Chapter 26:  Maternal and Perinatal Conditions

Definitions

The terms maternal and perinatal encompass a continuum of health states—from the most positive (complete physical, mental, and social well-being) to the most negative—and a huge number of clinical conditions. This chapter focuses on eight major conditions, hereafter referred to as the focus conditions, which are estimated to account for about 75 percent of maternal deaths and more than 60 percent of perinatal deaths. For the mother, these conditions are hemorrhage, sepsis, hypertensive disorders of pregnancy, obstructed labor, and unsafe abortion. For the baby, they are low birthweight, birth asphyxia, and infection (table 26.1).

We define maternal conditions as encompassing events occurring from conception to 42 days postpartum (WHO 1992a). The chapters on women's health, family planning, adolescent health, and surgery address the longer-term sequelae of pregnancy and childbirth; the preconception period; pregnancy at an early age; and specific interventions, such as repair of obstetric fistulas. Within the period from conception to 42 days postpartum, two broad categories of conditions can be distinguished: those arising specifically from pregnancy and parturition (direct obstetric conditions), and those aggravated by or aggravating to pregnancy (indirect obstetric conditions). Because the latter conditions, such as malaria, HIV/AIDS, or anemia, are not exclusive to pregnant or parturient women, they are not dealt with here but in the relevant disease-specific chapters.

Regarding perinatal conditions, we focus on those for which interventions can be directed to the baby through the mother during pregnancy or delivery. Our discussion is complemented by the discussion in chapter 27, which concentrates on the neonate, including special care of the small baby and emergency care of the sick newborn.

Formal definitions of perinatal conditions tend to vary by data source. Taken literally, they refer to conditions that arise in the perinatal period (Murray and Lopez 1998), which are not the same as events that occur in the perinatal period—that is, from 28 weeks of gestation to the end of the seventh day of life. For example, death resulting from conditions that arise in the perinatal period can happen at any age, although it tends to take place during the neonatal period (up to 28 days of life). By contrast, perinatal deaths include both stillborn babies and those who are born alive but die before the end of the seventh day. Early neonatal deaths only include live births.

Nature and Characteristics

Pregnancy and childbirth are not inherently pathological. Maintaining an effective balance, however, between preserving normality and ensuring a state of readiness to deal with abnormality represents a fundamental challenge to health systems and a tension in safe motherhood programming. Although this balance between prevention and treatment is not peculiar to maternal and perinatal conditions (or complications), the following additional characteristics are relevant to assessing the burden as well as the effectiveness of interventions:

• The principle of "first, do no harm" has particular significance in this area, because many preventive practices related to pregnancy and childbirth can readily become harmful in unskilled hands—for example, inappropriately early induction of labor or poor forceps technique. The iatrogenic burden of maternal and perinatal conditions is rarely factored into assessments of intervention effectiveness.

• The lives of two individuals, mother and baby, are potentially at stake (Stoll and Measham 2001); however, interventions will not necessarily benefit both equally, and indeed, some will be in direct conflict.

• A large number of maternal and perinatal conditions present clinically not as single entities but as complexes, such as hemorrhage and sepsis or preterm delivery and birth asphyxia. For the mother, the situation may be further complicated by the role of underlying conditions, such as HIV/AIDS underlying puerperal sepsis.

• The most extreme negative outcome, death of both the mother and the baby, is highly concentrated around the time of delivery, from the onset of labor or abortion to 48 hours postpartum or postabortion. Estimates indicate that about two-thirds of maternal deaths occur within this time window (AbouZahr 1998), and the proportion for perinatal deaths appears to be even higher (Bale and others 2003). For the mother, however, a growing number of studies highlight the contribution of direct and indirect causes of deaths, including violence, when a one-year postpartum reference period is used (Etard, Kodio, and Traore 1999; Hoj and others 2003).

• The initial clinical presentation of some conditions can be severe, with rapid escalation to a life-threatening state, and these conditions often require surgical intervention.

• A distinct clinical feature of some maternal conditions is their unpredictability (AbouZahr 1998). This fact has had a profound effect on the prioritization of interventions in safe motherhood, and it is an area in urgent need of further research. The situation is confused by the alternative endpoints, such as death or disability, and by the extent to which there are clear and predictable risk factors. Table 26.1 summarizes some of these key characteristics as they relate to the eight focus conditions.

Causes and Conceptual Frameworks

One of the most frequently quoted figures in safe motherhood is that 88 to 98 percent of maternal deaths are avoidable with moderate levels of health care (WHO 1986). This advocacy statement simplifies the multiple pathways leading to death and, thus, the multiple opportunities for primary and secondary prevention. In part, this simplicity is a further reflection of the grouping together of clinical conditions that in reality are distinctly different in terms of prevalence, case fatality, and scope for intervention, such as eclampsia and puerperal sepsis or congenital anomalies and birth asphyxia. The multiple endpoints and conditions, for both the mother and the fetus or newborn, have implications for what is regarded as an antecedent (a cause, a determinant, or a risk factor)¹ and what is regarded as a consequence (an outcome or a sequela).

A large number of conceptual frameworks depict pathways to adverse maternal and perinatal outcomes (Bale and others 2003; McCarthy and Maine 1992). Several identify three levels of contributory factors, which are also found in causal models for general health outcomes (WHO 2002): (a) distal, (b) proximal or intermediate, and (c) physiological or direct. Table 26.1 highlights the risk factors for the focus maternal and perinatal conditions. The distal determinants emphasize that maternal and perinatal well-being is not just a medical issue. Improvements throughout the health sector must be complemented by attention to wider social, economic, and cultural factors as well as to reproductive rights (CMH 2002). Many conceptual frameworks also differentiate between the timing of interventions: before pregnancy, during pregnancy, during labor and delivery, or during the postpartum period. Similarly, a further distinction can be made in terms of the timing of the outcome, although from a programmatic perspective, such a temporal focus may lead to fragmented care for women and their babies.

Levels, Trends, and Differentials

The latest regional estimates of maternal mortality are for 2000–1 (table 26.2), with most of the figures for the developing world produced by modeling (WHO 2004b). More than 99 percent of annual maternal deaths occur in the developing world. At a national level, the magnitude of the differential in terms of lifetime risk is almost 500-fold between the highest figure for a developing country (1 in 6) and the lowest estimate for a developed country (1 in 29,800) (WHO 2004b). This differential is often cited as the largest discrepancy between the developing and developed world of all public health statistics, reflecting major differences both in obstetric risk, as measured by the maternal mortality ratio, and in levels of fertility, as reflected in the total fertility rate.

In terms of medical causes of maternal mortality, even greater caution is needed regarding the reliability of any patterns observed, because of their dependence on whether the data are health service based or population based and on coding conventions. Figure 26.1a shows the percentage distribution among direct causes at a crude global level. Direct causes account for about 80 percent of all maternal deaths, with indirect causes responsible for the remainder. Of the direct causes, hemorrhage is generally regarded as the most common and may be underestimated, because health facilities are unaware of many such deaths, given the short interval between onset and death (see table 26.1). In terms of indirect causes, the pattern varies enormously between different parts of the world, primarily according to the prevalence of HIV/AIDS, malaria, and tuberculosis.

The published data on severe maternal morbidity are weaker still. A recent World Health Organization (WHO) systematic review indicates how prevalence figures vary hugely according to the criteria used to identify cases (Say, Pattinson, and Gulmezoglu 2004). Using disease-specific criteria, WHO found that prevalence ranged from 0.80 to 8.23 percent. Using organ system criteria, WHO found that the range was 0.38 to 1.09 percent. Finally, using management-based criteria, WHO found that the range was 0.01 to 2.99 percent. Estimates suggest that for every maternal death, at least 16 or 17 other women suffer a life-threatening complication during pregnancy or childbirth (Gay and others 2003) and at least 30 women are left with long-term disabilities, such as an obstetric fistula (UNFPA 2003). These estimates must be regarded as crude approximations, most originating from small-scale studies and most in urgent need of updating and verification. Given the varying case fatality rates shown in table 26.1, the fact that the distributional pattern for morbidity (figure 26.1b) does not completely mirror the one for mortality is not surprising.

As concerns mortality in babies, an estimated 5.7 million perinatal deaths occur each year, 47 percent as stillbirths and 53 percent in the first week of life (J. Zupan, personal communication, August 25, 2004). Many of those deaths are linked directly with complications experienced by the mothers, and several studies have shown that the survival prospects for a baby whose mother dies are generally poor—less than 1 percent in one study in Bangladesh (Koenig, Fauveau, and Wojtyniak 1991). In 2004, neonatal deaths represented 36 percent of all deaths of children under five in developing countries, with about 1 million of these 3.94 million neonatal deaths occurring in the first week of life (Jamison and others 2004). Table 26.3 presents modeled estimates for early neonatal deaths in 2001. The data on the magnitude and patterns of stillbirths remain particularly poor.

Given weak sources of information, the dearth of reliable trends data is hardly surprising. At a global level, a major difficulty arises from the need to use models to estimate maternal mortality. As the basic methodology for the models has changed over time, the data are not appropriate for trend assessment. AbouZahr and Wardlaw (2001) provide patchy support for downward trends in some parts of the world, mostly on the basis of civil registration data and mostly restricted to countries with maternal mortality ratios of less than 100 per 100,000 live births—thus notably excluding South Asia and Sub-Saharan Africa. Even where declines appear to have occurred, they did so prior to 1990. Countries with sustained falls since then, such as Argentina and China, cannot be regarded as representative of all developing countries. Cause-specific trend data are extremely rare, often gathered through small-scale hospital-based studies or special inquiries (see, for example, Pattinson 2002). Recent WHO (2004c) statistics on unsafe abortion show an apparent decrease in incidence in all world regions, although the risk of death remains high at 50 per 100,000 live births, and in parts of Sub-Saharan Africa the risk is as high as 140 per 100,000 live births (Rogo, Bohmer, and Ombaka 1999). These adverse events, however, are often also the most seriously under-reported, as elaborated further in chapter 57.

The availability of reliable trends data for perinatal mortality is even more problematic. A demand for population-based estimates for newborn mortality is comparatively recent; thus, there has been insufficient time to accumulate multiple data points. Demographic and health surveys (DHSs) are a key source for tracking trends in infant and child mortality. Several DHSs now have data that can be disaggregated to show neonatal deaths, but only a few have information on stillbirths, and the quality of that information is still being assessed. Information from WHO suggests that early neonatal death rates fell slightly, from 28 per 1,000 live births around 1980 to about 25 per 1,000 in 2000, for low- and middle-income countries, and the equivalent trend for stillbirths is suggested to be a drop from 36 per 1,000 deliveries to 22 per 1,000 deliveries (J. Zupan, personal communication, August 25, 2004).

Two types of differentials are particularly relevant: geographic (or regional) and socioeconomic. Table 26.2 indicates the wide variation in the magnitude of maternal mortality across regions, and a similar difference can be seen between countries. In terms of absolute numbers of deaths, just 13 countries account for 70 percent of the global total (WHO 2004b).² Caution is again needed, because the poorest countries also have the weakest information systems and, therefore, have estimates derived solely from modeling. One regression model (WHO 2004b), for example, uses independent variables, such as the percentage of deliveries with health professionals present and the proportion of deaths of women of reproductive age that are maternal deaths. Those variables are themselves subject to error and likely to be least reliable where information systems are weakest. Geographic differences in maternal mortality within countries are poorly documented, although remote populations are often assumed to suffer the highest levels because of poor access to emergency obstetric care. Although this assumption seems logical, few reliable data are available to confirm or refute it, and the possibility of high levels of mortality in urban areas linked to unsafe abortion (Thonneau and others 2002) makes the topic of geographic differentials a priority for research.

Until recently, socioeconomic differentials in mortality have tended to be inferred from utilization patterns for prenatal care and health professionals at delivery. The DHSs continue to provide the main data sources in this regard, for both international and national analyses, and they demonstrate huge differences between wealth quintiles. A relevant recent development, however, is the familial technique, which can be used to examine socioeconomic differences in maternal mortality using existing survey data (Graham and others 2004). Because maternal health and health care are clearly associated with stillbirths and early neonatal deaths, the same differentiating factors are likely to apply to perinatal outcomes. Indeed, data from many DHSs show large gaps between rich and poor in relation to neonatal mortality, with the greatest average disparity being found in Latin American and the Caribbean (http://www.worldbank.org/poverty/health/).

Attributable Burden

The estimation of maternal and perinatal conditions as part of international assessments of the burden of disease has long been controversial, and much has been written about the problems and potential distortions of priorities (AbouZahr 1998; Sadana 2001). Some of those criticisms relate to methods of valuation based on disability-adjusted life years (DALYs), especially in relation to discounting and the omission of stillbirths, and others to the inaccuracies and selectivity of the base data on the incidence of complications, on case fatalities, and on disabilities. Table 26.4 presents DALYs for South Asia and Sub-Saharan Africa for the focus conditions for 2001. Those two regions together account for 74 percent of the global burden of maternal conditions and 64 percent of the global burden of perinatal conditions.

The significance of the burden of maternal and perinatal conditions is clear from two recent global assessments (CMH 2002; WHO 2002). The approaches the two initiatives adopted have led to different conclusions about public health priorities. The former focused on avoidable mortality resulting primarily from direct obstetric conditions, whereas the latter considered population risk assessments and highlighted the contribution of indirect obstetric problems—especially micronutrient deficiencies—and the role for preventive strategies. Clearly, the choice between different measures of burden has a crucial influence both on the strategic approach to achieving health gains and on the prioritization of interventions.

Levels and Types of Interventions

Box 26.1 presents one example of a comprehensive strategy for safe motherhood. It illustrates the range of programmatic issues raised by maternal and perinatal conditions:

• the scope for both primary and secondary prevention

• the difference between the individual receiving specific interventions (here, the mother) and the beneficiary (the baby)

• the multiple effects of single (component) interventions on different outcomes

• the multiple benefits to the same outcome of different interventions

• the short- and long-term time frames for interventions and outcomes

• the balance between supply-side and demand-side interventions

• the role for interventions outside the health sector.

Three main pathways are available for averting adverse outcomes: preventing pregnancy, preventing complications, and preventing death or disability from complications. The first pathway is the only truly primary preventive strategy. It requires intervention to avert the occurrence or mistiming of pregnancy by means of effective family-planning methods, as discussed in chapter 57. This preventive approach is relevant for those women who are able to and wish to avoid or delay pregnancy, but it has a limited role for those not in this position, estimated at between 15 and 57 percent of women age 15 to 29 (WHO 2002). As concerns the primary prevention of complications, comparatively limited reliable evidence is available on the true size of the avoidable fraction for many conditions at a population level. The emphasis in this preventive pathway is on maintaining normality and on managing mild complications—and thus on good quality of care. Finally, maternal death and disability may be avoided by effective, timely, and appropriate clinical interventions, often referred to as emergency obstetric care.

Given this complexity and the multiple approaches used to address maternal and perinatal conditions, no perfect framework for categorizing interventions exists. We, therefore, cluster the alternative intervention pathways on the basis of the following three parameters:

• level of care—home, primary, and secondary

• time period—pregnancy, labor and delivery, and postpartum

• strategic approach—population-based versus personal interventions.

Quality of Evidence

Pregnancy and childbirth have been the subjects of medical investigation for centuries and, indeed, are among the oldest clinical specialties. As a consequence, a substantial body of opinion exists on the signs, symptoms, etiology, prognosis, natural history, and management and treatment options for many maternal and perinatal complications, particularly in developed countries. Much of it can be regarded as conventional wisdom acquired through practice. In contrast, a comparatively small proportion of interventions can be regarded as based on evidence, by contemporary scientific standards, and arrived at through the conduct of robust research. Thus, in specification of the content of intervention clusters, a built-in tension exists between using the best available knowledge and using only evidence that passes minimum quality criteria. Equally important is recognizing the fundamental distinction between knowing what is effective at an individual case-management level, for which an evidence base exists for maternal and perinatal conditions, and demonstrating effectiveness at the aggregate levels of composite strategies and entire countries or regions, for which robust evidence is extremely limited (Graham 2002).

Population-based Interventions

The primary aim of population-based interventions is to reduce the risks leading to adverse outcomes at the population level rather than at the individual level (WHO 2002). Population-based interventions are essentially preventive and seek to promote healthy behaviors, thereby reducing incidence in the entire population. In the case of maternal and perinatal conditions, such an approach could be adopted for two major risk factors: lack of contraception and maternal undernutrition. The grade of evidence for these population-based interventions is primarily level C for the former, but a mixture of A and B for the latter.³

Personal Interventions

When we consider interventions directed at individuals rather than whole populations, the need for a continuum of care for mother and baby in terms of time (before and after delivery), place (linking home and health services through an effective referral chain), and person (the provider of care) is important. A variety of conceptual frameworks emphasize this continuum and the dangers of fragmentation. Care to prevent or treat the vast majority of maternal and perinatal conditions can be provided at home, at the primary level (clinic or health center), and at the secondary level (district hospital),⁴ with the district or equivalent regarded as the essential planning unit for service delivery (WHO 1994). This system is comparable to the "close-to-client" health system that the Commission on Macroeconomics and Health (CMH 2002) has proposed, whereby trained staff members other than doctors provide much of the care, with an emphasis on primary prevention and management of acute conditions.

Policy Considerations and Approaches

The health of mothers and babies is a human right and needs to be underpinned by policies and laws that increase access to information and good-quality, affordable health services (Germain 2000). A positive policy environment is crucial for promoting maternal health and reducing the burden of maternal and perinatal conditions. Such policy considerations need to go beyond the health sector to include related issues, such as transportation, nutrition, girls' access to education, and gender biases in the control of economic resources. Through a human rights–based approach, programs can be fashioned to ensure that every woman has the right to make informed decisions about her own health and has access to quality services before, during, and after childbirth (Freedman 2001).

The ICPD marked a dramatic shift not only by putting the concepts of rights and choice center stage, but also by introducing the reproductive health paradigm. The first decade of the ICPD plan of action was marked by major improvements in policies related to maternal health in most of the 179 signatory countries. However, as observed at the ICPD + 10 Conference, many promised changes remain at the level of policy pronouncement and have not yet been implemented. The stagnation is most notable in relation to maternal mortality and the HIV pandemic, especially in Sub-Saharan Africa. The failure to fully implement the ICPD consensus can be attributed to lack of political will, inadequate funding for programs to further reproductive health, and weak health systems. It is too early to judge the effect of the MDG proclamation (Johansson and Stewart 2002), although it could well suffer the same fate unless special attention is given to maternal and child health in the context of sectorwide approaches and Poverty Reduction Strategy Papers (UNFPA 2003). Some suspect that both these modalities may not give reproductive health the focus and attention it requires, because competing needs may crowd it out. Others argue, however, that sectorwide approaches can be a boon for maternal health because they offer a more effective platform for addressing ailing health systems (Goodburn and Campbell 2001).

Whether at the national or international level, advocacy for maternal and perinatal health should focus on the following seven key message areas:

• magnitude of the problem

• factors influencing maternal and perinatal outcomes

• functions of maternal health programs and which interventions work

• consequences of not addressing maternal and perinatal health

• costs of improving maternal and perinatal health

• responsibilities at each level of the health system and beyond

• policy and legal impediments to implementing comprehensive safe motherhood and newborn health programs.

Major advocacy networks, such as the Partnership for Safe Motherhood and Newborn Health, the White Ribbon Alliance, and the Healthy Newborn Partnership, seek to promote maternal and newborn health at the global level. Their purpose is to create awareness by changing the language of discourse, building international political commitment, developing global guidelines, and improving access to technical information for providers and program managers.

Selected Intervention Packages

For some of the reasons mentioned in the previous subsection, researchers have made few attempts to model packages of interventions for maternal and perinatal conditions, and many of those attempts do not specify content in sufficient detail to replicate the package. Our approach is to define content by beginning with a literature search of best practices in preventing and managing the focus maternal and perinatal conditions, acknowledging that, by excluding conditions that impose a lesser burden, we ignore interventions that might be highly effective and cost-effective. We then grouped those interventions that are considered effective and that are either being or likely to be implemented on a substantial scale into packages of care, bearing in mind previous CEA work, such as the WHO mother-baby package (WHO 1994). Expert panels then reviewed the component interventions and the packages and assisted with identifying resource requirements. Given the complementary CEA elsewhere in this volume on interventions relevant to maternal and perinatal conditions such as family planning, we focus on care during pregnancy, postpregnancy care, and care immediately postdelivery—in other words, on clusters or packages of interventions typically referred to as prenatal care, delivery or intrapartum care, and emergency obstetric care. Table 26.5 outlines the content of those packages.

When one considers the intervention packages, contextual factors are clearly crucial. Given the particularly high burden in South Asia and Sub-Saharan Africa, we chose those two regions as the specific health system scenarios for this chapter. Those regions are also characterized by high levels of poverty and encompass some of the most heavily indebted countries in the world.

Resource Use and Costs

We adopted an ingredients approach (Creese and Parker 1994) to identify resource use. For this type of bottom-up costing, we prepared lists for primary- and secondary-level care facilities of types of personnel, drugs, supplies (medical and nonmedical), medical and surgical equipment relevant for the interventions, and capital items (vehicles, buildings, building space). For most of the scenarios, our identification of resources was based on the WHO mother-baby package costing tool (WHO 1999), with necessary modifications because of the content of care packages indicated in table 26.5. We estimated the costs for clinical personnel on the basis of salaries for different grades according to the guidelines provided by the volume editors for the two selected regions. The time required by different staff members for each care intervention and the changes in time and personnel because of varying content and coverage of packages were informed by expert panel reviews, and we then calculated the costs. We valued the other nontraded inputs using information primarily provided by WHO-CHOICE (2004).

Cost-Effectiveness Ratios

The CEA involves a number of fixed and variable assumptions (see annex 26.A). The most important assumptions concern the reducible burden of these conditions, the effectiveness of the interventions, and the availability of appropriate human resources. We have assumed that increases in care can be achieved without major capital investments and that human resources are not in short supply; therefore, more could be used (with given wage rates) as required for increased activity and enhanced coverage.

Table 26.7 summarizes the findings of the CEA in terms of incremental cost-effectiveness ratios (ICERs) for the six primary comparisons between the base scenario and alternative intervention packages for a population of 1 million. Table 26.8 gives details of total costs, deaths averted, life years saved, and DALYs averted. Table 26.9 shows the findings of the sensitivity analysis in terms of how the ICERs change when different assumptions (see annex 26.A) are made with respect to effectiveness, met need, and inpatient costs.

In interpreting the results, note that they are point estimates. Even though they are based on the best information currently available, all the inputs into the model are subject to some degree of uncertainty. Without access to robust data on individual costs and effects or without specifying distributions for each variable, it is impossible to identify confidence limits for the estimated ICERs. Thus, we do not know, for example, whether the difference in the incremental cost per DALY averted for Sub-Saharan Africa between increased coverage at the primary level (US$92) and improved quality of CEmOC (US$151) reflects a genuine difference in cost-effectiveness or whether there are overlapping confidence intervals (table 26.7).

With those important caveats in mind, at first sight the results for South Asia and Sub-Saharan Africa appear quite different. For each intervention package, regardless of the specific assumptions made, the cost per DALY averted is always lower in Sub-Saharan Africa. The higher costs of care in Sub-Saharan Africa (see annex 26.A) are thus more than compensated for by the higher effectiveness, which is a result of the region's greater burden. However, some important similarities are apparent between South Asia and Sub-Saharan Africa. Leaving aside options 3b and 5b (the options without nutritional supplements), the results for both regions show a consistent pattern. Improvements in the overall quality of care, especially at the primary level through the provision of BEmOC (option 3a), together with increased overall coverage (option 5a), are the most cost-effective intervention packages—and both include nutritional supplements. They are followed by increased coverage at the primary level (option 2). Improved quality of CEmOC (option 4) is the least cost-effective option. Removing nutritional supplements from the packages makes relatively little difference in Sub-Saharan Africa, slightly increasing cost-effectiveness, but in South Asia, options 3b and 5b become less cost-effective with the nutritional supplements removed. The explanation lies in the ICERs of nutritional supplements as such, which are US$48 or US$45 in South Asia and US$118 or US$110 in Sub-Saharan Africa, depending on whether the comparison is with or without increased coverage (options 5a and 3a, respectively). This difference reflects the high burden from low birthweight in South Asia and, thus, the gain from nutritional supplements.

Comparing the content of the three most cost-effective intervention packages (3a, 5a, and 2) suggests that much can be achieved through improvements at the primary care level. Improved quality in relation to managing complications—in other words, the provision of BEmOC—and increases in coverage (a combination of options 3a and 2) at the primary level are likely to have even lower ICERs than those shown in table 26.7. This finding is consistent with the Commission on Macroeconomics and Health's emphasis on close-to-client services (CMH 2002), and it is highlighted further in chapter 53. As noted earlier, given the importance of prompt intervention in the event of obstetric complications, the effectiveness of intervention packages that may reduce delays by bringing services closer to communities is hardly surprising.

The benefits from option 2 were achieved essentially by increasing prenatal care coverage from 50 to 70 percent, because our model assumes that those women taking advantage of professional delivery are those who have also had prenatal contact. Prenatal care is, thus, a crucial entry point to the health system. Small changes in prenatal care coverage (20 percent) lead to larger numbers of women also benefiting from the rest of the care package in terms of obstetric first aid and CEmOC.

This issue is important for safe motherhood and newborn health, because the role of prenatal care has been subject to intense debate about its benefits relative to resource use (De Brouwere and Van Lerberghe 2001; Maine and Rosenfield 1999). Much of this discussion has focused on the lack of evidence on the direct contribution of prenatal care to reducing maternal mortality (McDonagh 1996; Rooney 1992), which, in turn, is explained partly by the poor performance of at-risk screening tools. However, differentiating the contribution to the prevention of maternal deaths of the prenatal care component alone is difficult. Ultimately, life-saving interventions depend on the functioning of the entire health system, including an effective referral network.

Our model also made assumptions about women's willingness and capacity to respond to referral to higher levels of care in case of complications. This willingness and capacity depend on many factors and are undoubtedly also driven by communities' perceptions of quality of care. As noted earlier, coverage rates of prenatal care are already high in many Sub-Saharan African countries, but significant socioeconomic differentials are apparent within countries. Our model does not address this equity dimension but, given the recent work showing higher risks of maternal death among the poorest groups (Graham and others 2004), targeting disadvantaged women for improvements in uptake might be worth considering (Gwatkin and Deveshwar-Bahl 2002; De Brouwere and Van Lerberghe 2001).

Whereas option 2, increased primary-level coverage, relates predominantly to the demand side of the health system (Williams 1987), the most cost-effective packages (3a and 5a) focus on the supply side, particularly at the health center level. The latter packages are particularly relevant to the baby, including screening of the HIV status of the mother and treatment with antiretrovirals at the time of delivery to reduce the risk of mother-to-child transmission, as well as provision of anti-malarials. As a consequence, these options have a particularly marked effect on the burden from perinatal conditions, accounting for two-thirds to three-fourths of the total DALYs averted (table 26.8). Note that these cost-effective options include a doctor at the health center level to provide all six BEmOC functions. In some situations, highly skilled midwives will be able to act in this capacity, which would reduce costs and further increase cost-effectiveness.

The most comprehensive packages in our model provide for improved quality of care and coverage at both the primary and the secondary levels (options 5a and 5b). Costing US$1.79 and US$1.63 per capita, respectively, in Sub-Saharan Africa (as calculated from the total costs of these packages shown in table 26.8, and divided by the base of 1 million people), these are also the most expensive packages. Not surprisingly, therefore, these two options avert much higher numbers of DALYs, with the package that includes nutritional supplementation averting nearly three times as many DALYs as the base package (table 26.8). In CEA, generally the most comprehensive packages—that is, those that result in the greatest gain in quality and coverage and, thus, cost the most—are often not cost-effective, and yet our analysis found otherwise. This finding may partly be explained by the linear assumptions about effectiveness in the model and the assumption that the marginal cost of care is constant. Such a finding also stresses both the importance of a well-functioning health system (rather than an excessive focus on one element) and the absence of any quick fix. Moreover, we did not model these more comprehensive options as perfect but unrealistic scenarios. We also still allowed for 30 percent of pregnant women not attending prenatal care, 50 percent of severe complications at a primary level not reaching CEmOC, and 10 percent of those reaching secondary care not receiving the emergency treatment they need.

Finally, a note of caution is warranted on the interpretation of the CEA results. First, our model has necessarily used a number of assumptions for which data are extremely limited, and it remains fairly crude, having been subject to only a limited sensitivity analysis. Second, many comparisons are possible from our model, but we have selected only six. Thus, we may not have identified even more cost-effective intervention packages, such as a combination of options 3a and 2.