Estimates of Deaths from Malaria

Deaths are not well counted in much of the world, and the situation is worst where people are poorest—which is where malaria takes its greatest toll. This has begun to change—Snow and colleagues (2003) date recognition of the importance of reliable estimates to the World Health Organization's (WHO's) Global Burden of Disease program of the early 1990s—but the figures available are still approximate.

Some 50 years ago, Leonard Bruce-Chwatt estimated the annual African death toll from malaria at one million. The figure was an extrapolation to the continent based on the civil registration of deaths in Lagos in 1950 (Bruce-Chwatt, 1952). Other numbers have been produced since then— between 0.5 million and 2 million deaths per year—using a variety of more and less evidence-based methods (Sturchler, 1989; Greenwood, 1990; WHO, 1996; Schwartlander, 1997). The most plausible current estimates come from Snow and colleagues, and their work on the Burden of Malaria in Africa (BOMA) project (Snow et al., 2003), which builds on the MARA (Mapping Malaria Risk in Africa) risk mapping, as well as mining of a wide range of other data sources. Their best estimate is 1,144,572 deaths attributable directly to malaria in Africa in the year 2000 (Table 7-1).

TABLE 7-1

Estimated Malaria-Specific Mortality during 2000, best estimate and IQR range.

WHO's most recent estimate of malaria deaths is similar—a worldwide total of 1,124,000 deaths due directly to malaria in 2001, of which about 970,000 would have been in Africa (WHO, 2002). The estimate also includes about 90,000 malaria deaths in Southeast Asia, 56,000 in the Eastern Mediterranean region, 11,000 in the Western Pacific, and about 1,000 in the Americas. While in the same overall range, the estimates do differ in their distribution among age groups: Snow and colleagues figures suggest that about 65 percent of deaths among children under 5, and the corresponding figure from WHO is much higher at 86 percent.

Estimates of Malaria Cases

There are no direct counts of the number of cases of malaria that occur each year. A very wide range of estimates has been made, using a variety of definitions. The lack of precision is problematic for this report, because an estimate of how many courses of malaria treatment are used is needed to estimate how much a global ACT subsidy will cost. The number of cases and the number of treatments are not the same but are related.

Malaria cases may be overestimated because few cases are definitively diagnosed before treatment (whether prescribed or self-selected), so quite a large number of fevers that are not malaria may be counted as such. The number also may be underestimated because people may get no treatment, either through choice or economic default, and in any case, episodes of malaria are not kept track of formally anywhere.

Snow and colleagues, using the general pattern of risk and population as they used to estimate deaths for the Disease Control Priorities Project, also estimated the number of malaria episodes likely to occur in Africa in a year (the year 2000). In this case, they used the BOMA database to search for prospective studies of the incidence of fever (i.e., axillary or rectal body temperature ≥ 37.5°) since the 1980s.¹ Studies were included if they met the following criteria:

• They were cross-sectional studies with at least two rounds of observations on a single cohort to reflect seasonal differences in risk.
• Subjects were recruited randomly from the community, and not from a clinic population.
• For cohorts involved in randomized trials, only the control arm participants (i.e., those not receiving a study intervention) were included.
• Clinical episodes were defined by quantified parasite levels (although levels differed among studies, ranging from 1,000 to 10,000 parasites/µl blood).

Definitions and methods of detection differed among the studies, and it was not possible to standardize them; some lack of comparability and imprecision had to be accepted.

The results are summarized in Table 7-2, ending with a median estimate of 213,549,000 episodes per year in Africa (with the estimate ranging from 134,322,000 to 324,617,000). Among young children in stable endemic areas, data from 28 studies led to a median estimate of 1,424 episodes per 1,000 children each year (i.e., nearly 1.5 episodes per child). The 19 studies that included older children led to an estimate of 587 per 1,000 (on average, about one episode every other year). The median estimate for non-pregnant adults (based on only seven studies) was 107 per 1,000 people each year. Rates in low-transmission and fringe areas were correspondingly lower.

TABLE 7-2

Estimated Number [IQR] of Clinical Attacks (thousands) of Malaria during 2000.

The Effect of Antimalarial Drug Resistance on the Burden of Malaria

Given the fragmentary statistics on malaria morbidity and mortality, it may seem presumptuous to attempt an assessment of how the spread of resistance to chloroquine and sulfadoxine-pyrimethamine (SP) has affected these measures over the past decades. There is, in fact, little direct information on how the number of cases of malaria has changed in the most endemic areas. What has been documented, however, are stark bellwethers of worsening conditions: a well-documented instance is the malaria epidemic of 1999-2000 in KwaZulu Natal, South Africa, which was a direct result of failing antimalarial drugs—SP had a failure rate of 88 percent. Other factors—increased vector resistance to the pyrethroid insecticides that were introduced in 1996, and the reinvasion of the highly anthropophilic Anopholes funestus vector—further exacerbated this epidemic (Muheki et al., 2003). The introduction of ACTs (and other control measures) brought the epidemic under control. And the question of changes in mortality has been addressed by thorough reviews of the data that do exist, in two major efforts. The first, the long view across the 20th century, used a variety of historical records from the BOMA project (Snow et al., 2001). The other case is an examination of trends in the 1980s and 1990s on a finer scale, contrasting East and West Africa (Korenromp et al., 2003), using the relatively uniform data reported in African Demographic Surveillance Systems (DSS).

Trends Through the 20th Century

The BOMA project provides the best opportunity to chart malaria's past in Africa, and how drug resistance has affected its course over the final decades of the 20th century. The BOMA project began in 1998 with the aim of assembling in a single database all available evidence on morbidity, disability, and mortality associated with falciparum malaria in Africa, starting as far back as possible. The data come not only from the usual electronic databases, but from hand searches of early, unindexed papers in English and French tropical medicine journals, and a mass of unpublished material from local and regional conference proceedings, libraries, and Ministries of Health records (Snow et al., 2001).

In 2001, Snow and colleagues selected as much information as possible from the BOMA database to inform an analysis of malaria mortality in Africa from 1900 through the 1990s (Snow et al., 2001). They took mortality reports from areas with documented, stable endemic transmission, where the prevalence of parasitemia in children was at least 30 percent, and which had recorded both all-cause mortality and malaria-specific mortality. Thirty-nine studies in 13 countries of sub-Saharan Africa ² met the criteria, spanning the period from 1912 to 1995. A year-by-year analysis was not possible with these scattered, sparse data. Instead, the time span was divided into three periods, corresponding approximately to changes of probable significance to malaria control. The period before 1960 represents a time of limited access to primary health care and hence, to effective antimalarial drugs. From 1960 until 1990, after the beginning of independence for most countries, health care expanded across Africa and chloroquine became widely available, both from health services and as self-medication. The 1990s saw the widespread emergence of chloroquine resistance in many parts of Africa.

The picture painted by these data suggests a continuing downward trend in total child mortality over the three periods, but a downward and then ascending course for malaria-specific mortality, with the lowest rates in the middle period, and similar rates pre-1960 and post-1990 (Figure 7-1). With a 34 percent decline in overall mortality from before 1960 into the 1990s, and the fall and rise of malaria death rates into the 1990s, the proportion of all deaths due to malaria first fell from 18 percent pre-1960 to 12 percent in 1960-1990 but rose to 30 percent during the 1990s. Snow and colleagues (2001) cite data from Tanzania, Senegal, and Kenya comprising more detailed time series, which are consistent with the findings overall.

FIGURE 7-1

Malaria mortality and all cause mortality per 1000 children under 5 years in three time periods: pre-1960, 1960–1989, and 1990–1995. Central thick lines represent median estimates, box height represents 25%–75% quartile range, (more...)

The data used to describe these trends, are by their nature, limited and not entirely comparable. The pre-1960 data are mainly from colonial Anglophone Africa, where malaria deaths were tracked through civil notification systems operating in defined populations. The pre-1960 systems probably missed a greater proportion of deaths than the later prospective surveillance studies, which have high rates of ascertainment of the fact of death. However, identification of deaths from “malaria” may actually have been more accurate in the earlier period because deaths often were followed up by a medical officer to determine their cause. The later surveillance systems rely on assigning cause of death retrospectively, mainly through verbal autopsies.

The declines in childhood mortality during the second half of the 20th century are well documented and accepted. The biggest declines have been in deaths from diarrheal disease, widely attributed to the development and dissemination of oral rehydration therapy, largely through WHO-sponsored programs. Declines also occurred in deaths from pneumonia and other infectious diseases. The declines are not uniform, and there are even places where trends are reversing in overall child mortality, but by and large, the trends have been positive. Malaria appears to be an exception.

How can the observed trends in malaria mortality among African children be explained? The positive effects on total mortality are a plausible result of expanded basic health services, including vaccinations, the widespread adoption of oral rehydration therapy, expanded access to antibiotics, and at least in places, improved sanitation and general living conditions. In the mid-20th century, chloroquine would have counted among new beneficial drugs, even in the absence of a major campaign against malaria. The rise in malaria-specific mortality coincides with the spread of chloroquine-resistant strains of falciparum malaria. Although it is virtually impossible to demonstrate a causal link, in the absence of any other compelling reason, chloroquine resistance is the most plausible explanation for the increase.

Trends in the 1980s and 1990s

Resistance to chloroquine first became apparent in Africa in the late 1970s, more than a decade after its appearance in Asia. By 1990, resistant strains had been reported from all the endemic countries in Africa (Trape, 2001). The spread of drug resistance was implicated in the previous section as responsible for reversing the improvement in childhood mortality from malaria. The inference was made because mortality patterns in large population groups coincided with the general patterns of drug resistance, but the measurements of drug resistance were not available to make the link definitive.

There are data—albeit limited—that address the question more directly, although again, not definitively. Trape (2001) identified population-and hospital-based studies that recorded annual malaria mortality in Africa, based on continuous monitoring. He chose for analysis all those that spanned a period during which chloroquine resistance was known to be emerging in that area. Some of the hospital-based studies also had information on the prevalence of severe malaria over time. The remainder of this section is based on Trape's report.

Population-Based Studies

Three studies tracked childhood malaria mortality in populations over the period in which chloroquine resistance emerged, from the mid-1980s to the mid-1990s. All were in Senegal, in different climatic areas: Mlomp area (rain forest), Niakhar area (Sahel), and Bandafassi area (savanna). In all three areas, the malaria mortality rates increased from the early to the later periods (Table 7-3). They more than doubled in all three sites and in Mlomp, increased 10-fold. Over the same period, standardized clinical protocols for assessing drug failures were carried out at least twice. In all cases, the degree of resistance intensified.

TABLE 7-3

Childhood Malaria Mortality in Three Sites in Senegal.

In one other place, Bagamoyo, on the Tanzania coast, deaths were investigated in 1984-1985, and again in 1992-1994 (i.e., not continuous, as in Senegal), bounding a period of increasing chloroquine resistance. Overall child mortality remained the same, but the proportion attributed to malaria was twofold higher in the later period.

Hospital-Based Evidence

Data from hospitals in several countries offer another glimpse at trends in malaria incidence and death, with definite limitations. A hospital population consists only of those who decide—often through a complex decision-making process—to seek care there. Trends in the proportion of people with malaria who go to a hospital may relate to changes in incentives for hospital-based versus other sites of care (e.g., the population may know that hospitals are out of drugs at a particular time, a competing provider may be more attractive, or money for the hospital is unavailable in a given season or year) and changes in the incidence of other diseases, as well as to changes in the epidemiology of malaria. Hospitals also have advantages because of their range of expertise and technologies, and their record keeping.

Malawi. Malawi has national records for hospital admissions and deaths. From 1978 through 1983, the incidence of admissions for malaria for children under 5 years of age more than doubled. Throughout the period, about 5 percent of those children died (Khoromana et al., 1986). The spread of chloroquine resistance in Malawi during this time was well documented.

Tanzania. The mission hospitals in Tanzania recorded steep increases in the proportion of admissions for malaria from 1968 through 1985. In the 1970s, about 10 percent of admissions were for malaria, and by the mid-1980s, had risen to 23 percent (Kilama and Kihamia, 1991). This also coincides with the rapid increase in chloroquine-resistant malaria in Tanzania.

Congo. In the hospital that was the main referral center in Kinshasa, the proportion of overall pediatric admissions for malaria increased each year, from 29.5 percent in 1982, to 56.4 percent in 1986, and the proportion of deaths increased from 4.8 to 15.3 percent over the same period (Greenberg et al., 1989). Again, this was the period during which chloroquine resistance emerged and spread quickly in Kinshasa.

Malaria admissions and deaths at the four hospitals in Brazzaville were studied for the years 1983-1989, during which (in 1985) chloroquine resistance was first detected there. From 1983 through 1986, pediatric admissions for malaria increased from 22 to 54 percent of all pediatric admissions, and then remained stable. Deaths from cerebral malaria more than doubled from the first to the second half of the study period (Carme et al., 1992).

Nigeria. A similar pattern was seen in the pediatric emergency room of Calabar Hospital in Nigeria, where the number of cases of malaria-related convulsions doubled during the years 1986 through 1988. In 81 percent of these cases, chloroquine was ineffective (Asindi et al., 1993).

Recent Estimates of the Costs of Malaria

The remainder of this chapter describes recent work on defining the economic costs of malaria. The macroeconomic effects—more elusive but unquestionably greater than the aggregate of currently measured microeconomic effects—include economic losses to nations from lack of foreign investment, drains of human capital, and other large-scale effects that hamper overall economic development. The microeconomic costs most often measured include the direct expense, to both government and individuals, of preventing and treating the disease, and the indirect costs of being sick with malaria.

The Effects of Malaria at the Macroeconomic Level

Malaria and poverty occupy common ground. Where the burden of malaria is highest, economic prosperity is lowest. Both are concentrated today in tropical and subtropical zones. Quantifying the relationship between malaria and country-level economics has posed an analytical challenge, first addressed just a few years ago. Analyses approached with several different datasets representing malaria and economic indices confirmed a strong relationship, although the magnitude of effect has varied. A recent review by one of the leaders in the field, Jeffrey Sachs, estimates that the average per-capita gross domestic product (GDP) of malaria-endemic countries in 1995, at about $US1,500 (adjusted for purchasing power parity³), was roughly one-fifth the average across the nonmalarious world (Sachs and Malaney, 2002). Annual economic growth in malarious countries between 1965 and 1990 averaged 0.4% of per-capita GDP, compared with 2.3% in the rest of the world, after controlling for the other standard growth determinants used in macroeconomic models. Over the long run, this decrement suggests that malaria could reduce GDP by nearly one-half in highly endemic countries.

These analyses do not constitute proof that malaria is a cause of low incomes and poor economic growth, but that the disease must be considered at least a legitimate contributor, and possibly the major contributor. It is true at the microeconomic level that poverty can lead to a heavier burden of malaria, and that malaria can deepen poverty, if a lack of money equals an inability to protect oneself against malaria or to properly treat it. Relationships at the macroeconomic level are real, but—possibly because the chains of causation are obscured—easier to dismiss (Sachs and Malaney, 2002). Some of the ways in which malaria may hinder economic development are discussed in the sections that follow. An empirical study of the effect of reducing the numbers of malaria cases and related deaths in Vietnam (Box 7-2) found a generalized positive effect on household living standards (as measured by expenditures), which was greatest for areas with the greatest malaria reductions, and less so for areas with lesser malaria improvement (Laxminarayan, 2003).

BOX 7-2

During the 1990s the government of Vietnam invested in malaria control mainly through insecticide, re-treatment of bednets, and by providing effective treatment for clinical episodes. The result was a dramatic overall decline in cases of malaria, but (more...)

Some Pathways Through Which Malaria May Hinder Economic Development

Microeconomic Studies of the Effect of Malaria⁴

Most microeconomic studies of malaria use the “human capital” approach, which involves tallying the apparent costs of malaria borne privately by individuals and their families, by various levels of government, and by other providers of services (e.g., nongovernmental organizations of various kinds, organizations financing malaria programs). Other methods to get at the costs of malaria are possible, including the “willingness-to-pay” approach, an example of which is described later in this section.

In the human capital approach, the immediate costs of treating or preventing an episode of illness consist of the “direct” costs—money spent for malaria prevention and treatment, including, for example, the cost of bednets and mosquito repellents in the first instance, and of consulting a health provider, buying drugs, and paying for transportation, in the second. Indirect costs represent loss of income (or productive labor, even if the benefits are not monetized, e.g., lost agricultural production because of an inability to plant or harvest crops) due to illness. Indirect costs also include periods of being too sick to work, time spent caring for others who are sick, and time spent seeking care. Some drawbacks of the method, as used, are discussed after a review of the findings of recent studies.

Indirect Costs

When people are too sick to work, or their work capacity is reduced by illness, there are economic consequences: wage earners are paid less; agriculturists may produce less (particularly if illness coincides with the harvest); sick children themselves cannot work (even some very young children work) and may cause parents to lose work time. Deaths from malaria also have clear economic consequences. The death of an adult has obvious economic consequences, leaving families without needed resources. The death of a child involves a different set of economic losses. Illness and death from malaria diminish the stock of “human capital” in many ways. Conly (1975) conducted what is considered a classic study of the effect of malaria on agricultural settlers in Paraguay, finding major effects on the productivity of families (Box 7-3).

BOX 7-3

Conly's Classic Microeconomic Study in Paraguay. In the early 1970s, Conly spent 22 months in the field observing and recording data on malaria (and health more generally), work of all kinds, and finances among farmers in Paraguay in newly-settled areas (more...)

Studies of the indirect costs of illness vary in their methods, and in what is included as time lost. Although some include only time spent seeking treatment, most include some period of morbidity; some also include the costs of mortality (lifetime lost income) (Table 7-4). The value placed on lost time also varies among studies (as it does in life), but in all cases, simplifying assumptions are made to try to capture average experience, generally by age group. In the relatively few studies that attempted to calculate indirect costs of malaria, the cost for each episode ranges from US$0.68 for children under 10 years of age in Malawi (Ettling et al., 1994) to US$23 for an adult in Ethiopia (Cropper et al., 1999).

TABLE 7-4

Calculations of Productivity Costs of Morbidity from a Malaria Episode Using the Wage Rate Method.

Drawbacks of the Human Capital Method for Estimating Costs of Malaria

The studies reported here vary widely in the details of the methodology, their sources of data, and in their perspectives. Comprehensive studies of this type are difficult to conceive and to carry out, given inherent limits on data available in the places most affected, general difficulties in carrying out research in such places, and a lack of funding for such studies. Even when done well, however, the methods often require assumptions about such things as employment levels (which affect whether wages are actually lost), the value of leisure time, and substitutability of work by other families. Some of the aspects not generally captured by the human capital method relate to coping strategies adopted by families, which affect their economic well-being in ways not usually measured through direct and indirect costs usually measured.

The Willingness-to-Pay Approach to Disease Valuation

Individuals and families regularly make decisions with economic consequences, like deciding to seek treatment for disease as opposed to watching and waiting, deciding whether or not to buy a bednet to prevent malaria, and deciding to plant one crop versus another with different labor requirements and different economic yield when harvested. Clearly, economic realities constrain these decisions, although some are forced upon families, such as a child in convulsions, who must be given attention.

The willingness-to-pay approach attempts to estimate the costs of disease by determining the value people would place on avoiding it. The assumption is that this approach would incorporate the burden of treatment and prevention, of lost productivity, lost leisure time, and the value of not having to cope with malaria in other ways. But the way in which this valuation happens is subject to individual interpretation, and assumes that people's economic choices are rational and related to costs and other consequences. Nonetheless, the amount people are willing to pay to avoid disease—if captured in a reliable way—is a useful measure of how important the disease is, and can help make it a priority for public policy.

Studies of this type are very difficult to carry out reliably. The study described here is one of the few examples in malaria.

Results

Malaria and the Cost of Illness Malaria was common. More than half of the respondents (58 percent) reported having malaria in the previous 2 years; in half of the households, another adult (besides the respondent) had had malaria, and in half, a child or teenager had had malaria. Using the detailed information collected on these episodes, the costs of malaria per episode and per household were calculated (Table 7-5). Direct costs to the household included out-of-pocket expenditures to see a health practitioner, buy medicines, and pay for transportation, amounting to about US$1.60 for an adult episode, and about half that for a child. Indirect costs of illness are based on productive time lost by patients themselves, other family members caring for patients, and family members substituting for patients at their work. The average adult loss per episode was 21 days. Dollar values were assigned to the days lost, for adults, equivalent to the average daily wage for a healthy, unskilled worker. Under the “high productivity” assumption this amounted to US$24 for an adult, and under the “low productivity” assumption, half that amount. The rates for teenagers were half the adult rates, and for children, one quarter.

TABLE 7-5

Cost of Malaria to Households in Tigray, Ethiopia.

Willingness to Pay to Prevent Malaria As would be expected, for both the hypothetical vaccine and bednets, the lower the price offered to respondents, the more said they would buy, and the number they would buy was greater the lower the price (keeping in mind that each respondent was offered the intervention at one specific price). For vaccines, at the lowest price offered (US$0.80), three-quarters of households would buy at least one dose, but at the highest price (US$32), only 10 would buy at least one. For bednets, at the lowest price (US$1.30), about 80 percent of households would acquire at least one, and at the highest price (US$16), about 40 percent.

These results were compared with the cost-of-illness figures using three different models (see original paper for details), with similar results: people appear to be willing to pay about three times the annual cost of illness for total malaria prevention (the vaccine). For partial prevention (bednets), willingness to pay was less, about 72 percent of the value for vaccines.

Summing Up

The “true” economic costs of malaria are undeniably large, but just how large is not known. Admittedly, the information base is small, which accounts for part of the problem, but the methods themselves are not as well developed as needed. Adding up all of the effects from the best microeconomic studies using the human capital method, the totals do not begin to approach the magnitude of effect seen with top-down macroeconomic approaches. The human capital method, as practiced, underestimates costs in ways that are identifiable, as well as some that have yet to be defined. With macroeconomic methods, it is impossible to know whether other factors, inadequately controlled for, are inflating the numbers.

Understanding both the magnitude of malaria's economic effects as well as its operative pathways will accomplish two goals. It will better place malaria in its appropriate economic context and it will improve strategies by which to combat it.