Sources of Data

Vital registration systems are the ideal source of data with which to monitor mortality levels and trends. However, the statistics systems of most Sub-Saharan Africa countries continue to be underdeveloped; few countries are able to provide adequate vital statistics. In 2001 the World Health Organization (WHO) contacted all member states in Sub-Saharan Africa to try to obtain death statistics for a global assessment of the burden of disease (Kowal, Rao, and Mathers 2003). Just 9 out of 46 member states provided such data. Furthermore, coverage of registration was less than 60 percent in most of these 9 countries, with only Mauritius and the Seychelles providing vital registration data that were more than 90 percent complete. These two island states make up a tiny fraction of the population of Sub-Saharan Africa. Adequate vital statistics are far more widespread in other regions, particularly Europe and the Americas, and 60 percent of the countries for which the WHO could not obtain registration data are in Sub-Saharan Africa. Data series of any time length are particularly rare. Due to a lack of appropriate direct data, the Adult Mortality in Developing Countries project, which studied trends in adult mortality in 27 developing countries (Hill 2003), included only 2 countries from mainland Sub-Saharan Africa (Benin and Zimbabwe). Among the mainland countries, South Africa is an exemplary country because of the government's determined efforts in the last decade to improve its statistics. The lack of reliable mortality data had been dubbed the "black hole of vital statistics" (Botha and Bradshaw 1985), but national efforts during the 1990s resulted in over 90 percent of the adult deaths being registered (Dorrington et al. 2001). Yet, even in South Africa, the production of timely cause-of-death statistics remains a challenge.

The pervasive lack of vital registration data makes it necessary to derive estimates of mortality using indirect demographic techniques based on survey and census data. The most readily accessible data that can be used to estimate adult mortality come from the Demographic and Health Surveys (DHS). At least one such survey has been conducted in most African countries, and many of the African surveys include sibling histories that were developed to obtain estimates of maternal mortality (Rutenberg and Sullivan 1991). Unfortunately, almost all DHS samples are too small to enable the study of age-specific mortality in adulthood or mortality trends in any detail. Moreover, concerns exist about both the reliability and validity of reports on the survival of respondents' siblings. Although Bicego (1997) found good correspondence with other data, Stanton, Abderrahim, and Hill (1997) suggest that sibling survival may progressively underestimate adult mortality as the time since the event lengthens. These authors conclude that sibling histories can at best provide an estimate of adult mortality in the few years before the data were collected. Comparison of successive sets of sibling history data in the three African countries that have collected them in two DHS surveys also supports this conclusion (Timaeus and Jasseh 2004). Apart from the DHS surveys, several other international programs of surveys have been conducted in Sub-Saharan Africa during the past decade. However, although the Multiple Indicator Cluster Surveys (MICS) conducted by UNICEF are an important source of child mortality estimates, they have not included questions on adult mortality. Moreover, data from the World Health Survey have yet to become available.

Data on adult mortality can be collected in national censuses either by including questions about recent deaths in households or by asking about the survival of the parents of household members. Only some African countries have a tradition of asking one or both sets of questions in their censuses, and few additional countries responded to the growing concern about adult mortality from AIDS and other causes by adding them to their schedule for the 2000 census. Even fewer countries have yet published these data. Although reporting of recent deaths can often be incomplete and subject to reference period and age reporting errors, a range of methods exists for the evaluation of such information and, in favorable circumstances, the correction of inadequate data (Bennett and Horiuchi 1984; Brass 1975; Hill 1987). Data on the survival of parents can be used to estimate conventional life table indexes by means of the orphanhood method (Brass and Hill 1973; Timaeus 1992). The method can be used to estimate the historical trend in mortality and has been refined in various ways in order to produce more up-to-date estimates (for example, Timaeus 1991). Moreover, adjustments can be made to orphanhood data to correct for the biases introduced by the excess mortality of orphans who have been vertically infected with HIV (Timaeus and Nunn 1997).

Two consecutive censuses can also be used to estimate adult mortality from intercensal cohort survival or age-specific growth (Preston and Bennett 1983). The ratio resulting from matching age groups in the second population to the first population provides a summary measure of adult mortality. This requires that the population be closed to migration and that the census coverage be the same in both censuses. Few, if any, Sub-Saharan Africa countries come close to meeting either of these conditions. Thus, the method has not been used much or with great success.

Demographic surveillance of local populations has been used for many years in a handful of African populations to study mortality and provide a framework for epidemiological research and the trial of health interventions (for example, Pison and Langaney 1985). In response to the spread of HIV in Africa, many more such longitudinal surveillance systems have been established recently in which vital events are monitored at regular intervals (Gregson et al. 1997; Hosegood, Vanneste, and Timaeus 2004; Nunn et al. 1997; Sewankambo et al. 2000; Todd et al. 1997; Tollman et al. 1999; Urassa et al. 2001). Almost all these sites, however, cover only small and select populations. Therefore, they are of little use for the estimation of adult mortality in national populations. The partial exception is the surveillance system organized by the Adult Morbidity and Mortality Project (AMMP) in Tanzania, which covers three districts that include both rural and urban areas, although it is not statistically representative of the population of the country as a whole (Kitange et al. 1996).

Because of the shortage of reliable empirical estimates, World Population Prospects (United Nations 2003) uses estimates of child mortality and assumptions about the age pattern of mortality taken from a family of model life tables to determine life expectancy in all of mainland Sub-Saharan Africa. The system of model life tables adopted by the United Nations (UN) Population Division implicitly defines the level of adult mortality. In most of these countries the Population Division simply has assumed that non-AIDS mortality at all ages conforms to a Princeton North model life table. These models are derived from historical data largely on northern Europe (Coale, Demeny, and Vaughan 1983) and are fitted to an estimate of under-five mortality (United Nations 2002). In a few countries, Princeton West models (Coale, Demeny, and Vaughan 1983) or UN Far Eastern models (United Nations 1982) have been used in the same way. Moreover, Sub-Saharan Africa is the only area in which data are completely lacking for a few countries and mortality in them has to be guessed from estimates for neighboring countries. The UN Population Division bases its estimates of AIDS deaths in high-prevalence African countries on those produced by the Joint United Nations Programme on HIV/AIDS (UNAIDS) using epidemiological models fitted to HIV seroprevalence data collected in antenatal clinics (UNAIDS Reference Group 2002). Although these models are based on careful review of the data coming out of a range of research studies, at present no attempt is made by any of the UN agencies to use empirical data on mortality in national populations to estimate AIDS deaths in particular countries.

The WHO reviewed the various estimates and observed substantial variations, depending on the different procedures and judgments made (Lopez et al. 2002). It produced its own set of life table estimates for 191 countries in 2000 based on its own mortality database and a new family of model life tables, which in turn were based on an extension of the Brass logit system (Murray et al. 2003). Although the basis of these life tables was measures of mortality from direct registration wherever possible, the only mainland Sub-Saharan Africa countries for which such data were adopted were Zimbabwe and South Africa and in addition, for Tanzania, where the AMMP surveillance data for part of the country were used.

For the rest of Africa, the WHO, like the UN Population Division, was forced to estimate adult mortality and life expectancy by extrapolation from estimates of child mortality. It did this by modeling the relationship between adult and child mortality in its database of more than 1,800 life tables, using its modified logit life table system. Given the lack of such data, however, this database includes few life tables from Sub-Saharan Africa and few with levels of child mortality as high as those that are typical of the African region. Thus, although the WHO's estimates are based on models developed using contemporary developing country data, there is no guarantee that they will be more accurate for Africa than those made by the UN Population Division using model life tables based on historical data on high-mortality Western populations. Moreover, like the Population Division, the WHO was unable to locate any up-to-date mortality data on some eight African countries. To produce the final life tables, mortality levels were estimated without HIV/AIDS, and then estimates of the additional adult AIDS deaths were added, based on the impact observed in the direct estimates from the demographic surveillance sites in Tanzania and the national vital registration systems of Zimbabwe and South Africa (Salomon and Murray 2001).

Mortality Levels

Table 4.1 shows estimates of adult mortality based on the WHO life tables for Africa in 2000 (Lopez et al. 2002). The index presented is the probability of dying between exact ages 15 and 60 (₄₅q₁₅). It refers to a particular year but can be interpreted as the probability that someone who had survived childhood would die before old age if he or she went through life subject to the age-specific death rates of the year in question. This measure has been adopted by many international agencies, including the WHO, as a good indicator of the overall level of adult mortality. Unlike life expectancy at age 15, it is not influenced by death rates at age 60 and over, which are very difficult to measure in countries with defective vital statistics systems. One limitation of the measure is that it is heavily influenced by death rates in the upper part of the 45-year age range. This reduces its sensitivity to trends in death rates in early adulthood. In addition, without supplementary data or assumptions, neither sibling history data, which refer mainly to young adults, nor orphanhood data, which refer largely to middle-aged parents, can provide robust measures of adult mortality across the whole of this wide age range.

Table 4.1

Probabilities of Dying between Ages 15 and 60 (₄₅q₁₅) in Sub-Saharan Africa.

In its estimates of mortality, the WHO found that of the 40 countries with the highest mortality, 37 were from the Sub-Saharan region. The level of adult mortality is highly variable across African countries. Southern and Eastern Africa have particularly high adult mortality, whereas mortality in Western Africa is lower and the Indian Ocean Islands, which accommodate relatively small populations, have the lowest rates. According to these WHO estimates, the probability of surviving from exact age 15 to exact age 60 in 2000 was less than 50 percent in nearly half of the countries in Sub-Saharan Africa. Most of these are countries affected severely by the HIV/AIDS epidemic, but the list also includes war-torn countries, such as Sierra Leone. The estimates suggest that adult men have consistently higher mortality than adult women in Sub-Saharan Africa, with the female advantage in survivorship being smallest in the very high mortality countries and largest in the low-mortality island states.

The estimates of adult mortality produced by the WHO for 2000 and those made by the United Nations Population Division for the period 1995–2000 are compared in figure 4.1. It is clear that the WHO estimates tend to be significantly higher than those issued by the Population Division for both men and women. The discrepancies are especially large in several countries in the Southern African region, but are also significant for men and of borderline significance for women (p = 0.059) in the other countries of the region. As much of Africa is experiencing rising adult mortality from HIV/AIDS, one would expect the WHO's 2000 estimates to be slightly higher than those made by the Population Division for a period on average 2.5 years earlier. However, HIV/AIDS can explain only a small part of the discrepancies between the two sets of estimates. Instead, they are rooted in part in differences in what is assumed about the level of child mortality by the two organizations, in part in the use of different model life tables to extrapolate from child to adult mortality, and in part in different assumptions about the scale of AIDS mortality in particular countries. One cannot readily quantify the relative importance of these contributions to the scatter revealed by figure 4.1. It seems likely, however, that the large discrepancies between the estimates for some southern African countries result at least in part from different assumptions about the severity of AIDS mortality in them.

Figure 4.1

Probabilities of Dying between Exact Ages 15 and 60 in Sub-Saharan Africa, by Sex Source: Authors' calculations from United Nations 2003 and Lopez et al. 2002.

Trends in Adult Mortality

There is growing evidence of rising trends in adult mortality in the countries in Sub-Saharan Africa. The eastern and southern African regions have been particularly hard hit by the AIDS epidemic, and the available data show large increases in adult mortality rates. For example, data from Malawi show that adult mortality was declining until the mid-1980s but that this trend was reversed in the 1990s (table 4.2). These results are based on the intercensal survival method but are broadly consistent with estimates based on recent household deaths and DHS sibling history data (Blacker 2004). They suggest that mortality for men and women were at similar levels, rather than women having lower mortality, but long-standing evidence exists to suggest that Malawi may be exceptional in this regard (Timaeus 1993, 1998).

Table 4.2

Probabilities of Dying between Ages 15 and 60 (₄₅q₁₅), by Sex, according to Intercensal Survival, Malawi.

Analyses of data from censuses in Kenya also suggest that a reversal in the trend in adult mortality occurred in that country during the 1990s (table 4.3). The orphanhood estimates of person-years lived indicate that slight improvements were made between the 1970s and 1980s but that an even larger decline occurred in the next decade. In this country, the mortality of women is lower than that of men.

Table 4.3

Orphanhood Estimates of Person-Years Lived and Lost at Ages 35–75 (Men) and 25–75 (Women), Kenya.

Analysis of a range of data sources from Zimbabwe, including vital registration data, has also shown a marked rise in the adult mortality rates during the 1990s (Feeney 2001). This is in keeping with the rising prevalence of HIV observed in antenatal clinics during the late 1980s and early 1990s. Trends in the probability of dying between ages 30 and 65, ₃₅q₃₀, are shown in figure 4.2. The estimates from vital registration, household deaths, and survival of parents are reasonably consistent, although the last of these series may underestimate somewhat for the earlier period and place the subsequent rise in mortality slightly too early. Mortality probably fell during the 1970s and early 1980s, but age-specific death rates and probabilities of death subsequently rose by 200 to 300 percent between the late 1980s and late 1990s for both adult men and adult women.

Figure 4.2

Conditional Probabilities of Dying between Ages 30 and 65 in Zimbabwe, by Sex, from Different Data Sources, 1969–90 Source: Feeney 2001.

Analysis of a range of data sources from South Africa that also include vital registration data shows that adult mortality rates were fairly constant during the 1980s, followed by an increase in the late 1990s (Timaeus et al. 2001). Trends in ₄₅q₁₅, the probability of surviving from 15 to 60, are shown in figure 4.3 and reveal the initial signs of the impact of the major HIV epidemic that emerged during the 1990s. More recent data collated from the national population register show that there has been a continued rise in the number of adult deaths registered in South Africa (figure 4.4) and the increase has followed the distinct age pattern associated with AIDS. However, improved registration of deaths and improved registration on the population register could account for some of this increase. Taking into account these improvements as well as population growth, the increase in deaths of persons at ages 15 years or more has been approximately 40 percent. Age-specific rates can be calculated for South Africa from these data by adjusting for underregistration of deaths (Dorrington, Moultrie, and Timaeus 2004). This can only be done for the years up to 2001, the date of the most recent census, which provides an estimate of the population against which to assess the underregistration. It confirms that a large rise in the mortality of young adults has occurred. The relative increase in mortality rates has been greater for women than for men, although starting from a lower base. Also, the increase for women occurs at a younger age than that for men.

Figure 4.3

Conditional Probabilities of Dying between Ages 15 and 60 in South Africa, by Sex, from Different Data Sources, 1980–2000 Source: Timaeus et al. 2001

Figure 4.4

Age Distribution of Reported Adult Deaths on the National Population Register of South Africa, by Sex, 1998–2003 Source: Bradshaw et al. 2004.

Comparison of the empirically based estimates presented in figure 4.3 with those in table 4.1 suggest that the modeling procedure used by the WHO to estimate adult mortality in South Africa in 2000 has produced severe overestimates, particularly for women. This highlights the uncertainty attached to estimates of adult mortality made from data on children and to projections using epidemiological models fitted to HIV data. It may be the case that the WHO also overestimated women's mortality in the other southern African countries for which it published much higher estimates than the UN (figure 4.1). However, the estimates in table 4.1 are not consistently exaggerated. For example, the estimates for Zimbabwe produced by the WHO are broadly consistent with the most up-to-date empirical estimates shown in figure 4.2.

Although the rise in adult mortality in South Africa by 2000 remained moderate at the national level, adult mortality had already risen far more in some parts of the country. Data from a demographic surveillance site in a rural area of one of the provinces that has been most severely affected by HIV/AIDS show that a massive rise in adult mortality occurred between the mid-1990s and 2000. By the latter year the probability of dying between exact ages 15 and 60 reached 58 percent for women and 70 percent for men (Hosegood, Vanneste, and Timaeus 2004). Moreover, this site collects information on causes of death using verbal autopsies. As they are based on a review of symptoms, these data are less liable than the official statistics to attribute AIDS deaths to other causes. They show that the huge rise in adult mortality, which is concentrated among young adults, as in the national statistics, can be accounted for entirely by AIDS deaths.

Church records from northern Namibia have been used to estimate the mortality experience of the parishioners (Notkola, Timaeus, and Siiskonen 2004). Adult mortality did not change much during the 1980s but increased rapidly from about 1994. In 2000 the death rates of women at ages 20–64 were 3.5 times higher than in 1993; for men they were 2.5 times higher.

An analysis of the sibling histories collected prior to 1997 in 11 DHS surveys in Sub-Saharan Africa showed that, although adult mortality was falling or stagnant in Western Africa and in Namibia in the 1980s, it had begun to rise sharply in Eastern Africa (Timaeus 1998, 1999). Moreover, four of the six Eastern African countries considered were characterized by unusually high mortality of young adults relative to older adults. Unfortunately, DHS surveys collect sibling histories only from women of reproductive age (although respondents report on siblings older than themselves) and the surveys cover rather small samples for the study of adult mortality, especially at ages 45–59. Thus, these surveys lack the statistical power to enable one to produce meaningful estimates of how the age pattern of mortality in adulthood is evolving in each country as the overall death rate rises. Timaeus and Jasseh (2004) incorporate the results of more recent DHS surveys into an analysis based on 26 studies in 23 countries. They attempt to sidestep the relatively small size of DHS samples by estimating a common age pattern of mortality increase across all the countries in which HIV has become present while determining the size of that mortality increase separately for each country. Their smoothed summary estimates of the level and trend in the probability of dying between ages 15 and 60 are shown in table 4.4 and by the WHO region in figure 4.5, which also plots the WHO's own estimates of the same index for the year 2000 (Lopez et al. 2002).

Table 4.4

Estimates of the Probability of Dying between Ages 15 and 60(₄₅q₁₅), from DHS Sibling Histories (per person).

Figure 4.5

Trend in the Estimates of the Probabilities of Dying between Ages 15 and 60 (₄₅q₁₅), by Sex and WHO Region, from DHS Sibling Histories (per 1,000 people) Source: Timaeus and Jasseh 2004.

The results of the analysis by Timaeus and Jasseh (2004) show that the fastest rises in adult mortality have occurred in South Africa, Zimbabwe, Zambia, Uganda, Guinea, and Cameroon. Adult mortality has risen relatively slowly or continued to fall in the Sahel. A significant change in the trend about four years after the development of a generalized HIV/AIDS epidemic is observed in 16 of the 19 countries in which the sibling histories cover the relevant period. Of the three countries in which HIV has become prevalent but mortality has not risen markedly, the data for Nigeria are known to be of poor quality, but it is unclear why Ethiopia and Rwanda do not conform to the usual pattern. The increase in mortality is concentrated among women age 25 to 39 and men age 30 to 44. On average, men's death rates have risen by about a third more than those of women. However, as women are dying at younger ages and African populations have grown rapidly during the last few decades, the sex differential in the number of AIDS deaths is small.

Extrapolation of the trend in the sibling history estimates forward to 2000 and comparison of them with those made by the WHO (also shown in figure 4.5) reveal that they are in quite good agreement for some countries, including high mortality countries, such as Malawi, Zambia, and Zimbabwe. For men, the WHO estimates are much higher than predicted by the sibling history estimates in Mozambique and in countries where the prevalence of HIV infection is fairly low, that is to say Chad and western African countries such as Benin, Burkina Faso, and Mali. For women, the discrepancies between the two series are larger and more widespread. They extend to South Africa, the eastern African countries and, to a lesser extent, Côte d'Ivoire, Guinea, and Togo in western Africa. These discrepancies might result from severe underreporting of dead siblings in the DHS. However, it is unclear why this should be more serious in lower mortality countries and, in some cases, for sisters alone. Alternatively, the UN agencies may tend to overestimate adult mortality in Africa. An analysis comparing the numbers of orphans implied in the estimates by the Population Division with empirical data gathered by the DHS on the proportions of children orphaned provides evidence in support of the latter interpretation (Grassly et al. 2004). The DHS data find far fewer orphans than are predicted by models based on the Population Division and UNAIDS estimates of mortality. Just as in the comparison of the sibling history estimates with the WHO estimates presented here, these discrepancies tend to be larger for maternal orphans than paternal orphans and larger in low-mortality countries than in high-mortality countries. This pattern of discrepancies suggests that, rather than the problem being underreporting of orphanhood in DHS or exaggeration of the scale of AIDS mortality, the model life table systems used by the UN agencies may systematically overestimate background adult mortality in Sub-Saharan Africa, especially for women.

Conclusion

The lack of good quality vital registration data in Sub-Saharan Africa has resulted in uncertainty in the levels of adult mortality. Estimates of adult mortality vary, depending on the data source, the methodology used, and the assumptions made. Despite the uncertainties in the data, the evidence shows that adult mortality rates are generally high, reflecting poor levels of health in the region. By the mid-1990s the levels of adult mortality do appear to have become far more varied among the regions and countries and between the genders than they had been a decade earlier. Many countries have shown an increase, with death rates reaching unprecedented heights. This is in stark contrast to the findings by Hill (2003) that adult mortality has been declining in most developing countries—at a rate of 1 percent per year for men and 2 percent per year for women. Estimates of adult mortality for some countries in Sub-Saharan Africa are by far the highest in the world.

Because of the inadequacy of vital statistics in Sub-Saharan Africa, both the UN Population Division and the WHO estimate adult mortality and life expectancy in the countries of the region by extrapolating from estimates of child mortality using model life tables and adding in an estimate of AIDS deaths based on antenatal clinic data. Enough is known about adult mortality in Africa to make it clear that age patterns of mortality vary markedly between different populations in the region. The assumption that a fixed relationship between adult and child mortality holds across the continent is bound to produce estimates that are badly in error in particular countries, even if they are unbiased overall. In addition, evidence from DHS orphanhood data (Grassly et al. 2004), DHS sibling histories (Timaeus and Jasseh 2004), and earlier surveys and censuses (Timaeus 1993) all suggest that in recent years the Population Division's extrapolations using Princeton North model life tables have overestimated non-AIDS adult mortality in large parts of the region. The WHO's estimates infer the relationship between child and adult mortality from a database of populations that tend to have lower mortality than is typical of Africa and appear to suffer from very similar limitations to those produced by the Population Division.

Given the limitations of the demographic and epidemiological data, there are inevitably large errors in some of the estimates of AIDS mortality made by UNAIDS and the WHO. Nevertheless, the evidence that adult mortality is now rising in most of Sub-Saharan Africa is clear cut. Moreover, the empirical evidence as to the size and speed of the rise in adult mortality in different countries in the region and the evolving age pattern of the mortality impact is consistent with predictions from epidemiological surveillance data and models of the impact of HIV/AIDS on adult mortality in most countries. The annual toll of AIDS deaths in Sub-Saharan Africa continues to grow rapidly, and the urgency of responding to the poor health conditions in the region has been exacerbated by the emergence of the HIV/AIDS epidemic.

In some African countries significant numbers of people have been dying from AIDS since the 1980s. In such countries it is difficult to imagine what would have happened to rates of mortality by today in the absence of AIDS. Thus, it is no longer viable to estimate and forecast mortality by adding an estimate of AIDS deaths to a separate estimate of non-AIDS mortality. Instead, new approaches for making mortality estimates and projections are required that can be calibrated to data on both the prevalence of HIV infection and all-cause adult mortality.

While existing data do signal the broad scale of the impact of AIDS on the mortality of adult populations in Sub-Saharan Africa, they are clearly not sufficiently detailed, reliable, or timely to provide the description of changes in level and age pattern that is urgently needed to monitor the impacts of programs that are being funded by African governments, the Global Fund, the World Bank, and bilateral donors. The fundamental cause of the uncertainty about levels of adult mortality and the impact on them of AIDS in Africa remains the limitations of epidemiological monitoring and demographic data systems in the region. The rollout of antiretroviral care and treatment activities throughout the continent should have a major impact (short, middle, and long term) on mortality dynamics and the demographic landscape of the region. More than ever before, better epidemiological and demographic data are required to evaluate the impact of health programs and to assess whether these efforts are effective in reducing the mortality of young adults.

National statistical offices, technical advisers, donors, and researchers are urged to assign more priority to the collection of adult mortality data in this region. In addition to responding to the poor levels of adult health, reinvigorating vital registration and statistics is essential in the long term and must also be placed high on the health and development agenda. An effective registration system is the only way of producing detailed annual series of mortality statistics, including those on causes of death. Moreover, as countries in Sub-Saharan Africa experience the phenomenon of aging, improved vital registration systems will be critical to provide information about mortality trends at older ages, which are largely unknown at present and difficult to investigate using methods based on survey data. Thus, countries are encouraged to establish expert teams to critically review their national civil registration systems in regard to their legal framework, organizational issues, system design, training needs, and quality control issues and to implement strategies for their improvement.

Given the stresses on governments in the region (in part due to the impacts we are trying to measure), it is also vital to propose ways in which survey data, census data, and demographic surveillance data might be expanded and better used to track and understand changes in mortality while vital registration still remains defective. Demographic and epidemiological surveillance in localized sites has proved invaluable and can yield relatively reliable statistics on causes of death. However, it needs to be balanced by other sources that can provide nationally representative data. Although sibling history data have serious limitations, in particular their limited statistical precision in all but the largest surveys, they have proved useful. We believe that countries should continue to collect these data. Moreover, they should collect them repeatedly, as this greatly increases the scope for assessment of the quality of the statistics that result. However, the crucial source of demographic data for planning, particularly at the subnational or district level, is the population census. All African governments are faced with either the reality or the threat of massive rises in adult mortality and they should all include questions about household deaths by age and sex and about orphanhood in their 2010-round censuses. A major drive is needed on the part of the United Nations agencies and donor organizations to support African governments in this, so as to ensure that question and form design, enumerator training, data capture, and the analysis of these data are all conducted successfully.

References

1. Bennett N., Horiuchi S. Mortality Estimation from Registered Deaths in Less Developed Countries. Demography. 1984;21:217–33. [PubMed: 6734860]
2. Bicego G. Estimating Adult Mortality Rates in the Context of the AIDS Epidemic in Sub-Saharan Africa: Analysis of DHS Sibling Histories. Health Transition Review. 1997;7(Suppl. 2):7–22.
3. Blacker J. The Impact of AIDS on Adult Mortality: Evidence from National and Regional Statistics. AIDS. 2004;(Suppl. 2):S10–26. [PubMed: 15319740]
4. Blacker, J., P. Kizito, and B. Obonyo. 2003. "Projecting Kenya's Mortality: Using Spectrum to Project the AIDS and Non-AIDS Components." Paper presented at the Conference on Empirical Evidence for the Demographic and Socioeconomic Impact of AIDS, Durban.
5. Botha J. L., Bradshaw D. African Vital Statistics—A Black Hole? South African Medical Journal. 1985;67:977–81. [PubMed: 4002089]
6. Bradshaw D., Laubscher R., Dorrington R., Bourne D., Timaeus I. M. Unabated Rise in Number of Adult Deaths in South Africa. South African Medical Journal. 2004;94(4):278–79. [PubMed: 15150940]
7. Brass, W. 1975. Methods for Estimating Fertility and Mortality from Limited and Defective Data. Chapel Hill, NC: International Program of Laboratories for Population Statistics.
8. Brass, W., and K. H. Hill. 1973. "Estimating Adult Mortality from Orphanhood." In International Population Conference, Liège, 1973. Vol. 3. Liège: International Union for the Scientific Study of Population.
9. Coale, A. J., P. Demeny, and B. Vaughan. 1983. Regional Model Life Tables and Stable Populations. London: Academic Press.
10. Dorrington, R. E., D. Bourne, D. Bradshaw, R. Laubscher, and I. M. Timaeus. 2001. The Impact of HIV/AIDS on Adult Mortality in South Africa. Cape Town: Medical Research Council. http://www​.mrc/bod/bod.htm.
11. Dorrington, R., T. A. Moultrie, and I. M. Timaeus. 2004. Estimation of Mortality Using the South African Census 2001 Data. Monograph 11. Cape Town, South Africa: Centre for Actuarial Research, University of Cape Town. http://www​.commerce.uct​.ac.za/care/Monographs​/Monographs/Mono11.pdf.
12. Feeney G. The Impact of HIV/AIDS on Adult Mortality in Zimbabwe. Population and Development Review. 2001;27:771–80.
13. Grassly N. C., Lewis J. J. C., Mahy M., Walker N., Timaeus I. M. Comparison of Survey Estimates with UNAIDS/WHO Projections of Mortality and Orphan Numbers in Sub-Saharan Africa. Population Studies. 2004;58:207–17. [PubMed: 15204254]
14. Gregson S., Anderson R. M., Ndlovu J., Zhuway T., Chandiwana S. K. Recent Upturn in Mortality in Rural Zimbabwe: Evidence for an Early Demographic Impact of HIV-1 Infection? AIDS. 1997;11:1269–80. [PubMed: 9256946]
15. Hill K. Estimating Census and Death Registration Completeness. Asian and Pacific Population Forum. 1987;1(3):8–13. 23–24. [PubMed: 12280697]
16. ———. 2003. "Adult Mortality in the Developing World; What We Know and How We Know It." Paper presented at the Training Workshop on HIV/AIDS and Adult Mortality in Developing Countries, New York, September. Organized by the United Nations Population Division, Department of Economic and Social Affairs, UN/POP/MORT/2003/1. http://www​.un.org/esa​/population/publications​/adultmort/HILL_Paper1.pdf.
17. Hosegood V., Vanneste A. M., Timaeus I. M. Levels and Causes of Adult Mortality in Rural South Africa: The Impact of AIDS. AIDS. 2004;18:663–71. [PubMed: 15090772]
18. Kitange H. M., Machibya H., Black J., Mtasiwa D. M., Masuki G., Whiting D., Unwin N. et al. Outlook for Survivors of Childhood in Sub-Saharan Africa: Adult Mortality in Tanzania. British Medical Journal. 1996;312:216–20. [PMC free article: PMC2349992] [PubMed: 8563587]
19. Kowal, P., C. Rao, and C. Mathers. 2003. Report on a WHO Workshop: Minimum Dataset on Ageing and Adult Mortality in sub Saharan Africa. Geneva: WHO.
20. 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.
21. 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;57:165–82.
22. Murray, C. J. L., and A. D. Lopez. 1996. The Global Burden of Disease: A Comprehensive Assessment of Mortality and Disability from Diseases, Injuries and Risk Factors in 1990 and Projected to 2020. Vol. 1 of the Global Burden of Disease and Injury series. Cambridge, MA: Harvard University Press.
23. Murray, C. J. L., G. Yang, and X. Qiao. 1992. "Adult Mortality: Levels, Patterns, and Causes." In The Health of Adults in the Developing World, ed. R. G. A. Feachem, T. Kjellstrom, C. J. L. Murray, O. Mead, and M. A. Phillips. New York: Oxford University Press.
24. Notkola V., Timaeus I. M., Siiskonen H. Impact on Mortality of the AIDS Epidemic in Northern Namibia Assessed Using Parish Registers. AIDS. 2004;18:1061–65. [PubMed: 15096810]
25. Nunn A. J., Mulder D. W., Kamali A., Ruberantwari A., Kengeya-Kayondo J.-F., Whitworth J. Mortality Associated with HIV-1 Infection over Five Years in a Rural Ugandan Population: Cohort Study. British Medical Journal. 1997;315:767–71. [PMC free article: PMC2127535] [PubMed: 9345167]
26. Pison G., Langaney A. The Level and Age Pattern of Mortality in Bandafassi (Eastern Senegal): Results from a Small-Scale and Intensive Multi-Round Survey. Population Studies. 1985;39:387–405.
27. Preston S. H., Bennett N. G. A Census-Based Method for Estimating Adult Mortality. Population Studies. 1983;37:91–104. [PubMed: 22077368]
28. Rutenberg, N., and J. Sullivan. 1991. "Direct and Indirect Estimates of Maternal Mortality from the Sisterhood Method." In Proceedings of the Demographic and Health Surveys World Conference, Washington, D.C. Vol. 3. Columbia, MD: IRD/Macro International.
29. Salomon J. A., Murray C. D. L. Modeling HIV/AIDS Epidemics in Sub-Saharan Africa Using Seroprevalence Data from Antenatal Clinics. Bulletin of the World Health Organization. 2001;79:586–692. [PMC free article: PMC2566469] [PubMed: 11477962]
30. Sewankambo N. K., Gray R. H., Ahmad S., Serwadda D., Wabwire-Mangen F., Nalugoda F., Kiwanuka N. et al. Mortality Associated with HIV Infection in Rural Rakai District, Uganda. AIDS. 2000;14:2391–400. [PubMed: 11089628]
31. Stanton, C., N. Abderrahim, and K. Hill. 1997. DHS Maternal Mortality Indicators: An Assessment of Data Quality and Implications for Data Use. Demographic and Health Surveys Analytical Report 4. Calverton, MD: Macro International.
32. Timaeus I. M. Estimation of Mortality from Orphanhood in Adulthood. Demography. 1991;28:213–27. [PubMed: 2070895]
33. ———1992Estimation of Adult Mortality from Paternal Orphanhood: A Reassessment and a New Approach Population Bulletin of the United Nations 3347–63. [PubMed: 12317481]
34. ———. 1993. "Adult Mortality." In Demographic Change in Sub-Saharan Africa, ed. K. A. Foote, K. H. Hill, and L. G. Martin. Washington, DC: National Academy Press.
35. ———1998Impact of the HIV Epidemic on Mortality in Sub-Saharan Africa: Evidence from National Surveys and Censuses AIDS 12Suppl. 1S15–27. [PubMed: 9677186]
36. ———. 1999. "Mortality in Sub-Saharan Africa." In Health and Mortality: Issues of Global Concern, ed. J. Chamie and R. L. Cliquet. New York and Brussels: United Nations Population Division and Population and Family Study Centre, Flemish Scientific Institute.
37. Timaeus, I. M., R. E. Dorrington, D. Bradshaw, N. Nannan, and D. Bourne. 2001. "Adult Mortality in South Africa, 1980–2000: From Apartheid to AIDS." Paper presented at the Population Association of America's annual meeting, Washington, DC, March 29–31.
38. Timaeus I. M., Jasseh M. Adult Mortality in Sub-Saharan Africa: Evidence from Demographic and Health Surveys. Demography. 2004;41:757–72. [PubMed: 15622953]
39. Timaeus I. M., Nunn A. J. Measurement of Adult Mortality in Populations Affected by AIDS: An Assessment of the Orphanhood Method. Health Transition Review. 1997;7(Suppl. 2):23–43.
40. Todd J., Balira R., Grosskurth H., Mayaud P., Mosha F., ka-Gina G., Klokke A. et al. HIV-Associated Adult Mortality in a Rural Tanzanian Population. AIDS. 1997;11:801–7. [PubMed: 9143613]
41. Tollman S. M., Kahn K., Garenne M., Gear J. S. Reversal in Mortality Trends: Evidence from the Agincourt Field Site, South Africa, 1992–1995. AIDS. 1999;13:1091–97. [PubMed: 10397540]
42. UNAIDS (Joint United Nations Programme on HIV/AIDS) Reference Group on Estimates Modelling Projections. Improved Methods and Assumptions for Estimation of the HIV/AIDS Epidemic and Its Impact: Recommendations of the UNAIDS Reference Group on Estimates, Modelling and Projections. AIDS. 2002;16:W1–W16. [PubMed: 12045507]
43. United Nations. 1982. Model Life Tables for Developing Countries. New York: United Nations.
44. ———. 2002. World Population Prospects: The 2000 Revision. Vol. 3: Analytical Report. New York: United Nations.
45. ———. 2003. World Population Prospects: The 2002 Revision; Datasets in Excel and PDF Formats. New York: United Nations.
46. Urassa M., Boerma J. T., Isingo R., Ngalula J., Ng'weshemi J., Mwaluko G., Zaba B. The Impact of HIV/AIDS on Mortality and Household Mobility in Rural Tanzania. AIDS. 2001;15:2017–23. [PubMed: 11600831]
47. WHO (World Health Organization). 2000. The World Health Report 2000—Health Systems: Improving Performance. Geneva: WHO.