Malarial Risks in Children Age 0 to 14 Years in a Stable Endemic Area of the Kenyan Coast
Source: Snow and Gilles 2002.
Note: Scale of y axis has arbitrarily been fitted to demonstrate relative change in risk by age.
 |
Plasmodium falciparum is the most common of the four human malaria parasites across much of Sub-Saharan Africa. (The other three parasites are P. vivax, P. malariae, and P. ovale.) The distribution of P. vivax is concentrated in the Horn of Africa, covering Djibouti, Eritrea, Ethiopia, Somalia, and Sudan. P. falciparum accounts for almost all the malaria mortality in Sub-Saharan Africa, and it is often stated that the continent bears over 90 percent of the global P. falciparum burden. Recent bioinformatics analysis of changes in human ecology suggest that about 6,000 years ago, P. falciparum populations expanded rapidly in Africa and spread worldwide, coincident with human population growth and subsequent diasporas facilitated by the dawn of agriculture (Joy et al. 2003). This parasite has exacted a heavy mortality toll on Africa's population, evidenced by the selection for several human survival mechanisms, such as the genetic polymorphisms associated with red cell structure and function (Hill 1992).
Malarial Risks in Children Age 0 to 14 Years in a Stable Endemic Area of the Kenyan Coast
Source: Snow and Gilles 2002.
Note: Scale of y axis has arbitrarily been fitted to demonstrate relative change in risk by age.
Figure 14.1 shows a typical age-structured series of risks of infection, mild clinical disease, severe disease, and death due to P. falciparum for a population living on the Kenyan coast (Snow and Gilles 2002). Many individuals naturally acquire functional immune responses to severe disease and death early in life; immunity to the milder consequences of infection occurs later in childhood, but the ability to sterilize blood-stage infection probably does not occur until adulthood.
The relation between the frequency of parasite exposure and disease outcome is complex. The speed with which a population acquires functional immunity to the severe consequences of P. falciparum infection depends on the frequency of parasite exposure from birth as measured by the intensity of parasite transmission in a given locality (see Snow and Marsh 2002). The shape of the severe disease and mortality curves shown in figure 14.1 may therefore be shifted to the right for areas where parasite transmission is of low, stable intensity and to the left for areas where transmission is of high, stable intensity. Where infection is rare the risk of mortality is likely to be directly related to the risk of infection, because acquired functional immunity is unlikely to affect health outcomes. Understanding this relationship is important for defining the age-specific mortality burdens in Sub-Saharan Africa, an area able to support infection rates ranging from one infection every three years to hundreds of new infections per year (Hay et al. 2000).
Public Health Effects of Plasmodium falciparum Malaria
Source: Snow and Gilles 2002.
figure 14.2. In the absence of measures aimed at reducing the risk of infection, the risks shown in figure 14.2 largely depend on extrinsic factors, such as those that determine the speed with which a population develops acquired immunity and the access to effective case management, and on intrinsic factors, such as host genetics.
This chapter describes the determinants and distribution of P. falciparum infection risk in Sub-Saharan Africa and uses populations at risk to estimate mortality from malaria. It also considers the evidence for consequential and indirect mortality and describes P. falciparum as a risk factor for rather than a cause of pediatric mortality. The chapter concludes with a description of the relationship between poverty and malaria and recent trends in malaria mortality in Sub-Saharan Africa to provide some context for current international efforts to halve the malaria burden by the year 2010 (Nabarro and Taylor 1998).
Climate, local ecology, and active control affect the ability of malaria parasites and their anopheline mosquito vectors to coexist long enough to enable transmission. The frequency of transmission, or endemicity, depends on the density and infectivity of anopheline vectors. These features depend on a range of climatic, physical, and population characteristics, for example, rainfall, location of human settlements near or at rivers or other mosquito larval breeding sites, and the density of human populations in a village. The most significant determinant of the intensity of parasite transmission is climate.
The development of both the vector and parasite is temperature dependent. The optimum temperature range for parasite development in the female Anopheles (sporogony) is between 25°C and 30°C, and development ceases below 16°C. Intermittent low temperatures delay sporogony, and the period immediately after the infective bite by the mosquito on an infected human host is the most sensitive to drops in temperature. Above 35°C sporogony slows down considerably. Extremely high temperatures are associated with the development of smaller and less fecund adult mosquitoes. Thermal death of mosquitoes occurs at 40°C to 42°C. Altitude and temperature are strongly correlated: with every 100-meter increase in altitude, the temperature drops by 0.5°C. Overall, the use of altitude as a marker of endemicity or disease risk is vague, yet there is a tendency within the literature to refer to highland malaria in East Africa and the Horn of Africa.
Numerous studies have demonstrated the association between Anopheles gambiae sensus lato (the most important vector of P. falciparum in Africa) abundance and rainfall. Without surface water the female Anopheles cannot lay eggs. Rainfall is also related to humidity and saturation deficit, both affecting mosquito survival (adult vector longevity increases with humidities over 60 percent).
Using the climatic determinants of transmission identified above, the Mapping Malaria in Africa Project (http://www.mara.co.za) developed a series of risk maps for stable P. falciparum transmission across the continent (Craig, Snow, and le Sueur 1999). In brief, the project used long-term mean monthly temperature and rainfall data to define the limits of distribution of stable endemic malaria across Africa. Temperature and rainfall profiles in sample areas, where malaria endemicity was known, were translated into a model of "climate suitability." The temperature limits were related to the requirements for the extrinsic parasite development cycle. The model made the following assumptions: (a) the optimal temperature range was 18°C to 22°C; (b) optimal rainfall values were greater than or equal to 80 millimeters; (c) conditions of rainfall and temperature had to coincide on a month-to-month basis for at least five consecutive months (three months for the northern fringes of Sub-Saharan Africa); and finally, (d) a frost factor (mean monthly minimum temperature of less than 5°C for any one month) was used to eliminate transmission at any point. The model provided fuzzy climate suitability (FCS) values, ranging from zero (unsuitable, hence malaria absent) to one (very suitable, malaria endemic).
During earlier attempts to describe the malaria burden in Sub-Saharan Africa the climate suitability maps for P. falciparum transmission were combined with interpolated maps of population distribution (Deichmann 1996; Snow et al. 1999). The population data were initially constructed using population totals from the last available censuses for administrative units (communities, towns, or districts). These data were then converted into a regular raster grid of population totals, and auxiliary information was used to distribute the population within the administrative unit across its raster grid cells. The process heuristically incorporated information on where people tend to live: in or close to towns and cities, close to transportation infra structure, around protected areas, near water bodies, and not at very high elevations. Using the Geographical Information System (GIS)–based information on the location and size of towns and cities, roads, railroads, navigable rivers, and uninhabitable areas, population density was weighted, a high value implying a high density and a low or zero value implying low or no population. These weights were then used to proportionately distribute population to grid cells. The digital map extracted population distributions according to each cell in a regular raster grid with a resolution of 5 kilometers at the equator. It was then combined with the maps of malaria risk to provide simple population totals of those exposed to stable, unstable, or no malaria risk.
The P. falciparum risk-to-population distribution for Africa has subsequently been refined through use of an improved link between mortality data and transmission intensity (Snow et al. 2003). New criteria were adopted following an improved understanding of the variations in disease outcomes and risks congruent with variations in stable P. falciparum transmission. The relation between the exact number of new infections a population is exposed to each year (annual entomological inoculation rates, or EIR) and parasite prevalence is nonlinear, but quartiles of the prevalence of infection have been shown to approximate logarithmic increases in the EIR (Beier, Killeen, and Githure 1999). This principle has been used previously to categorize pediatric malaria and all-cause mortality according to parasite prevalence estimates from childhood cross-sectional surveys (Snow, Korenromp, and Gouws 2004; Snow and Marsh, 2002). When infection prevalence quartiles (less than 25 percent, 25 to 49 percent, 50 to 74 percent, and 75 percent and greater) are used, mortality indicators saturate at the highest two classes of endemicity and rise sharply from the x–y intercept within the first class. Analysis of 217 independent parasitological surveys among children in Kenya suggests that most areas (81 percent) with a FCS value of less than 0.75 are represented by communities supporting parasite prevalence rates of less than 25 percent (Omumbo et al. 2004). It seems appropriate, therefore, to distribute the population of Sub-Saharan Africa into areas of no risk, unstable transmission risk, low stable risk, and moderate to high stable endemic risk.
Fuzzy Climate Suitability Membership for Malaria
Source: Modified from Craig, Snow, and Ie Sueur 1999.
Note: Areas of no transmission or population: class1, FCS = 0; areas of negligible malaria risk: class 2, FCS > 0 and < 0.25; areas of low, stable endemic risk or epidemic prone: class 3, FCS ≥ 0.25 and < 0.75; and stable endemic malaria areas: class 4, FCS ≥ 0.75.
| Region | Age 0–4 years | Age 5–14 years | Age 153 years | Total |
|---|---|---|---|---|
| North Africa (exclusion) | 17,020 | 33,663 | 95,038 | 145,721 |
| Southern Africa (classes 1–3: no malaria risk) | 4,190 | 10,682 | 29,195 | 44,067 |
| Rest of Africa (class 1: no malaria risk) | 7,366 | 11,680 | 23,033 | 42,079 |
| Southern Africa (class 4: stable malaria risk) | 2,049 | 3,709 | 8,687 | 14,445 |
| Rest of Africa (classes 2, 3: epidemic to low stable risk; FCS < 0.75; parasite prevalence < 25%) | 22,018 | 34,668 | 69,126 | 125,812 |
| Rest of Africa (class 4: stable endemic risk; FCS ≥ 0.75; parasite prevalence ≥ 25%) | 73,351 | 115,261 | 228,105 | 416,717 |
Source: Compiled by authors.
Note: The medium variant assumptions of changing fertility, mortality, and migration on population growth were selected for the year 2000. The database also provided the estimated proportion of the population age less than 5 years, 5–14 years, and 15 years or older and the crude birth rate (CBR). The CBR was used to estimate the approximate number of live births in 2000.
The residents of Comoros, Réunion, and the Seychelles are assumed to be at negligible malaria risk and during the burden estimations formed part of the excluded Sub-Saharan Africa populations. Conversely it seems reasonable to assume that the islands of Cape Verde and São Tomé and Principe are subject to stable transmission; however, they have proved difficult to characterize at effective resolutions using the GIS malaria risk models, and their small contribution to populations at risk (less than 0.5 million people) have not been included.
figure 14.3). First, there are areas with no human settlement, or FCS values of zero (class 1). Some countries have minimal and only very localized risks of malaria transmission, with over 90 percent of their population residing in class 1 areas, the western Sahara region (Algeria, the Arab Republic of Egypt, Lesotho, Libya, Morocco, and Tunisia). It is assumed that these countries have a negligible malaria burden; thus, they are excluded from the Sub-Saharan Africa malaria burden estimations. Second, areas with marginal risks of malaria transmission are represented by a FCS greater than zero but less than 0.25 (class 2). Parasite prevalence rates are likely to be extremely low or zero; however, these populations often border areas of higher infection risk and may be subject to aberrations in climate that lead to localized transmission. Third, in some areas populations are exposed to a FCS value between 0.25 and 0.75 (class 3). Communities living in these areas are likely to support cross-sectional parasite prevalence estimates in childhood of less than 25 percent, and transmission is likely to be acutely seasonal with a tendency toward epidemics. In practice it would be hard to distinguish between classes 2 and 3 in many areas of Sub-Saharan Africa outside of southern Africa, and therefore they are grouped together here. Finally, some areas of the continent have stable endemic malaria transmission (FCS greater than or equal to 0.75), which may still vary seasonally, but support parasite prevalence rates greater than or equal to 25 percent (class 4). In southern Africa (Botswana, Namibia, South Africa, Swaziland, and Zimbabwe) a different approach was taken, because classes 2 and 3 represent the historical extents of transmission and do not reflect contemporary distributions of risk (M. Craig, personal communication). Class 4 areas are the only areas where malaria still poses a risk, and its extent and transmission potential are determined by aggressive vector control. The combined malaria-risk and population models were used to estimate populations resident in each malaria-risk class by using country-level data available from the World Population Prospects Population Database (United Nations Population Division 2001). The summed extractions for the continent are shown in table 14.1.
The current distribution maps of P. falciparum depend entirely on the biotic effects of climate on transmission. They fail to capture the more localized yet marked effects, such as urbanization and localized control. The Anopheles gambiae complex is less prolific in urban areas because there are fewer appropriate breeding sites. People living in urban areas are on average 10 times less likely to receive an infective bite than their rural counterparts (Hay et al. 2000; Robert et al. 2003; Trape et al. 1992). In 2000, 38 percent of Africa's population were urban dwellers (United Nations Population Division 2001). The current estimations of populations at risk under different endemicity conditions (table 14.1) will overrepresent those in the higher-risk classes. New satellite imagery makes it possible to map the extent of land use, land cover, and urbanization, a refinement important for subsequent iterations of the disease burden models for Sub-Saharan Africa (Hay et al. 2005).
Several aggressive, and highly successful, vector control programs in Africa have had significant effects upon transmission (Bradley 1991; Le Sueur, Sharp, and Appleton 1993). More recently, such countries as South Africa and Mozambique have successfully reduced or eliminated transmission in areas that are currently classified as prone to transmission (Barnes and Folb 2003; LSDI 2001). It will never be possible to account for all local vector control activities on a continent-wide map, but it is important to recognize that because current estimations of disease burden will be overestimated.
Other important factors that determine localized transmission intensity are currently not captured in the climate-driven models shown in figure 14.3. These include widespread use of insecticide-treated bednets, leading to a decrease in local parasite transmission in only a few areas of Sub-Saharan Africa; drug resistance, leading to increased parasite transmission carriage; population displacement due to conflict or resettlement, altering transmission characteristics of newly settled areas; man-made ecological changes, such as deforestation; and agricultural practices, such as rice-field cultivation, that change the breeding site availability for local mosquito vector populations. More detailed models of malaria risk in Mali and Kenya have included some of these factors on more recent maps (Kleinschmidt et al. 2001; Snow et al. 1998). However, at present these more complex models are not available at a continental scale.
Finally, because population estimates in the United Nations Population Division's figures are based on estimations incorporating assumptions about fertility and mortality in a country, these totals may differ from the total populations in each administrative unit reported by the country. More refined microcensus data or models of population projections below the national level are required for future mapping of disease burdens.
Malaria is often difficult to diagnose on a purely clinical basis. Fever is common to almost every infectious disease, and the severe pathology caused by P. falciparum, such as acidosis, anemia, and altered consciousness, are also complications of other infections. When a person is ill, demonstrating the presence of malaria infection increases the likelihood that symptoms are directly due to the infection, but the high prevalence of asymptomatic infections makes it difficult to exclude other diagnoses. Our earlier understanding of the pathophysiology of malaria derived from clinical descriptions among adults in Southeast Asia and only recently have the mechanisms of death been more precisely defined for pediatric African populations (Marsh et al. 1995; Warrell 2002). Several detailed clinical studies in African hospital settings have described the principal, sometimes overlapping, routes to a fatal outcome. These include cerebral involvement from sequestered infection in the vasculature of the brain, metabolic disturbances, respiratory distress, and severe anemia. For epidemiological purposes it is convenient to define two major syndromes, cerebral malaria (CM) and severe malarial anemia (SMA). CM is a condition in which patients present in coma with several underlying causes, ranging from a primarily neurological condition to a systemic metabolic disturbance (Marsh and Snow 1999; Newton and Krishna 1998). Severe anemia is a pathology of life-threatening malaria with a complex etiology combining rapid hemolysis during acute infection or a slow insidious process compounded by antimalarial drug resistance. SMA is a life-threatening condition in young children and often warrants blood transfusion.
Africa presents an epidemiological challenge, because although it has high mortality rates, existing vital registration systems are not reliable or comprehensive. Most deaths occur outside the formal health service, and national government systems of civil registration in Sub-Saharan Africa are incomplete. Epidemiologists interested in defining malaria-specific mortality have established demographic surveillance systems (DSS) of large populations (between 20,000 and 100,000 people) to prospectively monitor population migration, births, and deaths (INDEPTH Network 2002; Smith and Morrow 1996).
The attribution of causes of death during DSS surveys is often performed through a verbal autopsy (VA) interview with bereaved relatives about symptoms and signs associated with the terminal illness. The VA details are either reviewed by a panel of clinicians or subjected to diagnostic algorithms. The sensitivity and specificity of VA diagnosis for malaria as a cause of death have been estimated in seven hospital-based validation studies in Africa. There is considerable variation in both the specificity (77 to 100 percent) and the sensitivity (45 to 75 percent) (Korenromp et al. 2003). The sensitivity of verbal autopsies for malaria will depend on the intensity of malaria transmission: in high-transmission areas, severe malaria is more likely to present as SMA than as CM (Snow et al. 1997a). SMA in a young child is often difficult to distinguish from acute respiratory infections (ARI), although CM among older children is less ambiguous. Similarly, the sensitivity and specificity of verbal autopsies for malaria will vary with the local spectrum of other diseases. ARI, acute gastro enteritis, and meningitis all share common clinical features with malaria, including cough, difficulties in breathing, diarrhea, or cerebral dysfunction. Despite these limitations, the VA-diagnosed risks of malaria mortality represent our only source of contemporary information on the direct, fatal consequences of infection in areas with stable transmission in Africa.
The data presented in this chapter derive largely from the Burden of Malaria in Africa (BOMA) project, supported by the Wellcome Trust, United Kingdom, and the Bill & Melinda Gates Foundation. The BOMA project began in 1998 and has completed a comprehensive search for published and unpublished empirical measures of malaria morbidity, disability, and death. The data include historical accounts of mortality from colonial administration records and detailed descriptions of recent DSS surveys. Details of the search methods and applications of the data are presented elsewhere (Snow, Korenromp, and Gouws 2004; Snow and Marsh 2002; Snow, Trape, and Marsh 2001). The estimates of malaria-specific mortality in childhood outside of southern Africa use only mortality reports based on a VA within a DSS framework after 1989. This time restriction was made because of recent temporal changes in pediatric risks of mortality from malaria in Sub-Saharan Africa (see "Trends in Malaria Mortality" section). Where DSS and VA methods were used during community or household randomized intervention trials, only control communities were included. In order to structure risks according to transmission classes shown in table 14.1, DSS surveys of malaria-specific mortality were also limited only to those in which there was a congruent estimation of malaria transmission intensity.
Adult malaria-specific mortality data from DSS sites are rare, and VA methods have not been well developed nor widely validated for the description of malaria deaths in this group. At Kilifi, Kenya, in the absence of obvious signs of respiratory disease an algorithm was used based on detailed exclusion criteria (for example, well-defined deaths or hospital-diagnosed conditions) involving an acute febrile event (R. W. Snow, unpublished data). In Tanzania a similar approach was used (Y. Hemed, personal communication). Both approaches are likely to overestimate the number of adult deaths from malaria in endemic areas. Owing to the paucity of empirical data in this age group, these two contemporary estimates have been included with other sources of historical data on adult malaria mortality from detailed, preindependence civil registration systems when causes of death were investigated by medical personnel from circumscribed, censused populations. It seems reasonable to assume that under stable transmission conditions adult mortality from malaria is unlikely to have changed much over time, as deaths would normally occur in individuals who, for some ill-defined reasons, failed to acquire functional immunity in childhood or lost their functional immune response. The intention has been to attempt, where possible, to fill an important data gap using imperfect data, and therefore the risks reported must be interpreted with caution.
Civil and vital registration systems in southern Africa are more comprehensive than those in other parts of Sub-Saharan Africa. Data from southern Africa, however, also underreport events and must be viewed as minimum estimates. Data have been extracted from reports provided during subregional malaria control program meetings and malaria deaths (reported only for all age groups combined) and expressed per projected population estimates for administrative areas of known malaria risk (figure 14.3). Data were available for the three areas of KwaZulu-Natal Province in South Africa, seven areas of Botswana, four areas of Namibia, and eight areas of Zimbabwe.
| Region | Age 0–4 years | Age 5–14 years | Age 153 years | Bibliographic sources |
|---|---|---|---|---|
| Southern Africa (malaria risk areas; class 4) | 0.13 [0.08–0.21] (study sites 22) | Diseko 1994; Kamwi 1998; Piotti 1997; South African Medical Research Council 1998, unpublished data | ||
| Rest of Africa (low, stable epidemic risk; classes 2, 3; FCS < 0.75; parasite prevalence < 25%) | 2.62 (287/109,412 PYO) (study sites 3) | 0.94 (158/167,598 PYO) (study sites 2) | 0.71 (265/372,777 PYO) (study sites 2) | Charlwood et al. 2001; Government of Tanzania 1997 |
| Rest of Africa (stable endemic risk; class 4; FCS ≥ 0.75; parasite prevalence ≥ 25%) | 9.33 [7.38–14.57] (study sites 12) | 1.58 [0.66–2.77] (study sites 10) | 0.6 [0.37–0.94] (study sites 15) | Delacollette and Barutwanayo 1993; Ghana VAST 1993; Jaffar et al. 1997; Snow, Mung'ala et al. 1994; Trape et al. 1998; Barnish et al. 1993; Premji et al. 1997; Government of Tanzania 1997; Salum et al. 1994; D'Allesandro et al. 1995; Pasha et al. 2003; Government of Colony of Gold Coast 1912–48; Colbourne and Edington 1956; Government of the Colony and Protectorate of Kenya 1935; Government of Colony and Protectorate of Nigeria 1934–35; Bruce-Chwatt 1952; Government of Colony and Protectorate of Sierra Leone 1913–17 |
Source: Compiled by authors. For low, stable endemic risks for "rest of Africa," G. D. Shanks and R. W. Snow provided unpublished data in Kenya; additional information from the Ghana VAST project were provided by the late Nicola Dollimare and Gilly Maude; unpublished VA data reconstructed by R. W. Snow to provide estimates of malaria-specific mortality in the study described by Snow, Mung'ala et al. 1994; additional information provided by Brian Greenwood for the Barnish et al. 1993 study.
Note: First numbers in each entry represent the median; numbers in square brackets represent the IQR; PYO = person-years of observation. For southern Africa, the figures in the first column include all age groups.
| Region | Age 0–4 years | Age 5–14 years | Age 153 years | Total |
|---|---|---|---|---|
| Southern Africa (malaria risk areas; class 4) | 1,878 | |||
| Rest of Africa (low, stable epidemic risk; classes 2 & 3; FCS < 0.75; parasite prevalence < 25%) | 57,688 | 32,588 | 49,079 | 139,355 |
| Rest of Africa (stable endemic risk; class 4; FCS ≥ 0.75; parasite prevalence ≥ 25%) | 684,364 [541,330–1,068,723] | 182,113 [76,072–319,274] | 136,863 [84,399–214,419] | 1,003,340 [701,801–1,602,416] |
Source: Compiled by authors.
Note: The total for southern Africa includes all age groups. Numbers in square brackets represent the IQR.
figure 14.3 and table 14.1. Median and interquartile ranges (IQR) have been used to describe the risk data per malaria risk class (table 14.2), and the estimated numbers of deaths directly attributed to P. falciparum in 2000 are shown in table 14.3 according to populations at various continental risk classes shown in table 14.1.
Twelve independent estimates of malaria mortality among children under five years living under conditions of stable transmission were used from DSS sites in Burundi, The Gambia, Ghana, Kenya, Senegal, Sierra Leone, and Tanzania. The median malaria-specific mortality rate among these communities was 9.33 per 1,000 children per year and represented 28.2 percent of all mortality among children under five. Under similar transmission conditions the median malaria-specific mortality among older children, age 5 to 14 years, was 1.58 per 1,000 per year (from 10 study sites), representing 52.2 percent of all deaths in this age group. Among adult populations surveyed as part of colonial administration civil registration systems or recent DSS sites in stable endemic areas, the median malaria mortality rates were 0.6 per 1,000 per year (15 reports), or 6 percent of all deaths among populations age 15 years or older.
There are few DSS or colonial administration civil registration data from low, stable endemic or epidemic conditions (parasite rate less than 25 percent; classes 2 and 3) outside southern Africa. Approximations to the DSS from the populations of a tea estate in the highlands of Kenya (Shanks and Snow, unpublished data) and a settled refugee camp in Sudan (Charlwood et al. 2001) provided data on malaria-specific mortality rates among censused childhood populations at low risk of malaria infection. In addition, the one true DSS site at Hai district in Tanzania provided a VA estimate of malaria mortality (Government of Tanzania 1997). All three studies covered periods after 1990. Only the Kenyan and Tanzanian studies provided information on older children and adults. Rather than being used to provide a median estimate from the limited data, the combined person-years of observation (PYO) and malaria deaths have been used to define direct risks of death from malaria under these transmission conditions. For children under five years the overall mortality rate was 2.62 per 1,000 per year (three studies); for children age 5 to 14 years the mortality rate was 0.94 per 1,000 per year (two studies); and for adults age 15 years and over the rate was 0.71 per 1,000 per year (two studies).
An ongoing rural DSS site at Agincourt, South Africa (Kahn et al. 1999), located at the fringes of malaria risk (Brink 1958), also provided additional data. A malaria-specific mortality rate of 0.065 per 1,000 per year (2 out of 216 deaths) was measured among children under five years between 1992 and 1995. Of the 785 deaths recorded in the population older than five years of age, only one was attributed to malaria. This single estimate is an inadequate representation of the malaria mortality across southern Africa. The civil registration data in malaria-risk districts of southern Africa suggest that the median estimate of malaria-specific mortality among the entire population in these areas was 0.13 per 1,000 per year. Although these estimates represent minimal approximations of the true mortality burden, they fall within the ranges of mortality described for the single DSS site (Kahn et al. 1999) and those sites described in the more detailed analysis of malaria mortality data from civil registers in two districts of KwaZulu-Natal Province between 1996 and 1999: 0.02 to 0.52 per 1,000 people per year (Tsoka, Sharp, and Kleinschmidt 2002).
Overall malaria-specific mortality in children is approximately 3.5 times higher in areas of stable endemic transmission than in areas of low intensity, stable, or epidemic-prone malaria in Sub-Saharan Africa, excluding southern Africa. Mortality declines rapidly with increasing age, and this is especially striking under conditions of stable endemic transmission. The mortality rates for all ages from P. falciparum in southern Africa are considerably lower than those described for the rest of Africa, reflecting a low risk of infection combined with effective control. Even with the use of all the tools described earlier in this chapter, approximately 1.14 million people might have died in Sub-Saharan Africa as a direct consequence of infection with P. falciparum in 2000.
The consequences of disease that are related to the clinical event include the consequences of clinical management, such as the immediate effects of adverse drug reactions or the longer-term effects of HIV-acquired infection through blood transfusion. Nonintervention-related consequences of clinical events also include the short- and long-term residual impairments resulting from cerebral malaria.
Drugs used to manage malarial fevers have adverse effects. We can assume that most of the severe adverse drug-related events are described as direct mortality during DSS VA studies, as they are likely to occur close to the febrile event. They are mentioned here, however, to emphasize that fevers are common in Africa and the use of antimalarial drugs is prolific. Adverse drug reactions (ADR) are likely to increase with the greater use of new, more complex drug combinations and agents, although the increase of ADR will probably never surpass the alternative of using ineffective drugs. There have been too few epidemiological studies of the human toxicity of many antimalarial drugs among African populations repeatedly exposed to these compounds. Most data derive from an examination of chemoprophylactic drug use among nonimmune travelers (Philips-Howard and Bjorkman 1990; Philips-Howard and West 1990). The majority of adverse reactions due to sulfonamides and 4-aminoquinolines are idiosyncratic. Severe adverse reactions to the commonly available antimalarial drugs in Africa, when used as recommended, include severe cutaneous reactions (Stevens-Johnson and Lyell's disease syndromes), aplastic anemia, severe neutropenia, thrombocytopenia, keratopathy, agranulocytosis, and hepatic failure (Reynolds 1993). It has been estimated that 2,350 deaths were probably caused by treatment rather than the disease itself in a single year among the children outside of southern Africa (Snow et al. 2003).
Severe anemia is a common feature of children hospitalized with complicated malaria. Transfusion is a common pediatric practice in Africa and adherence to guidelines for transfusion are often poor (English et al. 2004; Lackritz et al. 1992). Greenberg and colleagues (1988) examined the HIV and malaria status of 167 admissions of children to an emergency ward at the Mama Yemo Hospital in Kinshasa, the Democratic Republic of the Congo. The authors propose an unadjusted odds ratio for acquired HIV infection of 3.5 for malaria patients between 1 month and 12 years of age transfused once, 21.5 for those transfused twice, and 43.0 for those transfused three times during a single admission. In a study of transfusion practices in 1994 at six government hospitals in Kenya, Moore and colleagues (2001) calculated a 2 percent risk of transmission of HIV antibody positive blood from screened donations through blood transfusion to HIV-negative patients. This study did not include the risk from pre-seroconversion donations, nor did it allow for the sensitivity of test kits or the use of unscreened blood. The probability that an HIV antibody negative unit of blood is HIV infected has been estimated to be between 0.5 and 1.1 percent (Savarit et al. 1992). The probability of seroconversion following HIV-contaminated blood is assumed to be 96 percent (Colebunders et al. 1991). From estimates of hospitalized patients with severe malaria anemia, transfusion rates, and relative risks of seroconversion from HIV-infected blood donation, it has been estimated that between 5,000 and 8,000 children might become infected with HIV each year as a consequence of poor management of their malaria in the hospital (Snow et al. 2003).
The case-fatality rate of cerebral malaria in most hospital settings is high, often over 30 percent (Newton and Krishna 1998). Prolonged coma and seizures are associated with neurological impairment in survivors. The immediate and prolonged sequelae associated with CM among African children includes hemiparesis, quadriparesis or severe deficit, hearing and visual impairments, speech and language and nonverbal construction difficulties, behavioral problems, and epilepsy (Mung'ala-Odera, Snow, and Newton 2004). These impairments are estimated to occur in about 4,000 children each year in Sub-Saharan Africa.
Those with severe deficits have a higher mortality risk soon after the disease event. For less dramatic impairments, such as epilepsy, increased mortality rates have been described (Coleman, Loppy, and Walraven 2002; Jillek-All and Rwiza 1992; Snow, Mung'ala, et al. 1994). In Western countries, the risk of premature mortality could be as high as two to three times that described in age-comparable groups without epilepsy (Cockerell et al. 1994; Hauser, Annegers, and Elveback 1980; Zielinski 1974), but in Africa the risk could be as high as nine times (Coleman, Loppy, and Walraven 2002). This is likely to be caused by poorly managed epilepsy resulting in status epilepticus or accidents, such as drowning or burns.
Indirect consequences of P. falciparum infection include anemia (unless anemia is linked to acute high-density parasitemia as a direct cause), low birthweight, growth retardation, or undernutrition. In addition, malaria infection can increase the severity of other comorbid infectious diseases through immune suppression or enhanced invasive capacities across physical barriers to infection (for example, blood and tissue). Previous approaches to the global burden of disease have assumed that each death must be attributed to a single cause and can be fitted into the fixed disease-mix matrix of all causes (Murray and Lopez 1997).
During randomized controlled intervention trials aimed at reducing the incidence of infection (but not 100 percent protective), the all-cause mortality of children is often reduced more than would be attributed by VA diagnosis of malaria. For example, in Kilifi the proportion of deaths of children under five years attributed to malaria by VA was 34 percent (R. W. Snow, unpublished data). During a randomized controlled trial of insecticide-treated bednets in the same area, the incidence of malaria infection was reduced by 50 percent (Snow et al. 1996), which was sufficient to reduce all-cause mortality by 33 percent (Nevill et al. 1996). More dramatically, in The Gambia, insecticide-treated bednets reduced all-cause mortality by over 60 percent, and yet the VA-diagnosed contribution of malaria to all-cause mortality among control populations was only 16 percent (Alonso et al. 1993). This has led some to speculate that malaria infection is a contributor to broad causes of mortality beyond the direct fatal consequences of infection (Molineaux 1997).
Attempts have been made to distinguish between factors that affect the direct outcomes of malaria infection and the effects of malaria infection on the outcome of other health burdens. This is particularly important when the interaction between malaria and HIV or undernutrition is considered. Especially important is the role malaria infection plays as a risk for the extended, indirect consequences of infection on the public health burden posed by P. falciparum. A recent model relates infection prevalence to all-cause pediatric mortality outcomes in Sub-Saharan Africa (Snow, Korenromp, and Gouws 2004). The model regards P. falciparum infection as a risk factor for all-cause mortality rather than attempting to directly estimate malaria's contribution to all-cause mortality (one cause, one death).
Despite a poor understanding of the precise mechanisms of pathology (Menendez 1995), the morbid outcomes of malaria infection during pregnancy have been well described (Brabin 1983; Guyatt and Snow 2001a Guyatt and Snow 2001b; Steketee, Wirima, Hightower et al. 1996; Steketee et al. 2001). In endemic settings in Africa, pregnant women experience relatively little malaria-specific morbidity (for example, fever) but do have increased risk of infection and higher density parasitemia leading to anemia and placental sequestration of the parasite. These effects operate across a broader range of endemicities used to describe morbid and fatal risks among nonpregnant populations (that is, areas covered by classes 3 and 4 shown in figure 14.1). Maternal anemia has been shown to be an important contributor to maternal mortality with a relative risk for mortality of 1.35 for moderate anemia and 3.51 for severe anemia (Brabin, Hakimi, and Pelletier 2001). Brabin and colleagues further estimate that malaria contributes to maternal anemia and that 9 percent of anemia-associated maternal mortality can be attributed to malaria in Sub-Saharan Africa; this figure predicts approximately 5,300 maternal deaths annually from malaria anemia in areas of Sub-Saharan Africa outside of southern Africa classified as classes 3 and 4.
Prematurity and low birthweight (less than 2,500 grams) are associated with maternal malaria, including the contribution from both malaria-associated maternal anemia and placental infection. The contribution of malaria during pregnancy to low birthweight and subsequent mortality in the first year of life has been estimated to range from 3 to 8 percent of infant mortality (Greenwood et al. 1992; Steketee, Wirima, Hightower et al. 1996; Steketee et al. 2001). If this range had been applied to the expected numbers of live births in 2000 among populations in low endemic, epidemic prone, and endemic malaria areas of Sub-Saharan Africa outside of southern Africa, there may have been between 71,000 and 190,000 infant deaths indirectly attributable to malaria in pregnancy.
In very low endemic settings or areas where malaria is epidemic prone and in southern Africa, malaria in nonimmune pregnant women can be devastating and lead to maternal death and abortion. For example, in urban Mozambique, 15.5 percent of all maternal deaths in one hospital over a five-year period were attributed directly to malaria (Granja et al. 1998). In an epidemic-prone setting of Ethiopia, maternal malaria carried an approximately eightfold increased risk of abortion (R. Newman, unpublished data). Similarly, in a low-endemicity setting in Southeast Asia even single malaria infections were associated with increased risk of low birthweight (Luxemburger et al. 2001).
Anemia among African children is a hematological state determined by combinations of nutritional deficiencies (iron, folic acid, other micro nutrients, and protein-calorie malnutrition), iron loss through helminth infection, red cell destruction, red cell production decreased by infectious diseases, and the genetic constitution of red cell hemoglobin (Menendez, Fleming, and Alonso 2000; Nussenblatt and Semba 2002). Malaria has long been recognized as a major contributor to anemia in children, reducing hemoglobin concentrations through several mechanisms. The primary mechanism increases rates of destruction and removal of red blood cells and decreases the rate of erythrocyte production in the bone marrow. Other mechanisms are associated with acute clinical states (for example, hemolysis or cytokine disturbances), whereas chronic or repeated infections are more likely to involve dyserythropoiesis (Menendez, Fleming and Alonso 2000). Data from randomized controlled trials of malaria-specific interventions to reduce the incidence of new infections through insecticide-treated bednets or the prevalence of blood-stage infections through chemoprophylaxis or intermittent presumptive treatment suggest a halving of anemia risks through intervention (Korenromp et al. 2004). This substantial reduction in the risks of anemia demonstrates the importance of P. falciparum in maintaining the poor hematological health of children. This finding, however, cannot be extrapolated to subsequent morbid or fatal consequences, because few detailed prospective studies focus on describing the effects of reduced hemoglobin concentrations on health outcomes.
It has been postulated that nutritional status is related to the threats posed by infection and disease and to the role of infectious disease in perpetuating undernutrition (Pelletier, Frongillo, and Habicht 1993; Pelletier et al. 1995). Most studies have generally focused on severe malnutrition and specific nutrient deficiency, and only a few have examined the role of malaria. One striking feature of the global distribution of anthropometric markers of undernutrition is its congruence with the distribution of endemic malaria. Although P. falciparum malaria and malnutrition are both highly prevalent in Sub-Saharan Africa, the existence of a synergistic interaction has not been well established. Evidence from intervention trials aimed at reducing the frequency of new infections suggests that malaria infection might have some indirect effects upon the generalized nutritional status of African children. A study in Nigeria on the use of chemoprophylaxis in the treatment of malaria in children showed a reduction in the incidence of infection and clinical attacks that was accompanied by a reduction in the incidence of malnutrition (Bradley-Moore et al. 1985). Improved growth among young children has more recently been demonstrated in The Gambia and Kenya in studies comparing those protected by insecticide-treated bednets with those left unprotected (D'Allessando et al. 1995; Snow et al. 1997b; Ter Kuile et al. 2003). Despite the biological plausibility of synergism between infection and growth, the precise relationship between undernutrition and severe malaria continues to be difficult to quantify empirically within disease burden frameworks.
In Sub-Saharan Africa the HIV epidemic has been superimposed on the long-standing malaria pandemic. The wide geographical overlap and the concurrent high prevalence of both HIV and malaria mean that even modest interactions could substantially affect public health among populations exposed to both (Chandromohan and Greenwood 1998). Studies during the 1990s observed that malaria infection was more common and of higher parasite density in HIV-positive than in HIV-negative pregnant women in a range of malaria endemic settings, and in women of all gravidity, and that those who have had multiple pregnancies were most affected (Parise et al. 1998; Shulman 1999; Steketee, Wirima, Bloland et al. 1996; van Eijk, Ayisi, and ter Kuile 2001, p. 405; van Eijk et al. 2001; Verhoeff et al. 1999). Additionally, two longitudinal cohort studies in Uganda and Kenya and one hospital-based case-control study in Uganda have demonstrated that HIV-infection approximately doubles the risk of malaria parasitemia and clinical malaria in nonpregnant adults, and that increasing HIV-immunosuppression is associated with higher-density parasitemias (Francesconi et al. 2001; French et al. 2001; Whitworth et al. 2000). Thus, some evidence shows that HIV infection increases the incidence and severity of clinical malaria in adults.
In order to define the indirect consequences of P. falciparum it is important to know whether the high intensity of exposure to malaria infections increases the rate of progression of HIV disease in Africa. In a recent study in Malawi, HIV blood viral levels were found to be seven times higher in HIV-infected adults with acute uncomplicated malaria than in HIV-infected blood donors without malaria (Hoffman et al. 1999). As with other acute infections, the increased viral burden was reversed by effective malaria therapy (Hoffman et al. 1999). These findings are consistent with in vitro laboratory studies in which HIV-1 replication was increased 10-fold to 100-fold in peripheral blood mononuclear cells exposed to soluble malaria antigens or malaria pigment (Xiao et al. 1998).
The evidence on the potentiation of malaria by HIV or HIV by malaria is currently insufficient to quantify the specific HIV-malaria interaction risks for inclusion in malaria burden estimates.
The effects of malaria infection on low birthweight seem conclusive; the overall role of malaria infection on anemia is clear, but its extrapolation to indirect mortality is difficult; the effects of infection on undernutrition and HIV is at present less conclusive and impossible to enumerate. It seems reasonable to assume that these additional consequential or indirect risks contribute to all-cause mortality beyond that described from estimates of the causes of death directly attributed to malaria.
To examine these conclusions and assumptions further, data on all-cause mortality of children under five from DSS studies undertaken across a broad range of malaria transmission settings in Sub-Saharan Africa were analyzed against the prevalence of P. falciparum infection at each site. Weighted least-squares regression was used to model the contiguous relationships between all-cause mortality and parasite prevalence rates, allowing for the square of parasite prevalence (for possible saturation of parasite prevalence), timing, location, and the sampling precision of each study (Snow, Korenromp, and Gouws 2004). The unadjusted median all-cause child mortality rate for low prevalence areas of childhood infection (less than 25 percent) was 10.9 per year per 1,000 children under five (IQR 7.8–17.6). This rose dramatically to 39.1 per year per 1,000 children (IQR 32.8–52.2) among populations exposed to childhood parasite prevalence risks greater than or equal to 25 percent. In the regression model, mortality increased significantly with parasite prevalence, but this effect leveled off at higher prevalence rates. The model suggested that, in rural DSS sites throughout Sub-Saharan Africa, all-cause mortality increases by more than twofold (25–30 deaths per 1,000 children under five years old) over the prevalences of malaria infection covered by the DSS sites, and parasite prevalence explained 64 percent of the variation between sites in all-cause under-five mortality. By contrast, the direct estimation of malaria-specific mortality presented earlier for children living under stable endemic conditions was only 28.2 percent.
These comparisons of direct versus indirect contributions of malaria infection to child survival must be viewed with some caution. Ecological analyses are constrained by multiple confounders, notably, influences of socioeconomic status, access to and effectiveness of health services, prevalence of HIV, exposure to other parasitic diseases, nutrition, and the genetic makeup of the population. Still, such an approach to estimating malaria's contribution to child mortality in Sub-Saharan Africa suggests that the influence of P. falciparum infection is greater than that described by the direct, single-cause attribution of childhood deaths. The effects of malaria during pregnancy on low birthweight and to a lesser extent the nutritional and hematological influences of malaria infection may explain the difference between direct and indirect estimates.
Recent global, cross-country regression analysis of malaria risk and gross domestic product have found the disease to be a significant influence on long-term economic growth. As much as half the gross domestic product is lost in highly endemic areas over 25 years as a direct result of malaria (Gallup and Sachs 2001; McCarthy and Wu 2000). Although persuasive to international donors, such a macroeconomic analysis fails to identify the mechanisms of these economic losses (Malaney, Speilman, and Sachs 2004). It seems reasonable to assume that national levels of economic loss are a composite of economic burdens at the household level. Shepard and colleagues (1991) estimated the household cost of malaria in Burkina Faso, Chad, the Republic of Congo, and Rwanda. The authors concluded that a case of malaria in Africa cost US$9.84 in 1987, of which US$1.83 was direct and US$8.01 was accrued indirectly as a result of forgone income associated with malaria morbidity and mortality. The total estimated cost of US$0.8 billion represents 0.6 percent of the gross domestic product of the economies in Sub-Saharan Africa. Thus, microeconomic analyses of household economic burden posed by malaria are only a fraction of macroeconomic analyses of national economic loss. Malaney, Speilman, and Sachs (2004) suggest that this difference might arise either because of inadequate methodologies or because the macroeconomic national burden encompasses household financial costs and ill-defined financial "externalities," such as trade, tourism, and investment.
Whether malaria drives household or national poverty requires better definition. However, the converse is irrefutable: poor people are less able to prevent infection or afford effective disease management. In Africa, malaria is largely a disease of the rural populations, and often these communities are home to some of the poorest of the poor in Africa. There is increasing evidence that strategies promoted to prevent infection, such as insecticide-treated bed-nets, are not reaching the poor when cost-retrieval is part of the strategy. The recent Kenyan Demographic and Health Survey showed that less than 7 percent of children described as living in households at the lowest wealth index quartile sleep under an insecticide-treated bednet compared with 35 percent of children in the top wealth quartile households (http://www.measuredhs.org/). Similar findings have been reported for Uganda (Mugisha and Arinatwe 2003). In Tanzania, poor children were less likely to receive antimalarials when febrile than children from wealthier families (Schellenberg et al. 2003). A household survey in Malawi focused on low-income households whose mean annual income was US$115 and where the costs of malaria prevention and treatment represented about 20 percent of annual income (Ettling et al. 1994).
Malaria-Specific and All-Cause Mortality Estimates per Year for Children under Five (per 1,000)
Source: Snow, Trape, and Marsh 2001.
Note: Central horizontal lines 3 median quartile range; box width and T= median and upper/lower limits of mortality estimates, respectively. Single outlier is more than three times the box width.
figure 14.4. The period 1960–89 was characterized by a median malaria-specific mortality of 7.8 per 1,000 children under five per year, a decline of 18 percent from the estimates recorded before 1960 (9.5 per 1,000 children under five per year). However, the 11 studies of childhood malaria-specific mortality during the 1990s provided a median estimate of 10.2 per 1,000 children under five per year, similar to that described in the preindependence surveillance data (figure 14.4a). In contrast, the annualized risks of all-cause mortality recorded at the same sites demonstrate a continued decline over the three time periods (figure 14.4b), and overall childhood mortality during the early 1990s was 34 percent lower than the median estimates of all-cause mortality recorded during the preindependence surveys. The net result has been a rise in the proportion of all deaths attributed to malaria: 18 percent before 1960, 12 percent between 1960 and 1989, rising to 30 percent during the 1990s.
Malaria-Specific, Nonmalaria, and All-Cause Mortality Rates among Children under Five (per 1,000 per year from 28 DSS sites)
Source: Adapted from Korenromp et al. 2003.
Note: Data analyzed by least-squares linear regression to allow for interactions of determinants of VA-adjusted mortality, the square of parasite prevalence, and region.
Figure 14.5 summarizes the combined Sub-Saharan Africa data for 1982–98, allowing for parameters of region, parasite prevalence, and VA performance.
There are several possible explanations for the observed trends in malaria mortality in Sub-Saharan Africa over the last 20 years, such as declining household wealth linked to a changing health sector based on cost retrieval and the general deterioration in the quality of clinical care. Although access to and quality of health care is undoubtedly suboptimal in many areas, this has not had a similar impact on non-malaria childhood mortality. Global warming has been debated as a general cause of expanding malaria risk and thereby increasing malaria-specific mortality in several parts of Africa. However, this seems an unlikely explanation for the areas included in figures 14.4a and 14.5, as long-term data for areas likely to be affected by changes in ambient temperature do not support a trend in favor of increased malaria transmission (Hay et al. 2002). For many years chloroquine provided a cheap, effective, and easily available treatment. The past 20 years have seen a precipitous decline in the efficacy of chloroquine across Africa (EANMAT 2003; Talisuna, Bloland, and D'Alessandro 2004), and this represents the most likely factor contributing to the change in malaria-specific childhood mortality. Such an explanation would also be entirely consistent with observations of increased hospitalization and morbidity in other parts of Africa (Asindi et al. 1993; Greenberg et al. 1989; Shanks et al. 2000) and the sharp rises in malaria morbidity associated with declining first-line therapy in KwaZulu-Natal Province of South Africa and its reversal following the introduction of effective artemisinin-based combination therapies (Barnes and Folb 2003).
During 2000 approximately 1.14 million people may have died as a direct result of infection with P. falciparum. Eighty-eight percent of these deaths would have occurred in areas of stable endemic malaria and the majority of these would have been among young children. The empirical evidence for indirect or consequential mortality may explain an additional 10 percent of mortality directly attributed to malaria infection. This would be consistent with observations made during randomized controlled trials that suggest that reducing the risks of infection in a community has an impact beyond what might readily be described by verbal autopsy as direct malaria mortality. Recent ecological analyses of all-cause pediatric mortality suggest that the difference between direct and indirect attribution of malaria as cause of death might be substantially higher than 10 percent.
The Roll Back Malaria movement proposes to halve malaria mortality by the year 2010 (http://www.rbm.int). This goal has been set even though existing, affordable therapeutics are rapidly failing, health service provision is breaking down, there are no immediate prospects of widespread vaccination, and poverty continues to afflict most endemic countries. There are strong reasons to believe that over the past 15 years malaria-specific mortality has risen and now accounts for an increasing proportion of overall childhood mortality. The starting point for new efforts to "roll back malaria" is not a level playing field but a mortality burden that has returned to levels described before Africa gained independence.
One million deaths due to malaria each year in Africa resonate with earlier claims of a similar figure proposed as far back as the 1950s (Bruce-Chwatt 1952; Greenwood 1990; Schwartlander 1997; Sturchler 1989). These estimations are hard to comprehend without a methodological framework or empirical evidence to support them. How much further forward are we in estimating malaria's contribution to the mortality burden in Sub-Saharan Africa? The estimates provided in this chapter continue to be driven by informed approximations, in part because of the paucity of reliable and accurate data, but also due to the inherent difficulties of unique diagnosis. The estimates, however, have been made from empirical epidemiological measures of mortality risks, structured according to age and malaria transmission. The method is similar to one previously used to define the malaria burden in Sub-Saharan Africa (Snow et al. 1999) but takes into account new data, refined endemicity classifications, broader health consequences, and the temporal effects associated with changing antimalarial drug sensitivity. This data-driven method allows for a more transparent review of the evidence.
Such reviews of available evidence reveal as much about what we do know as about what we do not know. Many aspects of the malaria burden in Sub-Saharan Africa require further empirical investigation or modeling. In the continued absence of wide-area, reliable coverage of cause-specific mortality data there is a need for increased research to help in an understanding of the spatial determinants of risk. The assumption that urban versus rural, East versus West African, or poor versus affluent populations experience similar risks of poor health outcomes from malaria infection is clearly an oversimplification of complex interactions. The precise cartography of risk requires improved population distribution maps. This is now being addressed at a global scale through a new initiative, the Malaria Atlas Project, funded by the Wellcome Trust, United Kingdom (http://www.map.ox.ac.uk). The emphasis within the literature on measuring childhood risks of disease and death from malaria would seem justified from the current understanding of immunity, but there is comparative ignorance about the true public health consequences of malaria among older children and adults. The impact of infection and disease on undernutrition, anemia, and HIV remain at best speculative and require further investigation before these composite risks can be defined. The most significant challenge facing the malaria epidemiologist today is to describe malaria health outcomes in Africa and how these risks change from birth through adulthood according to the dependent factors of infection, immunity, and control.
Free Full text in PMC]Free Full text in PMC]Free Full text in PMC]Free Full text in PMC]Free Full text in PMC]
This review was made possible because of a large contribution over the years from malaria epidemiology colleagues, notably Marlies Craig, Uwe Deichmann, Simon Hay, Eline Korenromp, Claire Mackintosh, Kevin Marsh, Charles Newton, Dennis Shanks, Rick Steketee, and Jean-François Trape. The Burden of Malaria in Africa Project is principally supported by the Wellcome Trust, United Kingdom (project number 058992), with additional support from the Bill & Melinda Gates Foundation (project number 17408); the Disease Control Priorities Project, World Bank; and the Kenyan Medical Research Institute. Robert W. Snow is supported by the Wellcome Trust, United Kingdom, as a senior research fellow.
