A Supplementation in Ghana: Time Since Dosing, Number of Doses, and Time of Year. American Journal of Public Health. 1995;85(9):1246–51. [PMC free article: PMC1615567] [PubMed: 7661232]
  20. Tanzania Ministry of Health. 1997. "Policy Implications of Adult Morbidity and Mortality—End of Phase 1 Report." In Adult Morbidity and Mortality Project. Dar es Salaam: Government of Tanzania.
  21. Timaeus I. M. Impact of the HIV Epidemic on Mortality in Sub-Saharan Africa: Evidence from National Surveys and Censuses. AIDS. 1998;12(Suppl. 1):S15–27. [PubMed: 9677186]
  22. United Nations Department of Social Affairs. 1955. Age and Sex Patterns of Mortality: Model Life Tables for Underdeveloped Countries. New York: United Nations.
  23. United Nations Department of International Economic and Social Affairs. 1982. Model Life Tables for Developing Countries. Population Studies 77. New York: United Nations.
  24. WHO/UNICEF (World Health Organization/United Nations Children's Fund). 2003. The Africa Malaria Report—2003. Geneva: WHO and UNICEF.



1q0 represents the probability of dying before age one for newborns; 4q1 represents the probability of dying before age five for those who survive to age one. In general, nqx is the probability of dying between the ages of x and x + n for those who survive to age x.


. 5q0 represents the probability of dying before age five for newborns.


. See INDEPTH Network 2002 for a discussion of why the age pattern and level of mortality are orthogonal in the principal components model. This allows the clustering technique to identify clusters based on age pattern only, regardless of the level of mortality; essentially the principal components model controls for level.

Copyright © 2006, The International Bank for Reconstruction and Development/The World Bank.
Bookshelf ID: NBK2300PMID: 21290661


  • 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...