RESEARCH

Health transitions in sub-Saharan Africa: overview of mortality trends in children under 5 years old (1950–2000)

Transitions sanitaires en Afrique sub-saharienne : présentation succincte des tendances de la mortalité chez les enfants de moins de 5 ans (1950-2000)

Transiciones sanitarias en el África subsahariana: panorama de las tendencias de la mortalidad en los menores de 5 años (1950–2000)

Michel Garenne^I,1; Enéas Gakusi^II

^IInstitut Pasteur, Unité d'Epidémiologie des Maladies Emergentes, 25–28 rue du Docteur Roux, 75724 Paris Cedex 15, France
^IIAfrican Development Bank, Tunis, Tunisia

ABSTRACT

OBJECTIVE: To reconstruct and analyse mortality trends in children younger than 5 years in sub-Saharan Africa between 1950 and 2000.
METHODS: We selected 66 Demographic and Health Surveys and World Fertility Surveys from 32 African countries for analysis. Death rates were calculated by yearly periods for each survey. When several surveys were available for the same country, overlapping years were combined. Country-specific time series were analysed to identify periods of monotonic trends, whether declining, steady or increasing. We tested changes in trends using a linear logistic model.
FINDINGS: A quarter of the countries studied had monotonic declining mortality trends: i.e. a smooth health transition. Another quarter had long-term declines with some minor rises over short periods of time. Eight countries had periods of major increases in mortality due to political or economic crises, and in seven countries mortality stopped declining for several years. In eight other countries mortality has risen in recent years as a result of paediatric AIDS. Reconstructed levels and trends were compared with other estimates made by international organizations, usually based on indirect methods.
CONCLUSION: Overall, major progress in child survival was achieved in sub-Saharan Africa during the second half of the twentieth century. However, transition has occurred more slowly than expected, with an average decline of 1.8% per year. Additionally, transition was chaotic in many countries. The main causes of mortality increase were political instability, serious economic downturns, and emerging diseases.

RÉSUMÉ

OBJECTIF: Reconstituer et analyser les tendances de la mortalité chez les enfants de moins de cinq ans en Afrique sub-saharienne sur la période 1950-2000.
MÉTHODES: 66 enquêtes démographiques et de santé, ainsi que des enquêtes mondiales sur la fertilité provenant de 32 pays africains, ont été sélectionnées pour analyse. Les taux de mortalité ont été calculés par période d'un an pour chacune des études. Dans les cas où l'on disposait de plusieurs études pour un même pays, les chiffres relatifs aux années de recouvrement ont été combinés. Des séries chronologiques spécifiques à chaque pays ont été également analysées en vue d'identifier les périodes correspondant à une tendance uniforme : déclin, stabilité ou augmentation. Les changements de tendance ont été étudiés à l'aide d'un modèle logistique linéaire.
RÉSULTATS: Pour un quart des pays étudiés, on a observé un déclin régulier, c'est-à-dire une transition sanitaire progressive. Un autre quart des pays ont enregistré des baisses à long terme, ponctuées par quelques hausses mineures sur de courtes périodes. Huit pays ont connu des augmentations importantes de la mortalité, imputables à des crises politiques ou économiques et dans sept autre, la mortalité a cessé de baisser pendant plusieurs années. Dans huit autres pays encore, la mortalité a augmenté au cours des dernières années en raison des cas pédiatriques de SIDA. Les valeurs et les tendances reconstituées ont été comparées à d'autres estimations établies par des organisations internationales, reposant dans la plupart des cas sur des méthodes indirectes.
CONCLUSION: Globalement, de grand progrès ont été obtenus dans la survie des enfants d'Afrique subsaharienne pendant la deuxième moitié du vingtième siècle. Néanmoins, la transition s'est opérée plus lentement qu'on ne l'attendait, avec une baisse moyenne de la mortalité de 1,8 % par an. En outre, cette transition a été chaotique dans plusieurs pays. Les augmentations de la mortalité étaient imputables principalement à des phases d'instabilité ou de forte récession économique, ou encore à des maladies émergentes.

RESUMEN

OBJETIVO: Reconstruir y analizar las tendencias de la mortalidad en los niños menores de 5 años en el África subsahariana entre 1950 y 2000.
MÉTODOS: Seleccionamos para el análisis un total de 66 encuestas de demografía y salud y encuestas mundiales de fecundidad de 32 países africanos. Las tasas de mortalidad se calcularon por periodos anuales para cada encuesta. Cuando había varias encuestas para un mismo país, se combinaban los años superpuestos. Las series cronológicas de cada país fueron analizadas para identificar los periodos con tendencias monotónicas, ya fueran decrecientes, constantes o crecientes. Se utilizó un modelo de regresión lineal para analizar los cambios de tendencia.
RESULTADOS: Una cuarta parte de los países estudiados presentaban tendencias de disminución monotónica de la mortalidad, lo que significa una transición sanitaria suave. Otra cuarta parte presentaba disminuciones a largo plazo con algunos aumentos de menor importancia durante periodos breves. Ocho países presentaban periodos de aumentos importantes de la mortalidad como consecuencia de crisis políticas o económicas, y en siete países la mortalidad dejó de disminuir durante varios años. En otros ocho países la mortalidad ha aumentado en los últimos años de resultas del SIDA pediátrico. Los niveles y tendencias reconstruidos se compararon con otras estimaciones realizadas por organizaciones internacionales, basadas generalmente en métodos indirectos.
CONCLUSIÓN: En términos generales, durante la segunda mitad del siglo XX se lograron avances muy importantes en materia de supervivencia infantil en el África subsahariana. Sin embargo, la transición ha sido más lenta de lo esperado, con una disminución media del 1,8% anual. Además, la transición ha sido caótica en muchos países. Las causas principales de aumento de la mortalidad fueron la inestabilidad política, graves crisis económicas y las enfermedades emergentes.

Introduction

The health transition, defined as a steady decline in mortality, has been one of the most important features of demographic changes in the twentieth century, and has had many economic and social consequences.^1–5 In sub-Saharan Africa, the health transition began somewhat later than in other countries. Although much change has occurred there since 1930, most African countries still have high levels of infant and child mortality compared with other regions, with much variation between countries.

To judge the health transition in Africa fairly, it seems most appropriate to consider trends in mortality rather than simply looking at current mortality. In countries where the health transition started later, a relatively high death rate after a period a steady decline in mortality could still indicate a favourable health transition, as is the case in some west African countries. By contrast, a situation of increasing mortality, but with a relatively low current mortality could hide a negative change, as is the case in some countries in southern Africa.

Documentation of mortality trends will allow a better understanding of the status of the health transition in Africa, and help to identify gaps where further action is needed. The monitoring of mortality trends is particularly imports tant in children younger than 5 years old, a group that is the main target of public health policies and the most common indicator of mortality levels in developing countries.

Comprehensive vital registration data remain the best source for assessing mortality trends, but these data are not currently available in most African countries. To assess mortality trends, analysts rely on data from demographic sample surveys, or other sources such as mortality data collected in censuses.^6–10 A synthesis of indirect mortality estimates in Africa was conducted at the World Bank for the period before 1985,^11–12 and showed a steady decline in mortality in almost all countries investigated. This study was repeated and extended to other countries, and included new estimates for African countries.^13–14 However, these syntheses suffer from the lack of precision associated with the use of indirect methods to estimate trends, and especially trend reversals. Ahmad et al. have completed analysis using direct and indirect estimates, and reconstructing trends by 5-year periods from 1955–59 to 1995–99.¹⁵ This compendium made better use of all available data, in particular direct estimates provided by Demographic and Health Surveys (DHS). However, although the use of 5-year time periods provided reasonable estimates of mortality levels and major trends, it often obscured the specific time periods when changes in mortality trends occurred. Being able to establish the precise date of reversals in mortality trends is important if the cause of these changes is to be identified.

In this study we aimed to provide new estimates of trends in mortality in children younger than 5 years in African countries with data from demographic sample surveys. The rationale of this analysis is to identify periods of monotonic change, and precise times at which trend changes occur. Here, we present a synthesis of our work; more details at country level are provided in a companion working paper.¹⁶

Methods

To reconstruct mortality trends, we used data from demographic sample surveys with maternity histories. These data provide dates of birth, and when applicable, age at death, for large samples of live births, which allowed us to compute age-specific death rates for periods many years before the survey. A total of 56 DHS and 10 World Fertility Surveys (WFS) were selected, covering 32 sub-Saharan Africa countries. In addition, a Multiple Indicator Cluster Survey (MICS) was included to cover Angola, which had no DHS or WFS survey. This MICS survey was based on a simplified methodology: only birth histories of the last three pregnancies were included, which shortens the retrospective period for estimating mortality trends. The sample included about 70% of all countries and 80% of the total population of continental sub-Saharan Africa and Madagascar.

Country mortality estimates

In the first step, we calculated death rates by computing person-years at risk and date of death. When several surveys were available for the same country, deaths and person-years were simply added for each age and period, with the respective weights of each survey. Two age groups were considered: infanthood (<12 months) and early childhood (12–59 months). Life table calculations provided the final estimate of the underfive mortality ratio, q(5), which expresses the probability of dying before the fifth birthday, as is done in DHS surveys.

Search for monotonic periods and inflexion points

In the second step, we searched for monotonic periods of mortality change (i.e. periods of either declining, constant or increasing mortality) and inflexion points (the times at which a change in slope occurred). In a smooth health transition, the rate at which death rates decline (the slope) tends to be constant over long periods of time, with decreases of 4% to 5% per year considered favourable indicators. The search for inflexion points was first done graphically, and then tested statistically. Once monotonic periods were identified, a linear-logistic model (Logit) was applied, the slope was estimated (the b coefficient of the regress sion line), and a test was performed to verify whether the change in slopes from one period to the next was significant. We discarded non-significant changes in slopes caused by erratic values, and made final estimates for all significant monotonic periods. Monotonic trends also allowed us to identify short periods with excess mortality by testing the difference between observed and expected value. The Logit model could be written as:

Logit[q(5)] = a + b × year

Where a = intercept and b = slope.

HIV seroprevalence and expected impact of paediatric AIDS

Paediatric acquired immunodeficiency syndrome (AIDS) has become a leading cause of death in many African countries in recent years. The number of AIDS-related deaths is adding to that of other causes of infant and child mortality, and by itself could reverse mortality trends in countries with high human immunodeficiency syndrome (HIV) prevalence. Paediatric AIDS comes almost entirely from the vertical transmission of the HIV virus from mother to child either during pregnancy, at the time of delivery, or through breastfeeding.¹⁷

We calculated the net effect of paediatric AIDS on mortality in under-5-year-olds assuming probabilities of vertical transmission, and probabilities of death in children infected with HIV. First, the dynamics of HIV seroprevalence in pregnant women was reconstructed for the 1980–2000 period. Trends in HIV prevalence were estimated from data provided by the AIDS database of the US Bureau of Census, International Programs. For many countries, HIV seroprevalence estimates were available for only a few years between 1980 and year 2000. Only one country — South Africa — had a complete series of data for all years between 1990 and 2000. In other countries, only a few data points were available. Trends in seroprevalence were fitted with a hyperbolic function of the Logit of the proportions of seropositive women, a model that we tested successfully on the South African data, and also applied to other countries. In one case where an increase in seroprevalence was halted and even reversed (Uganda), two curves were fitted, one ascending and the other descending. Although based on an ad hoc model, we found that this fitting procedure was suitable for all countries with a sufficient amount of data to test.

Once the dynamics of the HIV epidemic in pregnant women had been reconstructed, we calculated HIV prevalence at birth assuming a vertical transmission rate of 25%. To calculate paediatric AIDS mortality, we used a life table framework and the assumption that 30% of infected children died of HIV/AIDS each year between age 0 and 5 years. These parameters correspond roughly to the mean of several studies conducted in sub-Saharan Africa, and are consistent with a pooled analysis of African data.^18-19 For mortality, our estimates match the pooled estimates (52% versus 51% mortality, respectively, within 2 years), although some studies find somewhat lower mortality, so we might have overestimated AIDS mortality in some countries.

To calculate mortality trends without AIDS, it was assumed that causes of death were additive. Our calculations led to a final estimate of 16 per 1000 for Africa as a whole by 2000, which is consistent with the 13 per 1000 estimate published by other authors.^18-19

Results

The list of surveys used for our analyses is shown in Table 1, and results of the Logit adjustment (intercept and slope) are shown in Table 2, web version only, available from: http://www.who.int/bulletin. Where there were available data, we started the reconstruction of trends from 1950; reconstruction began later for countries with incomplete data. Estimates stop in the year of the last survey, which ranged between 1987 and 2000.

We calculated average estimates for sub-Saharan Africa by weighting the country estimates by the number of births in 2000. Results show a steady, and relatively slow, decrease in mortality since 1950, by about 1.8% per year. The decline would have been more rapid in recent years had it not been for the emergence of HIV/AIDS (Fig. 1). The calculations of the African average have some minor biases since the list of included countries is not exactly the same from year to year, especially for the first 10 and the last 10 years. However, if we had included only countries with full information, the bias in mortality calculations would have been even larger and the number of years covered much smaller.

Typological profiles of health transitions in Africa

The reconstructed trends allowed basic typological profiling of national health transitions in sub-Saharan Africa (Fig. 2).

Steady mortality decline

Only a few countries had a smooth health transition (Fig. 2a), i.e. a steady mortality decline in under-5-year-olds during the period covered by the surveys: Botswana (up to 1988), Comoro Islands, Ethiopia, Guinea, Liberia (up to 1986), Malawi, Mali and Togo. These countries had high levels of mortality in 1950, and exhibited no reversals in mortality trends and no minor accidents. The case of Liberia is unusual because no data are available after 1986. This country went through an ordeal of savage civil war in the 1990s, a period during which a rise in mortality might be expected. Similarly, data from Botswana stop in 1988, and it is likely that mortality rose in the 1990s because of HIV/AIDS, since this was the case in neighbouring countries. For Ethiopia, only one survey is available (for 2000), and evidence of trends during the troubled period of the 1970s is very limited. In any case, results of our analysis indicate that only a small number of African countries have had a smooth health transition.

Short periods of excess mortality

Some countries with a steady decline in mortality experienced several years of excess mortality, which although they were statistically significant did not substantially affect overall trends (Fig. 2b). In Burkina Faso mortality increased during drought in 1973–74 and again in 1994–95 in rural areas, but not in urban areas. In Chad, mortality increased significantly in 1980–81, at the height of the civil war. In Ghana, under-five mortality increased between 1979 and 1983 during a severe political crisis in the country that was later resolved with the return of Jerry Rawlings. In Lesotho, mortality increased significantly between 1962 and 1965. In Namibia, mortality temporarily rose during the struggle for independence (1977–82), to return to a trend of declining mortality until the arrival of HIV/AIDS in the 1990s. Likewise, mortality increased in Zimbabwe during the struggle for independence (1978–82) in that country. In Nigeria, mortality increased around the time of the Biafra war (1964–1968); between 1978 and 1988, another rise occurred in rural areas, but rates were stagnant in urban areas during the same period. In the United Republic of Tanzania, mortality increased during the period between 1979 and 1985, particularly in urban areas. In Togo, mortality rose in rural areas in 1993–94. However, for all the countries mentioned in this section the changes seem to have been transitory, since they did not have any substantial effect on long-term mortality trends.

Changing trends in underfive mortality: political and economic crises

Several countries went through prolonged periods of mortality increase as a result of a political crisis, an economic downturn or civil war (Fig. 2c). In Angola, the limited data available showed that under-five mortality declined during the late colonial period and the years following independence (1975–79), then increased during the civil war (1980–89), to decline again later when the situation became more stable (1993–96). A similar pattern was noted in Mozambique, a country that went through political change, and was also deeply affected by civil war in the 1980s. In Madagascar, mortality increased during the Malagasy revolution period (1975–86), when per capita GDP underwent a rapid decline. In Uganda, mortality increased markedly during the Idi Amin Dada years (1971–79) and the few years of political uncertainty following his departure (1980–83), until a stable situation returned with the arrival of Yoweri Museveni in 1986. In Rwanda, as in Burundi, mortality increased for about a decade after independence (1965–76). In Rwanda, however, a period of steady decline in mortality ended in 1991 following the attacks of the Tutsi refugees, peaked in 1994 during the period of genocide, and rose again in 1997–98. As a result, no trend is visible during the very troubled period of the 1990s in Rwanda. In Zambia, mortality increased between 1975 and 1992, especially in urban areas. This rise was a consequence of the major drop in international prices for the country's main export commodity, copper, in 1975. A serious economic crisis followed, inducing several political changes culminating with liberalization of the economy in 1992. This large increase in mortality occurred even after discounting the effect of HIV/AIDS.

Stagnation during 1980–2000

We noted periods of stagnation in under-five mortality in the 1980s and 1990s (Fig. 2d) for several countries. In Benin, mortality stopped declining between 1979 and 1989 in urban and rural areas despite favourable economic indicators. In the Central African Republic, mortality stagnated between 1977 and 1989, a difficult period following the installation and fall of Jean-Bédel Bokassa. In Gabon, mortality stagnated after 1985, even after the effect of AIDS had been discounted. In Niger, mortality plateaued in rural areas between 1972 and 1992. In Northern Sudan, mortality remained steady between 1974 and 1984, and even increased slightly in rural areas.

Epidemiological crisis (malaria)

Only one country showed evidence of a reversal in mortality trends due to a disease other than HIV/AIDS. In Senegal, mortality in under-5-year-olds increased between 1960 and 1970, an increase closely associated with a rise in malaria morbidity and mortality that followed the failed attempt to eradicate malaria from 1955 to 1959 (Fig. 2e). Malaria mortality seems to have increased again in the 1990s according to local studies,²⁰ although there was no discernable effect on all-cause mortality.

Mortality increases caused by HIV/AIDS

Paediatric AIDS has had a strong effect on death rates in several countries, sometimes causing or exacerbating mortality increases, and in other cases slowing the rate of mortality decline. The effect is especially evident in southern African countries (Botswana, Namibia, South Africa, Zimbabwe, Zambia), in some eastern African countries (Kenya, Uganda, United Republic of Tanzania), and in western African (Cameroon, Côte d'Ivoire) (Fig. 2f). In Namibia, United Republic of Tanzania and Zimbabwe, the increase in under-five mortality in recent years seems to be entirely attributable to HIV/AIDS, whereas in other countries (Cameroon, Côte d'Ivoire Kenya, South Africa), paediatric AIDS contributed to only part of the increase. In Uganda and Zambia, mortality has declined in recent years, despite a large effect of HIV/AIDS. Our findings show that the net effect of HIV/AIDS in African countries is complex and dependent on a matrix of factors.

Reversal in health transition after discounting HIV/AIDS

In several countries, under-five mortality remained steady or increased after discounting the effect HIV/AIDS. In South Africa, even after subtracting the effects of paediatric AIDS, under-five mortality stopped declining after 1993. In Cameroon, under-five mortality after discounting HIV/AIDS was estimated at 145 per 1000 in 1998, whereas a lower rate (101 per 1000) was expected from previous trends. In Côte d'Ivoire, under-five mortality was still decreasing from 1983 to 1998 after discounting the effect of HIV/AIDS, but was declining more slowly than it was before 1983. In Kenya, under-five mortality increased after 1985, even after the effect of HIV/AIDS was discounted, and data from the 2003 DHS survey shows that the rate seems to be still increasing.

Divergent changes in urban and rural areas

In most countries, mortality change did not differ greatly between urban and rural areas. However, several cases of divergent changes were documented. In Burundi, rural mortality declined between 1976 and 1986, whereas it increased in urban areas. In the Central African Republic, mortality increased in rural areas between 1976 and 1990, but declined in urban areas during the same period. In Malawi, urban mortality increased substantially after 1984, but remained unchanged in rural areas. In Mozambique, urban and rural areas experienced different trends during the civil war, which had the greatest effect in the countryside. In Niger, mortality remained unchanged between 1972 and 1992 in rural areas, but continued to decline in urban areas. More details on urban/rural differences in Africa are given elsewhere.²¹

Recent trends

With data from the new round of DHS surveys becoming available, we were able to check the consistency of previous trends. In Burkina Faso, the downward trend has resumed after the 1994–95 crisis. In Cameroon, the rise in mortality since 1990 stopped around 1996, and a rapid mortality decline has resumed. In Kenya mortality continued to rise at least until 2001. In Ghana, mortality remained steady from 1995 to 2002. In Nigeria, data from the 2003 survey were found to be incompatible with previous estimates, which makes any inference difficult. The last three surveys conducted in Nigeria suggest that mortality remained steady since 1985, but mortality levels vary by a ratio of 1 to 2 from survey to survey. In the United Republic of Tanzania, indirect estimates from the 2004 survey suggest a rise in mortality since 1995, probably due to HIV/AIDS.

Discussion

Despite their limitations, data from demographic sample surveys were sufficient to allow us to reconstruct the main trends in under-five mortality. We were also able to identify some erratic changes, as well as mortality increases due to HIV/AIDS and other causes such as political or economic crises. Results from this reconstruction indicate that the health transition in Africa has been sustained between 1950 and 2000, with under-five mortality declining by about 1.8% a year. This decline for the whole continent has been affected by many situational factors, with minor or major mortality increases, and especially by the effect of AIDS.

Explanation of these changes, both positive and negative, requires a renewed look at mortality determinants. On one hand, major changes have occurred in the field of public health, including the increased number of physicians per capita, improved vaccination coverage, and widespread use of modern medicines, as well as urbanization and modern education, all of which have contributed to reductions in mortality. On the other hand, political crises, civil wars, poor state management, and, to a lesser extent, economic crises, have contributed to mortality increases. HIV/AIDS emerged in Africa in the mid 1980s to complicate the health situation, and now accounts for many increases in mortality seen since this time. Fluctuations in the prevalence and severity of malaria may also have played a role in mortality, although this contribution seems small compared to the other factors. Recent mortality increases after discounting the effect of HIV/AIDS require further research. In particular, increases in poverty in large cities might explain mortality increases in places such as Cameroon, Côte d'Ivoire and Kenya.

This attempt to reconstruct under-five mortality trends has several advantages over earlier estimates. Compared with indirect estimations, which often are out of date and lack precision, the direct estimates seem to provide more robust trends and inflexion points. Compared with earlier WHO estimates, which are based on a mixture of direct and indirect estimates, reconstructed trends are more consistent and often reveal patterns that had been smoothed by the five-year estimates. Furthermore, discounting the effects of HIV/AIDS allows for better identification of the proportion of mortality trends attributable to emerging diseases, and the proportion due to other causes.

The complexity of demographic dynamics is apparent when examining country data. There is much variation between countries' health transitions in sub-Saharan Africa, and are the results of the interference of several factors, the most important of which seem to be political stability, state management, and emerging diseases. The diversity of situations among the countries should be taken into account to better understand the various patterns of under-five mortality trends seen in sub-Saharan Africa during the second part of the twentieth century.

Acknowledgements

We thank the Centre d'Etudes et de Recherches sur le Développement International (CERDI), Clermont-Ferrand for administrative support.

Funding: This research was sponsored by the Wellcome Trust, Health Consequences of Population Change Programme (grant number: 062885/Z/00/Z).

Competing interests: none declared.

References

1. Stolnitz G. A century of international mortality trends, pt.1, Popul Stud (UK) 1955;9:24-55.        

2. Stolnitz GJ. Recent mortality trends in Latin America, Asia and Africa: review and re-interpretation. Popul Stud 1965;19:117-38.        

3. Preston SH. Causes and consequences of mortality declines in less developed countries during the twentieth century. In: Easterlin RA (editor). Population and Economic Change in Developing Countries. Chicago: University of Chicago; 1980. p. 289-353.        

4. Preston SH. Mortality patterns in national populations. New York: Academic Press; 1976.        

5. McKeown T. The modern rise of population. New York: Academic Press; 1976.        

6. United Nations, Population Division. Levels and trends of mortality since 1950. UN Population Study No. 74. New York, United Nations; 1982.        

7. Brass W,. The demography of tropical Africa. Princeton (NJ): Princeton University Press; 1968.        

8. Feachem R, Jamison D (editors). Disease and mortality in sub-Saharan Africa. New York: World Bank/Oxford University Press; 1991.        

9. Mandjale AE. Mortalité infantile et juvénile en Afrique: niveaux et caractéristiques, causes et déterminants. Louvain la Neuve: CIACO; 1985. p. 270.        

10. van de Walle E, Ohadike PO, Sala-Diakanda MD, (editors). The state of African demography. Liège: IUSSP-Derouaux; 1988.        

11. Hill A, Hill K. Mortality in Africa: levels, trends, differentials and prospects. In: van de Walle E, Ohadike PO, Sala-Diakanda MD (editor). The State of African Demography. Liège: IUSSP-Derouaux; 1988. p. 67-101.        

12. Hill A. Infant and child mortality: levels, trends and data deficiencies. In: Feachem R Jamison D (editors). Disease and mortality in sub-Saharan Africa. New York: World Bank/Oxford University Press; 1991. p. 37-76.        

13. Hill K, Yazbeck A. Trends in under-five mortality, 1960-90: estimates for 84 developing countries. Washington, DC: World Bank; 1994.        

14. Hill K, Pande R, Mahy M, Jones G. Trends in child mortality in the developing world: 1960 to 1996. New York: UNICEF; 1999. Available from: http://www.childinfo.org        

15. Ahmad OB, Lopez AD, Inoue M. The decline in child mortality: a reappraisal. Bull World Health Organ 2000;78:1175-91.        

16. Garenne M, Gakusi E. Reconstructing under-five mortality trends in Africa from demographic sample surveys. DHS Working Papers No 26. Calverton (MD): IRD-Macro. Available from: http://www.measuredhs.com        

17. Schmid GP, Buvé A, Mugyeni P, Garnett GP, Hayes RJ, Williams BG, et al. Transmission of HIV-1 infection in sub-Saharan Africa and effect of elimination of unsafe injections. Lancet 2004;363:482-8.        

18. Newell ML, Brahmbhatt H, Ghys PD. Child mortality and HIV infection in Africa: a review. AIDS 2004;18 Suppl 2:S27-S34.        

19. Newell ML, Coovadia H, Corina-Borja M, Rollins N, Gaillard P, Dabis F. Mortality of infected and uninfected infants born to HIV-infected mothers in Africa: a pooled analysis. Lancet 2004;364:1236-43.        

20. Trape JF. The public health impact of chloroquine resistance in Africa. Am J Trop Med Hygiene 2001;64 Suppl 1:12-7.        

21. Garenne M. Migration, urbanisation and child health in Africa: a global perspective. In: Tienda M (editor). African migration and urbanisation in a comparative perspective. Johannesburg: Witwatersrand University Press; 2006.        

(Submitted: 15 December 2005 – Final revised version received: 20 February 2006 – Accepted: 22 February 2006)

1 Correspondence to this author (email: mgarenne@pasteur.fr).