1 – An age-old scourge of the Southern countries

3Plasmodium falciparum malaria is a potentially lethal parasitic disease whose epidemiology varies widely between regions, depending on the climate, the environment and land use. These characteristics influence its impact on morbidity and mortality by affecting the frequency and intensity of exposure. In regions of low endemia or where malaria occurs in epidemic form, the disease can be deadly at all ages. In endemic zones, malaria mortality is high, but most victims are young children, since individuals develop premunition which reduces the risk of death as they grow older.

4Known since ancient times, malaria was eradicated in Europe – including along the coastal zones of mainland France – in the 1940s by massive DDT spraying. In the early 1950s, it was still a health risk in the inter-tropical regions of the world and in most Southern countries. But the effective use of insecticides to destroy malaria-carrying mosquitoes and, above all, the wide availability of a low-cost medication, chloroquine, led to hopes that malaria would be eradicated world-wide. For example, the virtual disappearance of malaria from Sri Lanka in just two years was a spectacular success story, increasing life expectancy by an estimated twelve years. [1]

5But chemical use led to the development of resistance: the mosquitoes became resistant to insecticides – which also proved toxic to humans – and the parasites to the anti-malarial drugs. In 1961, chloroquine-resistant strains of the malaria parasite emerged almost simultaneously in tropical Asia and America (Payne, 1987). But because the countries affected by malaria are poor, chloroquine remained in use as first-line treatment, and chloroquine resistance reached eastern Africa in 1978 and the west of the continent a decade later. In parallel, the malaria parasite developed resistance to the other drugs used as second- or third-line treatments, such as quinine or sulfadoxine-pyrimethamine [2] (Sibley et al., 2001; Gregson and Plowe, 2005).

6The launch of the WHO Roll Back Malaria project in 1998 marked a turning point in the fight against malaria, with the development of a new strategy encompassing the many different factors at play: clinical, biological, but also sociocultural and economic. Prevention was reinforced – notably through the use of insecticide impregnated mosquito nets – and new treatments were introduced in countries still using chloroquine despite its inefficacy. [3] In therapeutic terms, the combined use of several different molecules to minimize risk of resistance is probably the most promising solution (Trape, 2001). WHO currently recommends artemisinin-based combination therapy (ACT), a much more expensive form of treatment than chloroquine [4] (WHO, 2005), and less affordable for populations in developing countries (Agnamey et al., 2005; Marquet, 2003).

2 – Impact of malaria on mortality in Africa

7In sub-Saharan Africa, two-thirds of the population is exposed to malaria. Around 80% of malaria deaths occur in this region of the world, mainly among children under five (WHO, 2005). In 2000, the disease was responsible for an estimated 15-20% of infant and child mortality, though the exact number of malaria deaths in Africa is very difficult to determine [5] (Rowe et al., 2006; Smith et al., 2004, 2006; Snow et al., 1999a, 1999b; MARA/ARMA, 1999).

8Mortality in sub-Saharan Africa has increased in recent decades. This is due mainly to AIDS, although the resurgence of malaria has also contributed to the worsening health situation (WHO, 2005). Despite certain limits, [6] surveys representative at national level, such as the Demographic and Health Surveys (DHS), provide a means to assess general mortality trends in childhood (Barbieri, 1989). They show that child mortality increased in many countries in the 1990s, including in the African countries least affected by the AIDS epidemic, such as Senegal, where malaria is the leading cause of morbidity and mortality among children under five (Ndiaye and Ayad, 2006). In the absence of national cause-of-death statistics, however, the contribution of malaria to the increase in child mortality in the 1990s is difficult to measure.

9In most African countries, no civil records are kept outside the major cities: most children are not registered at birth (UNICEF, 2002), death registration is incomplete and reported ages at death are not always reliable (Lohlé-Tart and François, 1999; Hill, 1999). In addition, cause-of-death data from health infrastructures are of poor quality. In rural areas, few people are seen by a doctor before they die, and no autopsy is performed after death. [7] In 2001, only four countries of sub-Saharan Africa produced high-quality national data on causes of death [8] (Mathers et al., 2006).

3 – The role of demographic surveillance sites in measuring malaria mortality

10Few data sources are thus available for the detailed study of mortality, especially in rural areas. Demographic surveillance systems (DSS) have been developed to provide detailed and reliable population data (Pison, 2005; Delaunay, 2002). A geographically delimited population is monitored by means of multi-round surveys following an initial census. DSSs do not provide data representative at national level, but as malaria epidemiology is strongly linked to local characteristics, they are well suited to the study of the disease.

11DSS sites can provide detailed information on mortality. Whenever someone in the study population dies, the age at death is recorded, along with the probable cause of death if a verbal autopsy is carried out. A questionnaire is administered to the family of the deceased individual to record information on the reason for death: the disease, its treatment and its symptoms. A physician is then consulted to interpret the information obtained and identify a probable cause of death. This verbal autopsy method was first used in the 1950s (Biraud, 1956). It has been widely developed since then and is now used on numerous DSS sites in sub-Saharan Africa and Asia (Adjuik et al., 2006; Indepth, forthcoming).

12Medical causes of death are difficult to classify, however, even in a context where they are systematically diagnosed and recorded (Meslé, 2006). Although highly informative, the method has limits which have been widely studied by researchers working on this type of data, and the scientific literature on the question has been abundant since the late 1980s (Garenne and Fontaine, 1988; Snow et al., 1992; Chandramohan et al., 1994; Anker, 1997; Fauveau, 2006; Chandramohan et al., 2005; Garenne and Fauveau, 2006; Soleman et al., 2006; Setel et al., 2006). [9]

13The reliability of the method depends on various factors ranging from data collection to cause classification – choice of respondent, questionnaire used, experience of diagnosing physicians, etc. – but also on the cause of death itself. Its accuracy is generally measured in terms of sensitivity [10] and specificity [11] (Anker, 1997; Chandramohan et al., 2001). These indices serve to assess the divergence that may exist between the actual cause of death and the cause diagnosed by verbal autopsy, since this discrepancy may have a major impact on cause-specific mortality rates (Maude and Ross, 1997). Generally, deaths due to injury or diseases with very characteristic symptoms, such as scarlet fever or rabies, are accurately classified. For malaria, however, there are no pathognomonic symptoms (Rogier et al., 2005): fever, anaemia, coma, respiratory distress and neurological disturbances may be observed in a variety of pathologies (Marsh et al., 1996; Berkley et al., 1999). The verbal autopsy method may thus be less reliable for diagnosing malaria deaths. Local studies which have compared verbal autopsy diagnoses with clinical diagnoses show that the method is not very sensitive and moderately specific (Rowe, 2005; Anker et al., 1999; Todd et al., 1994; Snow et al., 1992). In practice, it is often impossible to measure these indices precisely because the actual cause of death is unknown.

14Measurement of malaria mortality in sub-Saharan Africa thus poses a dual problem: the shortage of data on causes of death in general, and the uncertain reliability of existing data based on verbal autopsies. Although limited, current estimations of malaria mortality in Africa are nonetheless based primarily on this kind of data collected at local level (Rowe et al., 2006).

1 – A rural population of Senegal monitored since 1985

15The Mlomp health and demographic surveillance system (HDSS) in southwestern Senegal was set up in 1985 (Pison et al., 2002) to measure demographic levels and trends in a rural population of sub-Saharan Africa. In 1985, two other DSS sites – Niakhar (Garenne and Cantrelle, 1997) and Bandafassi (Pison et al., 1997), were already functioning in the centre-west and south-east of Senegal. The addition of Mlomp ensured fuller coverage of the country’s demographic diversity (Map 1).

Map 1

Location of Mlomp and the two other DSS sites in rural Senegal

Location of Mlomp and the two other DSS sites in rural Senegal

16Since the initial census in 1985, the population of Mlomp has been surveyed annually. At each survey round, the demographic events of the preceding year – births, deaths, unions and migration – are recorded. The verbal autopsy method is used to diagnose biomedical causes of death.

2 – Malaria research in Mlomp

17The Mlomp region has a sub-tropical climate, with a rainy season from June to October and a dry season. Plasmodium falciparum malaria is mesoendemic in the area: transmission occurs throughout the year, but is especially intense over a 6-8 month period. [12] Only children are exposed to the severe and potentially fatal forms of the disease.

18A range of medical research programmes have been conducted in partnership with the village dispensary since the late 1980s, notably on chloroquine resistance and the efficacy of anti-malarial treatments. Many patients who come to the dispensary with fever symptoms are given a thick blood smear test to measure the Plasmodium falciparum density in the blood and to diagnose a malaria attack.

19The pooled findings of these programmes have revealed the impact of chloroquine resistance on mortality attributable to malaria in western Africa (Trape et al., 1998), and have confirmed the efficacy of ACT combination therapy (Adjuik et al., 2002) and its role in reducing mortality (Cissé et al., 2005). Indeed, the first two studies are very regularly cited. [13]

20As the medical information recorded at the dispensary in Mlomp is used to diagnose causes of death among the population monitored by the demographic surveillance system, the reliability of diagnoses is greatly enhanced. Yet this information is neither complete nor exhaustive. Given the difficulty of determining cause of death – and more specifically death from malaria – on DSS sites, a more in-depth study of diagnosis reliability is needed to measure malaria mortality trends since 1985 and to estimate its impact on overall mortality.

3 – Mlomp and its population

21Mlomp is a rural community located in a woodland area surrounded by paddy fields. On 1 January 2005, the monitored population totalled 8,008 people belonging to households grouped into family concessions based on the patrilinear system. The houses have adobe walls with corrugated iron or straw roofing. The inhabitants have no electric power supply and drinking water is drawn from the area’s many wells (Pison et al., 2002).

22Most inhabitants of Mlomp are of the Diola (Joola) ethnic group, and are animists or Catholics. They speak, Diola, the local language, many speak Wolof, the country’s lingua franca, while French, the institutional language is taught in school. The educational infrastructure is relatively well developed and in 2000, 55% of women aged 15-49 had been to school for at least one year, compared with 20% of rural Senegalese women overall in 2005 (Pison and Enel, 2005; Ndiaye and Ayad, 2006).

23Rice growing, which occupies the population during the wet season, is the main local resource. As no local cash crops are produced, the inhabitants of Mlomp migrate to earn money. The men often harvest palm wine or fish during the dry season, and the young women work in town as maids before returning to the village to marry (Pison et al., 2001). For example, on 1 January 2005, in the middle of the dry season, half of all young men and women aged 15-30 were absent from the village, and 43% of men aged 30-59.

24While the total fertility rate exceeded 6 children per woman in the early 2000s in rural Senegal (Ndiaye and Ayad, 2006), women in Mlomp had 4 children on average over the period 1985-2004. This fertility level is linked to a late age at first live birth (23 years). The first birth generally occurs before marriage, when the partners are still living separately. After marriage, births are spaced by means of contraception (17% of women aged 15-49 in 2000) and breastfeeding (median duration 19 months) (Pison et al., 2001).

25Over the period 1985-2004, life expectancy at birth was 60.5 years. [14] [15] Life expectancy is difficult to estimate at the national level, [16] but comparison with observed life expectancy on the other rural DSS sites in Senegal over the same period shows that mortality in Mlomp is relatively low for a rural region of Senegal. Comparison of under-five mortality produces the same conclusion (Table 1). A retrospective survey of birth histories of women enumerated in Mlomp in 1985 revealed a sharp drop in the risk of dying before age 5 from the 1960s, in parallel with the development of health infrastructures in the area (Pison et al., 1993).

Table 1

Life expectancy and under-five mortality at the three rural DSS sites in Senegal, 1985-2004

Bandafassi Mlomp Niakhar Life expectancy at birth (e0) 48 years 61 years 54 years Under-five mortality (5q0) 239‰ 101‰ 202‰ Source: Pison et al. (2005).

Life expectancy and under-five mortality at the three rural DSS sites in Senegal, 1985-2004

4 – Health infrastructure and measures to combat malaria

26As in many rural areas of sub-Saharan Africa, healthcare services in Mlomp are sparse. The nearest doctor is around 10 km from the area and the regional hospital equipped with surgical facilities is 50 km away. There is nonetheless a dispensary, opened in 1961, and a maternity clinic, opened in 1968.

27These two private facilities function well and have done so for many years: working under the supervision of the district physician, a nurse is present every day at the dispensary [17] and two matrons ensure a permanent presence at the maternity clinic. In the event of complications, patients are taken by the nurse to the district hospital, or by ambulance from the district hospital to the regional hospital. All women in Mlomp give birth at the maternity clinic [18] (Enel et al., 1993). As a result, there are fewer maternal deaths on the Mlomp site than on the two other rural DSS sites in Senegal [19] (Pison et al., 2000) and neonatal mortality is relatively low. The risk of dying in the first month of life was 27‰ in 1985-2004 against 46‰ in rural Senegal as a whole over the period 1996-2005 (Ndiaye and Ayad, 2006).

28For many years, monthly weighing [20] and vaccination [21] sessions have been organized. Practically all children in Mlomp receive the vaccinations recommended by the health authorities, which is rarely the case elsewhere. [22] These sessions also provide an opportunity for “get-togethers” where the matrons advise the mothers about their children’s health. They focus mainly on breastfeeding, hygiene, vaccination and prevention of the main diseases not controlled by vaccination, such as intestinal infections and malaria.

29Chloroquine was used as first-line treatment against malaria in Mlomp from 1961, when the dispensary first opened. In 1975, a major programme of chloroquine-based prophylaxis was launched during the rainy season, and presumptive treatment was given to children in the form of self-medication at the first signs of fever. But with the emergence and rapid spread of choloroquine resistance from 1990, chemoprophylaxis was limited in 1993 to children under 4 and pregnant women (Trape et al., 1998; Pison et al., 1993). Chloroquine is still used for presumptive treatment of young children and quinine is used in cases of high fever and for adults. However, the chloroquine resistance of Plasmodium falciparum has increased steadily, and exceeded 70% in 1997 (Sokhna et al., 1997; Trape et al., 2002).

30In 1999, an international study with WHO funding was set up to test an ACT combining amodiaquine+artesunate on four African DSS sites (Adjuik et al., 2002). Several hundred children aged between 6 months and 10 years were treated at the Mlomp dispensary, either with amodiaquine only, [23] or with amodiaquine+artesunate. The drugs were supplied by pharmaceutical laboratories. Encouraging results were obtained and in 2000 the entire population was treated with this ACT during the rainy season. Since 2002, treatment has been administered on a year-round basis. All these changes have been made with the agreement of the health authorities and the dispensary healthcare personnel are actively involved. Their participation is key to the success of the programme. This new first-line therapy currently costs between €0.74 for a child under five and €1.30 for an adult (Agnamey et al., 2005), plus the price of the consultation (€0.15 for an adult, and half that price for a child in 2002) and of the blood test to detect Plasmodium falciparum (€0.30 in 2002). Though expensive for the villagers, the treatment is highly effective and avoids the need for second-line intramuscular quinine injections which are even more costly.

31In parallel, the local health organization encourages the population to use insecticide-impregnated mosquito nets, especially during the rainy season, to protect pregnant women, babies and small children in particular.

32Throughout the 1990s, the health authorities had little scope for action, and a new upturn in malaria mortality was feared. The arrival of new treatments, introduced progressively since the early 2000s, has raised new hopes of progress in combating the disease.

1 – Protocol for determining causes of death

33In Mlomp, verbal autopsies are recorded during the annual data collection round, at the earliest after the mourning period, and at the latest one year after death if the person died just after the previous round. Verbal autopsies are conducted by interviewers [24] from Mlomp itself, to ensure maximum comprehension in both linguistic and sociocultural terms and to guarantee a relation of trust which is vital for successful data gathering [25] (Chandramohan et al., 2005). It is the interviewers’ task to find the friend or relative most capable of answering questions on the disease history and symptoms. Since 1985, the same verbal autopsy questionnaire [26] has been used for all deaths, though it includes a special section for children and women of reproductive age. [27] Once the names of the deceased and the respondent have been recorded, the latter reports what he/she believes to be the cause of death, any treatments received by the deceased during illness and the history of the deceased’s illness. Questions are then asked about a whole series of symptoms, [28] and details are recorded where relevant to the deceased.

34The medical information obtained from the local health institutions with the consent of the district physician enhances the reliability of verbal autopsies. The protocol for determining causes of death remained largely unchanged up to 2003. A single malaria specialist determined causes of death by visiting Mlomp at the end of the annual survey to consult the various registers [29] kept by the dispensary nurse and to check them against the verbal autopsies. Causes of death were coded in accordance with the ninth revision of the International Classification of Diseases (WHO, 1977). Only the underlying cause of death was identified and recorded.

35To facilitate comparative studies, the three rural DSS sites in Senegal all adopted the same protocol in 2004. In Mlomp, this harmonization involved recruting a new interviewer with medical training and entrusting autopsy analysis to several physicians with different fields of expertise to give complementary medical opinions. Any medical information available at the dispensary is now recorded on the autopsy at the time of data collection and submitted to the diagnosing physicians in Dakar. For the study of malaria mortality, we focus on the period 1985-2003 which is homogeneous in terms of data sources, collection and archiving.

2 – Reliability of malaria diagnosis

36Our analysis is limited to deaths occurring before age 20, since beyond that age the probability of dying from malaria is very low thanks to acquired premunition. From 1985 to 2003, no malaria deaths were diagnosed beyond age 20, and more than 80% of such deaths occurred before age 5 (Table 2).

Table 2

Distribution of diagnosed malaria deaths by age group in Mlomp, 1985-2003

Age group Number of deaths % aggregate % Below age 1 8 13 13 Age 1-4 41 69 82 Age 5-9 8 13 95 Age 10-19 3 5 100 Overall 60 100 Source: Mlomp database, 2005.

Distribution of diagnosed malaria deaths by age group in Mlomp, 1985-2003

37If we consider the causes of death as diagnosed, the risk of dying from malaria before age 20 has varied greatly over time, reflecting the sequence of epidemiological and therapeutic changes described above. There was practically no malaria mortality up until the early 1990s, and although it fluctuated from 1992, it rose to a mean level of over 30‰ in the middle of the decade before gradually declining (Figure 1).

Figure 1

Annual variation in risk of dying from malaria before age 20, Mlomp, 1985-2003

Annual variation in risk of dying from malaria before age 20, Mlomp, 1985-2003

38The knowledge and experience of the physician responsible for determining cause of death may affect his/her decision to classify a death as due to malaria or otherwise (Desgrées du Loû et al., 1996). The hypothesis that the availability of medical information improves the reliability of malaria diagnosis in Mlomp endorses the validity of these data. In etiological terms, although the presence of the parasite in the blood confirms that a malaria attack occurred, it does not necessarily mean that malaria was the underlying cause of death. Moreover, thick blood smear tests were not introduced until 1989 and are not available for all children who died. Certain cases of malaria death may therefore have gone undiagnosed. In addition, not all deaths are clearly diagnosed, and those with ill-defined causes are not distributed independently across the various causes of death, whose ease of diagnosis varies (Desgrées du Loû et al., 1996; Kahn et al., 2000). Over the period 1985-2003, one-third of all recorded deaths did not have a clearly defined cause. Though the proportion is only one-quarter for deaths occurring before age 60, this is still a far from negligible share. Before age 15, half of deaths from ill-defined causes are characterized by general symptoms (fever, convulsions, chills, coma, etc.) (Table 3).

Table 3

Distribution of deaths from ill-defined causes by age group, Mlomp, 1985-2003

Under 15 15-59 60+ Overall Percentage of ill-defined deaths 25 24 39 32 Including general symptoms 12 4 8 8 Senility, old age 0 0 9 4 Other symptoms 4 9 9 8 Sudden death or unknown cause 9 11 14 12 Number of ill-defined deaths 105 83 286 474 Total number of deaths 426 344 727 1,497 Source: Mlomp database, 2005.

Distribution of deaths from ill-defined causes by age group, Mlomp, 1985-2003

3 – Definition of deaths potentially linked to malaria

39To estimate the number of deaths potentially linked to malaria, we divide deaths into four groups: deaths diagnosed as malaria; deaths diagnosed as non-malarial; deaths from ill-defined causes but associated with symptoms and last, deaths from unknown causes. Within each group, we identify possible malaria deaths on the basis of available information (symptoms, medical information, age and season of death, comments of the physician making the cause-of-death diagnosis) (Table 4).

Table 4

Deaths potentially due to malaria occurring before age 20 by type of diagnosis, Mlomp, 1985-2003

Degree of certainty Number of deaths % Diagnosed malaria deaths Malaria attack and positive thick blood smear test Certain 27 45.0 Typical malaria attack (symptoms, season, age) Presumed 26 43.3 Malaria attack or other cause, or ill-defined death Possible 7 11.7 Total 60 100.0 Diagnosed non-malaria deaths Contributing cause or doubt Possible 26 9.3 No doubt or negative thick blood smear test Ruled out 253 90.7 Total 279 100.0 Deaths from ill-defined causes with symptoms Indeterminate fevers or other general symptoms Possible 45 62.5 Other symptom, no doubt or negative thick blood smear test Ruled out 27 37.5 Total 72 100.0 Death from unknown cause No indication 39 100.0 Source: Mlomp database, 2005.

Deaths potentially due to malaria occurring before age 20 by type of diagnosis, Mlomp, 1985-2003

40The 60 diagnosed malaria deaths from 1985 to 2003 are classified as certain (confirmed by a positive thick blood smear test), presumed (typical malaria attacks at high-risk age and in a high-risk season), or possible (contributing or concurrent cause, borderline ill-defined cause). Under this classification, almost half the diagnosed malaria deaths are certain, 43% are presumed and 12% are possible.

41Deaths diagnosed as non-malarial are classified either as possible malaria deaths (doubtful diagnosis or associated malaria) or as deaths for which a malaria diagnosis is ruled out (certain diagnosis or negative blood smear test). Out of 279 deaths, more than 9% – around 30 – could be due to malaria. Most were diagnosed as deaths due to pneumonia or bronchial pneumonia, ill-defined perinatal or intestinal infections, or meningitis. Indeed, the clinical signs of certain infections are very similar to those of a malaria attack.

42More than half the deaths from ill-defined causes but with associated symptoms could be linked to malaria. Of these 45 deaths, 41 correspond to description of general symptoms: 30 cases of fever, 7 of coma and 4 of convulsions.

43Last, among death from ill-defined causes, we identify those for which no details are available. Deaths are classified in this category in the total absence of information from verbal autopsies or from the dispensary. These deaths cannot be linked to any specific cause and their proportional distribution across the other categories – certain, presumed, possible and ruled out – is probably the most neutral redistribution method. Over the period 1985-2003, 39 deaths before age 20 are of unknown cause.

44The grouping of these deaths suggests that the share of malaria mortality in deaths before age 20 over the period 1985-2003 lies somewhere between less than 7% and almost 32%. There is a very broad range of possibles, from certain malaria deaths to those which might possibly be associated (Table 5). In what follows, these deaths – i.e. certain, presumed or possible malaria deaths – will be qualified as deaths potentially due to malaria.

Table 5

Deaths potentially due to malaria, Mlomp, 1985-2003

Degree of certainty Number of deaths Number after redistribution of deaths from unknown cause Percentage Aggregate percentage Certain 27 29.6 6.6 6.6 Presumed 26 28.4 6.3 12.9 Possible 78 85.4 19.0 31.9 Ruled out 280 306.6 68.1 100.0 Unknown cause 39 – – Total 450 450.0 100.0 Source: Mlomp database, 2005.

Deaths potentially due to malaria, Mlomp, 1985-2003

4 – Impact of malaria on mortality

45To study the impact of malaria on mortality, we use mortality rates based on the various predefined categories, which have the advantage of being additive. These rates are expressed in person-years. [30] Figure 2 shows the change by period of the standardized mortality rates [31] before age 20, distinguishing deaths potentially due to malaria, with deaths of unknown cause redistributed proportionally by age and period. Whatever the degree of certainty of malaria mortality, the distinction between the first period and the second is very clear, with mortality potentially due to malaria reaching 3 deaths per 1,000 people aged under 20 per year. Overall, we can thus confirm that there was an increase in malaria mortality in the early 1990s that coincides clearly with a rise in mortality levels. However, in 1995-1999, overall mortality started to fall again, while malaria mortality remained largely stable. Certain or presumed malaria mortality decreases in the last period, while possible malaria deaths remain high.

Figure 2

Variation in standardized rate of mortality potentially due to malaria at ages 0-20 by period, Mlomp, 1985-2003

Variation in standardized rate of mortality potentially due to malaria at ages 0-20 by period, Mlomp, 1985-2003

(per thousand person years)

46To estimate the impact of mortality potentially due to malaria, the contribution of each category to variations in life expectancy from one period to another can be calculated by age group and sex. [32] Life expectancy at birth showed large variations over the period, with a sharp decline in the early 1990s. Female life expectancy, in particular, fell dramatically by 7 years between 1985-1989 and 1990-1994 (Table 6). Female mortality increased at all ages between these two periods, primarily due to higher infant and child mortality. Because the survey protocol has not changed over time, and because this variation affects females only, it cannot be viewed as an artefact, although the especially low female mortality in the first period probably accentuated the effect of a worsening health situation. Malaria may also have affected girls more severely than boys.

Table 6

Sex-specific trends and variations in life expectancy at birth by period, Mlomp, 1985-2003 (in years)

1985-1989 1990-1994 1995-1999 2000-2003 Variations (1) (2) (3) (4) (2)-(1) (3)-(2) (4)-(3) Males 58.4 56.5 55.9 55.1 – 1.9 – 0.6 – 0.8 Females 68.5 61.5 64.5 63.8 – 7.0 + 3.0 – 0.7 Overall 63.1 58.9 59.8 59.2 – 4.2 + 0.9 – 0.6 Source: Mlomp database, 2005.

Sex-specific trends and variations in life expectancy at birth by period, Mlomp, 1985-2003 (in years)

47For the sake of clarity, we have grouped the first two categories of deaths potentially due to malaria (certain and presumed). For males, between 1985-1989 and 1990-1994, mortality potentially due to malaria contributed to the reduction in life expectancy in all age groups, though its effects were largely offset by lower mortality from other non-malarial causes. Between the beginning and end of the 1990s, malaria continued to reduce life expectancy, in early childhood especially. In 2000-2003, progress was achieved in certain and presumed malaria mortality, although deaths potentially due to malaria before age 1 continued to reduce life expectancy (Figure 3). For females, the differences in life expectancy from one period to another are much larger. Malaria was responsible for an estimated 3-year decrease in life expectancy in 1990-1994 with respect to the first period, though the malaria resurgence is not the only reason for this fall. In the late 1990s, the life expectancy decrease due to non-malarial causes had been largely recouped, but not that linked to certain or presumed malaria which, for its part, was only partially recouped in 2000-2003 (Figure 4).

Figure 3

Age-specific contribution of malaria to variations in male life expectancy in Mlomp

Age-specific contribution of malaria to variations in male life expectancy in Mlomp

Figure 4

Age-specific contribution of malaria to variations in female life expectancy in Mlomp

Age-specific contribution of malaria to variations in female life expectancy in Mlomp

48Overall, at ages 0-20, male and female life expectancy at birth fell by more than a year between 1985-1989 and 1990-1994 due to certain or presumed malaria. The drop is respectively 2.4 and 3.3 years if possible malaria is also included. At the end of the 1990s, males lost a further 6 months, while females recovered more than 8 months. In the early 2000s, the potential contribution of malaria changed very little for males, while females gained a year (Table 7).

Table 7

Contribution of malaria to variations in life expectancy at birth between periods by sex and at ages 0-20 in Mlomp

Difference (years) Males Females (2)-(1) (3)-(2) (4)-(3) (2)-(1) (3)-(2) (4)-(3) Malaria certain/presumed – 1.1 – 0.4 + 0.6 – 1.2 – 0.1 + 0.9 Malaria possible – 1.3 – 0.1 – 0.4 – 2.1 + 0.8 + 0.1 Malaria ruled out + 1.9 + 0.4 – 0.3 – 2.3 + 2.1 – 0.6 Periods: (1) 1985-1989; (2) 1990-1994; (3) 1995-1999; (4) 2000-2003. Source: Mlomp database, 2005.

Contribution of malaria to variations in life expectancy at birth between periods by sex and at ages 0-20 in Mlomp