PHN Technical Note 85-15 EFFECTS OF FERTILITY CHANGE ON MAT, AND CHILD SURVIVAL: PROSPECTS FOR SUB-SAHARAN AFRICA by Deborah Maine, Regina McNamara, Joe Wray Abdul-Aziz Farah, and Maril-yn Wallace October 1985 Population, Health and Nutrition Department World Bank The World Bank does not accept responsibility for the views expressed herein which are those of the author(s) and should not be attributed to the World Bank or to its affiliated organizations. The findings, interpretations, and conclusions are the results of research supported by the Bank; they do not necessarily represent official policy of the Bank. The designations employed, the presentation of material, and any maps used in this document are solely for the convenience of the reader and do not imply the expression of any opinion whatsover on the part of the World Bank or its affiliates concerning the legal status of any country, territory, city area, or of its authorities, or concerning the delimitations of its boundaries, or national affiliation. EN Technical Nbte 85-15 EFE S CF FIIfla cE aI mkmeuL AND flD SJRVAL: ARSBT KR SUBT A R M A ACT ABSTRACT In this paper, the potential effects of fertility chage on maternal and child mortality in sub-Saaharn Africa are estimated. Three. aspects of both maternal ard child mnrtality are examined: the curnt level of mortality; relative risks associated with family formation patterns; and exposure to risk factors. Information on maternal mortality in sub-Saharan Africa is wDefully sparse. Nevertbeless, there is eough information to make the following statements. Materal mortality in sub-Saharan Africa is stiJlJ very ligh-probably labove 200 deaths per 100,000 live births in most areas. Using World Fertility Survey (WFS) data on child- bearing patterns ardl desires, and existing information on relative risk associated with maternal age, se estimate that if mmen wmb say they wnt no mre children (aid are not using effective contraceptives) had no wore children, fran 5 to 18 percent of maternal deaths in sub-Saharan Africa would be prevented (depending on the coutry). While mortality rates among children younger tl"an five have declined substantially during the last several decades, they are still the highest in the world. Published studies of WFS data, and our own analyses of WFS data tapes from Cameroon, Gbana and Kenya, show the following: * short birth intervals are associated with substantial excesses of mortality amnog children bom either at the start or the end of such an interval. The excess in risk is strongest and most coilsistent durirg the first twoyears of life. * Te- excess in deaths associated with short intervals is not explained by socio- econormc status (as idicated by maternal education), maternal age, or birth order. In fact, birth interval is always a stronger predictor of death under age two than these other factors. Furthermore, control for these factors and for survival of the older child of the pair hardly reduced the relative risks of sbnrt interval. * A number of possible explanations of the birth spacing effect were explored, including concentration of short-interval births among disadvantaged groups, conpetition for family resources among closely-spaced children, aid maternal depletion. Nne of these theories appears to explain the poerful and persistent effect of birtl spacing on child survival. If all children wre preceded by a birth interval of at least 24 months, infant mortality rates in sub-Saharan Africa would be reduced by an estimated 5-22 percent. Ionger intervals would produce even greater reductions. Prepared by: Deborah Maine, Regina McNara, Joe Wray, Abdul-Aziz Farah and Marilyn Wallace, COnsultants to the Wkrld Bank Center for Population and Family Health, Columbia University October 1985 TABLE OF CONTENTS ABSTRACT ii ACKNOWLEDGEMENTS I NTRODUCTI ON 1 MATERNAL HEALTH 4 The Overall Risk of Maternal Death 5 Relative Risks of Maternal Death 10 Maternal Age 10 Parity 12 Birth Interval 13 Exposure to Risk of Maternal Death 13 Maternal Age 14 Parity 14 The Potential for Averting Maternal Deaths 15 INFANT AND CHILD HEALTH The Overall Risk of Infant and Child Death 18 Relative Risks of Infant and Child Death 19 Birth Order and Maternal Age 19 Length of Birth Interval 21 Preceding Birth Interval 22 Subsequent Birth Interval 24 Possible Confounding Factors 27 and Explanations Exposure to Risk of Infant and Child Death 31 Birth Order and Maternal Age 31 Factors Affecting Birth Spacing 31 Birth Interval 33 The Potential for Averting Infant and Child 35 Deaths DISCUSSION 36 TABLES F IGURES REFERENCES APPENDIX i; ACKNOWLEDGEMENTS We are grateful to a number of people for their generous assistance: Shea Rutstein of Westinghouse Health Systems for help with methodology; Erika Royston and Jane Ferguson of the Division of Family Hygiene, WVHO, Geneva, and Judith Fortney and Barbara Janowitz for sharing their information on maternal mortality; and Nicholas Prescott of the World Bank for comments on statistics. ii.i INTRODUCTION The effects of family formation patterns on the health of women and children have been explored in scores of studies carried out over the last fifty years. In spite of differences in time, place and methodology, strong patterns are apparent. For womenr several factors are associated with increased morbidity and mortality during pregnancy: bearing children when younger than 18 or older than 35; having four or more children. For children, unfavorable maternal age, high birth order, and short birth intervals have been associated with increased morbidity and mortalityr and decreased physical growth and intelligence test scores [Wray, 1971; Nortman, 1974; Maine, 1981]. Since only a few of the early studies of family formation patterns and maternal/child health were done in developing countriesr it was recognized that better data from such countries were needed. In 1970 the World Health Organizatic.n (WHO) initiated an international collaborative study of family formation patterns and infant health. The results obtained in Colombia, Egypt, India, Iran, Lebanon, Pakistan, Philippines, Syria and Turkey, confirmed that unfavorable family formation patterns have negative effects in developing as well as affluent countries [Omran and Standleyr 1976, 1981]. In the last few years, the World Fertility Survey (WFS) has proviaed researchers with an unprecedented wealth of information on infant mortality. The advantages of WFS data are both in scope and detail. In terms of scope, it has provided comparable data from larger systematic samples in more than 40 developing 1 countries, including nine in SubSaharan Africa (SSA). In terms of detail, it permits us to study, in ways never before possible, the prevalence and relationship to child survival of such factors as short birth intervals. The health situation in Africa today is complex. On one hand, using almost every available indicator of health status -° crude death rates, infant mortality rates, proportion of deaths under 5, life expectancy at birth -- health status in Africa is the worst in the world. Indicators of socioeconomic and environmental factors that affect health are consistent with this. On the other hand, for at least the last several decades, mortality rates (both crude and infant) have declined substantially. Gwatkin, in a review of demographic change around the globe# observed that although life expectancy in Africa has increased rapidly since World War II, the overall level in SubSaharan Africa is still below 50 years [Gwatkin, 1981]. Furthermore, the rate of increase in life expectancy is beginning to slow at an appreciably lower level than in other regions. Gwatkin speculates that the initial declines in mortality were due to relatively simple Western medical technology that became available in the 1950s, such as antibiotics, immunizations and malaria control. Further decreases in mortality, especially among infants and young children, may require significant improvement in the standard of living. This is happening extremely slowlyr if at all, in much of SubSaharan Africa. Mortality and fertility are related in dynamic and complex 2 ways. In Africa, the decline in mortality has not been offset by decreased fertility. Birth rates have declined minimally, if at all [World Bank, 19841. The result, of course, is rapid population growth -- three percent per year for the continent as a whole [Kent, 1983]. In this paper, we will concentrate on one particular aspect of the relationship between mortality and fertility -- the direct or short-term effect of fertility on family health. "Direct" and "short-term'" are relative terms. An example of a short-term effect is the increased likelihood of maternal death with increasing birth order. An example of an indirect or long-term effect might be that high birth order spurs population growth, thereby increasing the load on the medical system and perhaps lowering the guality of obstetric care, resulting in more maternal deaths. Before we proceed, several qualifications are necessary. First, we will be dealing almost exclusively with the health of women and young children because theser so far as we know, are the family members whose health is directly affected by family formation patterns. Secondly, we will often be discussing survival, rather than health p se. This is because even in developed countries morbidity data are sparse. In Africa, more than any place else in the world, we must rely on mortality data since so many people have little or no contact with the modern medical system, and special health surveys are rare. 3 MATERNAL HEALTH Life expectancy can be considered the "bottom lineit in discussions of health. Table . shows that life expectancy in Africa as a whole is estimated to be 50 years -- 10-14 years less than in other iarts of the Third World, 24 years less than in North America [Kent, 1983]. In most of SubSaharan Africa, life expectancy is less than 50 years. Pregnancy and childbirth are major determinants of women's health. In developing countries, complications of pregnancy and delivery make up about one-quarter of all deaths among women of reproductive age [Chen, et al, 1974; Willia.risr 1979; Fortney, 1982]. By comparisony in the United States, maternal deaths comprise less than one percent of total deaths among women 15-49 [National Center for Health Statistics, 1984]. The second tier of Table 1 gi es WHO ranges for maternal mortality rates (MMP.). Again. the disadvaritagc faced by African women is clear. The min.imuM estimate for Africa is 160 deaths per 100,000 live births -- 10 times higher than the maximum in North America [WHO, 1983]. Only in South Asia do rates approach those in Africa. The bottom tier of Table 1 shows the proportion of deliveries attended by trained medical personnel [WHO, 1983]. This is far lower in Africa (28 percent) than in Asia or Latin America (51 and 62 percent, respectively). These data tell us something about the risks faced by women giving birth in Africa. From an epidemiologic viewpointr the effect of fertility on the mortality.among women can be measured in several ways. One is the oyerallLg k of maternal mortality, 4 iLe., the average risk of death associated with becoming pregnant in a given country. A second measure is the relat .jy associated with pregnancy among various groups of women, e.a.? maternal mortality among women of different ages. The public health importance of relative risk depends, in part, on the proportion of women in the population who fall into high risk groups -- i.e.r the levlj u to risk. Using data on risk and exposure, we can estimate the impact of family formation patterns on maternal mortality. The Overall Risk of Maternal Death While the WFS has provided a tremendous amount of information on child survival in Africa, very little has been learned about maternal mortality in the past decade. What information we do have comes from three sources: hospital studies, official records, and community surveys. By far the most common form of information on maternal mortality in Africa is the hospital series -- i.e., a report on the number of deaths in a hospital compared to the number of deliveries there during a specified period of time. The major problem with such reports is that we cannot identify the population from which the patients are drawn, since most women in Africa deliver at home. In other words, we do not have appropriate denominators with which to derive rates. Doubts about the representativeness of hospital series data are substantiated by the common complaint of cli-nicians in Africa -- that many of the women who die under their care arrive in the hospital already moribund [Bullough, 1981; Aggarwal, 19801. 5 Despite these drawbacks, we need not disregard all hospital data. Some researchers provide separate MMRs for women who received prenatal care at the hospital ("booked" cases). The advantage of information on booked cases is that women who come for prenatal care can be considered to be a population within which some deaths occur. This reduces the bias that results from a heavy concentration of complicated deliveries. Of the dozens of African hospital studies examined, only nine provide separate mortality rates for booked patients.* In these, the MMR among booked women ranged from 59 to 530 death.s per l00,000 deliveries, as Table 2 shows. The median MMR is 150. We do not know the reasons for the variation, but it is likely that patterns of "selection" into medical care are as important as true difference in the MMRs. For example, in countries where most women have no prenatal care, it is likely that booked women are more advantaged than average. While these data may still be flawed, they are more meaningful than are statistics that include unbooked patients as well. The MMR. among unbooked patients is oft 0 or mGrL- e mes- that among booked cases in the same facility [Akingba, 1977]. Another source of data is studies of official records. By this we do not mean routine government reports of maternal mortality (which are notorious for underreporting), but special studies of official records. Typically, in such studies the numerator is derived from surveys of death records in as many * Throughout this paperr all relevant studies that we could find are included in text or tables. 6 sources as possible in a given area, and the denominator is the estimated number of births in the area during the same period of time. Four such studies were found, yielding MMRs of 103 to 172r with the median being 1304 A basic assumption in these special studies is that a large majority of maternal deaths (unlike births) either take place in health facilities .r eventually come to the attention of the medical personnel. This is a questionable assumption. For example. one of the lower MMRs reported in Table 2 (103 deaths per 100,000 births) is from a study of official records in Malawi in 1977 [Bullough, 1981]. In this study, questionnaires were sent to hospitals and large maternity units, as well as any of the 92 smaller maternity centers that had noted a maternal death in its monthly reports. The reported deaths were divided by the expected number of births in the region during the study period. In this population, more than 60 percent of all births take place at home. However, all but six of the 118 reported deaths took place in medical facilities. Since a large proportion of maternal deaths in developing countries are caused by hemorrhage and eclampsia (which often lead to rapid death) it seems most unlikely thssat so few women died before they reached medical-care. Thus, underestimation of the MMR seems certain, but its magnitude cannot be known. Suf f ice it to note that a study by the Centers for Disease Control found that during 1974-1978, maternal deaths in the United States were underreported by 20-30 percent [Smith, et al, 19841. Assuming no greater underreporting of deaths in Malawi, the true death rate would be 124-134. By far the best evidence on maternal deaths comes from 7 community studies, but these are rare. The findings of recent community studies in SubSaharan Africa are shown in Table 2. The MMR ranges from 400 in Ghana to 700 in Senegal. (Another stu dy, in rural South Africa, produced a MMR of 550, but this was based on a sample of only two deaths [Larsen, 1983].) These figures fall around the middle of the range of MMRs in Africa provided by the WHO: 160-1,100 (see Table 1). For comparison, a community study in Egypt found a MMR of 190 tFortney and Rodgers, 19843. Considering that the average life expectancy in Egypt is 56 yearst compared to less than 50 years in SSA, this difference in MMRs does not seem unreasonable [Kent, 1983]. What can we deduce from this relatively meager collection of data? Most of the studies shown in Table 2 present conservative estimates, avoiding the wildly inflated rates found in hopital studies that include unbooked patients. Among the studies of booked patients and official records, the median MMR is 138. The recent community studies (the preferred source of data) produce a range of 400-700 deaths per 100,000. It is credible that maternal deaths in SubSaharan Africa are this common? Quite possibly, considering that in l920, when the life expectancy for women in the United States was 55, the MMR was a staggering 800 for the total population, and even higher among minorities [New York, 19333. When discussing the risks of childbearing to women's health, it is necessary to consider the risks of avoiding pregnancy. Over the two decades since oral contraceptives reached the market, 8 probably more information has been gathered on their effects than on those of any other substance in human history. Hundreds of studies have been done on IUDs and other methods as well. To the best of our knowledge, use of any one of these methods by women in developed countries is far safer than pregnancy and childbirth, with the exception of use of the pill by older women who smoke (Oryr et al, 1980; Oryr 1984]. Not surprisingly, data from developing countries on the side effects of modern methods of contraception are virtually nonexistent, and the issues are complex. On one hand, it might be that common health problems in poor countries interact with contraceptives to produce side effects that are either not seen or not common in industrialized countries. Howevery this possibility has not been substantiated. On the other hand, it may be that use of effective contraception is even more beneficial to women, in poor countries whose health is more fragile to begin with. There is some support for this view. For example, preliminary results of the Menoufia study found that "in spite of less than optimal medical supervision, the practice of contraception in a developing country is safer than the alternative," i.e., pregnancy EFortneyr 1982]. Finallyr the above remarks refer to the use of modern methods of contraception. Women who resort to illicit abortion in order to avoid having a child are at very high risk of serious complications and death in most developing countries. Tens of thousands of Third World women die from illegal abortions every year [Rochat, 1980; Liskin, 19801. While deaths from illicit abortion are less common in SSA than in other parts of the world, they are certainly not unknown COjo and Savage, 1974]. Furthermore, clinicians in SSA .report an increase in young women suffering the effects of illegal abortions (Akingba, 1977]. Only if birth planning is accomplished by safe methods do the health benefits to women discussed below apply. Relative Risks of Maternal Mortality Ma ter Aa The overall risk of maternal death is largely determined by factors that affect the standard of livingr such as era (e.g.r before or after the invention of antibiotics), country, socioeconomic status and access to medical ca.re. Unfortunately, there is usually little an individual or family can do to change these factors. There are a number of factors, however, that also affect the risk of maternal mortality and arer to some extent, under women's control: the age at which they give birth, and the number and spacing of births. It is these factors we will consider under relative risks. Numerous studies spanning continents and decades have shown that adolescents and women 35 or older are much more likely than are women in their 20s to die as a result of pregnancy [Nortman, 1.974; Maine, 19813. This relationship is not dependent on socioeconomic status . In fact, in her landmark review and analysis of the literature, Nortman observed that the effect of age on maternal mortality is stronger in countries that have low rates of maternal deaths (e.g., European countries) than in those with high rates [Nortmanr 1974]. Regardless of the level of maternal mortality, the effect of 10 age is impressive. Figure 1 shows the relative risk of maternal death by age in Nortman's "high mortality" group of countries. In this group of countries (as in the low- and medium-mortality groups) women in their early 20s had the lowest rates of maternal death. In Figure 1, women younger than 20 are half again as likely to die of complications of pregnancy or childbirth as are women aged 20-24. At age 25 the rate starts to rise again, until women 40-44 have a MMR nearly five times that of women 20-24. The shortcomings of African data on maternal mortality are especially pronounced when one tries to specify rates for subgroups, such as age groups. Consequentlyf most analyses of this type use Asian and Latin American data. For example, none of the high mortality countries in Nortman's study were African. To what extent, then, does age affect maternal mortality in Africa? Ficure 2 shows maternal mortality by age in Zaria Hospital in Nigeria during 1976-1979 [Harrison and Rossiter, 1980B. These data are drawn from a series of 22,000 deliveries among both booked and unbooked patients -- as is usually the case, there were not enough booked patients to derive FiMRs Lor different age groups. The excess of deaths among women in their early teens is clear. Fifteen-year-olds had almost seven times the risk of death of women aged 20-24. Women on either side of the "ideal" age group -- those aged 17-19 or 25-29 -- had MMRs elevated by 30 percent. Among 16-year-olds and women 30 and older, the risk of dying was 2.5 times that of women aged 20-24. (The available data did not permit creation of more conventional age groups.) These data indicate that maternal mortality in Africa is affected by age in much the same way as in other areas. 11 For decades, studies have consistently shown that maternal mortality varies with parity as well as with maternal age. In general, maternal mortality is slightly eleevated among women having their first birth, lowest among women having their second or third child, and increases sharply thereafter [Nortman, 1974; Maine, 1981]. As is the case with age and mortality, this relationship has been documented in a wide variety of settings, and has held up as maternal mortality rates declined dramatically during this century. The relationship between parity and maternal mortality in Africa is even less well documented than that between mortality and age. Even so, the few fragments of inLormation we do have indicate that Africa is no different than the rest of the world in this respect. Figure 3 shows the relative risk of maternal death by parity in a hospital in Senegal during 1971-1975 [Correa, 1978). As usual, the curve is J-shaped, although there are several notable features. First, the relative risk of death among women having their first birth is higher than in some other studies. This might be due to the young age at f irst birth in Senegal. Secondlyf the relative risk is slightly higher among women having their second birth than among those having their third. This might be because approximately one-fifth of women younger than 20 were having their second or later birth. This brings up the issue of interaction between age and parity as they affect maternal mortality. Of necessityr as women have more children they are also getting older. Consequently, 12 when one looks at either age or parity, some of the-observed effect is probably due to the other. There have, however, been large studies in industrialized countries where the effects of age were controlled while those of parity were examined [Nortmant 1974]. These showed that both age and parity influence maternal mortality independently. Birth ;L£S:s Over the years, as evidence of an effect of birth spacing on infant mortality has accumulated, people have speculated about a similar effect on maternal mortality. Having children very close together must be hard on women, especially poor women, the reasoning goes. It was hypothesized that this strain on the mother's body (named the "maternal depletion syndrome" by Jelliffe), might be one of the wavs in which birth spacing affects infant health [Jelliffe, 1966]. As sensible as this seems, no direct evidence of the maternal depletion syndrome has been foundr but there is indirect evidence. Nortman observed that the optimal age for maternal survival increases with parity [Nortman, 19741. Since this cannot be either an age or a parity effect, it seems to be a spacing effect. Thus, while we can infer that closely spaced births are harmful to women's health, the effect is not nearly as strong as on infant and child health. Exposure to Risk of Maternal Mortality .The importance of the relative risks discussed above depends, to a large extent, on the proportion of pregnancies that carry high risk. For example, the high risk to maternal health associated with giving birth after age 40 is of littie consequence in the United States because it has become an 13 uncommon event. In SubSaharan Africar on the other hand, women generally marry early, have their first birth soon afterwards, and continue to have children until they are in their late 30s or early 40s. (There is, of course, considerable variation within and among countries.) Table 3 shows the proportion of all births that are to women in high-risk age groups in nine SubSaharan and two North African countries, The proportion of total births that are to women 15-19 ranged from six percent in Tunisia to 23 percent in Cameroon. There was less variation in births to older women: 8-15 percent among those aged 35-39; 3-13 percent among women 40-49. The last column of Table 3 shows the proportion of all births that take place in unfavorable age groups. This is lowest in Benin (25 percent) and highest in Kenya (38 percent). In general, about one-third of all births occured at ages when the woman is at increased risk of serious complications and death. Unless pregnancies are very widely spaced, continuing childbearing into their late 30s and early 40s is a sign that women are attaining very high parity. This is the case in Africa where fertility is higher than in other parts of the world. The the Total Fertility Rates (TFRs) * are 4.1 and 4.3 in Asia and The Total Fertility Rate (TFR) is a synthetic rate, representing childbearina in an imaginary cohort. It is the number of children an dverage woman would have during her lifetime if she conformed to the age-specific fEertility rates . current at the time of the survey. 14 Latin America, respectively, compared to 6.5 in Africa [Kent, 1983). Kenya has the highest TFR in the world -. 8.0. Actually, the TFR (being a kind of 'average') doesn't express what large families many African women have. As Table 4 shows, among women who have? for the most partr completed their childbearing (i.e., aged 45-49), at least one-third of women in the African countries for which we have data have borne seven or more children. In most of these countries, the proportion is more than one-half. In some societies, family size in one's grandparents' generation seems large when compared to current standards. However, it is important to remember that the large families cited above are not relics of the past. We have no evidence that fertility rates are declining substantially in these countries. The Potential for Averting Maternal Deaths A number of researchers have made estimates of the proportions of maternal deaths that might be averted with a given change in childbearing patterns [Nortman, 1974; Trussel and Pebley, 19841. However, they either did not have or did not take advantage of the kinds of information that the WFS provides. They assumed that, for example, women 35 or older no longer gave birth. But they did not take into account whether or not such births were wanted. The WFS provides data on the number of women in each age group who say they want no more children. Therefore, we can use women's own fertility desires to estimate the proportion of births that might be averted in the future through the use of 15 contraception.* Although these data have shortcomings, using women's stated desires definitely seems preferable to the alternatives. In addition, using WFS data we can determine the proportions of women who want no more children that are already using effective methods of contraception. If we didn't have this information, we would be doublecounting -- i.e., counting among the deaths that could be averted some that already are being averted. The proportion of women who say they want no more children is much higher in North Africa (47-53 percent) than in SSA (4-17 percent), as Table 5 shows. Table 5 also shows the proportion of women who want no more children that were using an efficient method of contraception at the time of the WFS survey. This proportion is 4-17 percent in the seven SSA countries. compared to 46 percent in Egypt, and 48 percent in Tunisia. Table 6 show estimates of the proportion of deaths that could be prevented if all unwanted births were averted through use of effective methods of contraception if women in all age groups who say they want no more children (and are not currently using an efficient method of contraceptioon) realized their desires and had no more births. (The methodology used is * This is not to say that these data are necessarily predictive. Some women who say they want no more children, and are exposed to the risk of pregnancy. do not use effective methods of contraception even when there is no apparent physical or financial barrier to access. On the other hand, it is quite likelv that there are a good many women (perhaps especially in Africa) who will not say to-a stranger that they want no more children, even though that is the truth. 16 presented in Appendix 1.) In Egypt, where half of currently married, fecund women say they want no more childrenr the reduction would be largest (28 percent), followed closely by Tunisia (26 percent). In the six SSA countries for which we have the necessary data, an estimated 5-18 percent of maternal deaLhs would be averted. While lower than in Egypt, these are still substantial reductions, especially when one remembers how common maternal deaths are in these countries. Another way to conceptualize the reduction in maternal deaths possible with the use of family planning is to focus on the risk to an individual woman. While infant death is a one-time risk, women face the risk of maternal death repeatedly. Consequently, even if we ignore the increasing risk per pregnancy with age and birth order, a woman's chaances of dying are multiplied by the number of births she has. Thuse reducing the average completed family size Kenya from eight to six children would reduce the number of maternal deaths by at least one- quarter. 17 INFANT AND CHILD HEALTH The Overall Risk of Infant and Child Death. In 1982. infant mortality in SSA ranged from 77 deaths per 1,000 live births in Kenya, to 190 in Sierra Leone, with most countries having rates between 100 and 140 [World Bank, 1984]. (See Table 7.) (These figures are estimates based on census and survey data since vital registration in most African countries is far from complete.) Child mortality rates (deaths at 1 to 5 years) ranged from 13 deaths per 1,000 children per year in Kenya, to 39 in Angola, with most countries having rates of 20 to 30. These rates mean that in many SSA countries, one-fifth to one-quarter of children die before their fifth birthday. Although these rates are among the highest in the world, they are considerably lower than they were two decades ago. As Table 7 shows, in about half of the 32 SSA countries, infant mortality declined by 20-30 percent during 1960-1982. Child mortality declined by at least 40 percent in two-thirds of the countries shown. A-t -h--discussion of the determinants of child mortality is beyond the scope of this paper. As is the case with maternal mortalityp the national level of infant and child mortality reflects many of the factors that comprise standard of living. The focus of this paper is risks associated with certain aspects of family formation -- i.e.r birth interval, birth order, and -maternal ager but it is important to remember that socioeconomic status has a strong influence on mortality even when these factors are controlled t[ob1zraft, et al.. 1984; McNamara., 1985]. 18 Relative Risks of Infant and Child Death Bir td Ordena Age. Traditionally, studies examining the relationship of birth order to infant mortality have shown U-shaped or J-shaped curves. That is, mortality is lowest among second and third children (Nortmant 1974; Wray, 1971; Maine, 1981]. The relationship between infant/child mortality and family formation patterns in SubSaharan Africa has been explored with WFS data for nine countries [Rutsteinr 1984]. The familiar patte.rn emerges, as Figure 4 shows (Rutstein, 19841. In these countries, infant mocrtality is generally lowest among second and third order children, while seventh and higher order children generally have the highest mortality rates. In Mauritania and Kenya, infants of birth order 4-6 have the lowest rates of death. In the Sudan. infant mortality appears to decline steadily with birth order -- a very unusual pattern. After the first year of life, birth order has no consistent effect on mortality. Figure 5 shows the relationship of maternal age to infant mortality. The relationship is roughly what would be expected from the findings of previous studies -- i.e.r mortality is lowest among infants born to women in their 20s. Until a few years ago. birth order and maternal age were the main family formation factors studied for effects on infant wellbeing. A number of practical considerations contributed to this. Information on birth order and maternal age is easier to co-llect, often being available from birth certificates. Furthermore, without computers, simultaneous examination of a number of variables in large groups was exceedingly difficult. 19 Now that computer analysis is standard, and the WFS provides data from large and varied populations, that has changed. And, at the momentr the trend seems to be to discount the importance of birth order and age. The reason for this is that when, for example, the effects of maternal education, and infants' sex, birth order and interval are controllede the effect of maternal age on infant mortality is often reduced and, in some countries, disappears (Hobcraft, et al, 1984]. Similarly, when education, sex, interval and maternal age are controlled, the effect of birth order is diminished and inconsistent. In our analysis of the WFS data from Cameroont Ghana and Kenya, we found that adjustment for birth interval tends to slightly increase the effect of older maternal age and weaken that of high birth order. (See Appendix Table 6.) We believe that it is a mistake to discount maternal age and birth order on the strength of such analyses. The reason for this is that there is a difference between research conducted to discover etiology and research done to shape policy and programs. In this case, controlling for various other factors while examining the effect of, for example, birth order allows us to search for the root "cause." Having done that, we can see that there is no simple cause. What then? In realityr maternal age and birth order are not separable. For example, it is unusual for old women to be having their f irst child, especially in developing countries. On the oth'!r hand, to individual women or the program personnel who have contact with them, age and birth order are still extremely useful markers, indicating the need for 20 special counselling and care. The same argument, we believe can be made in terms of the recent attention to the idea that completed family size, rather than birth orderr is an important influence on infant mortality [Gray, 1984]. Midwives, physicians and health educators, when faced with an individual woman. have no way of knowing what her completed family size will be (as, indeed, she herself does not). They can only know that she is 35 years old and has already given birth to five children. If statistical analyses are to be helpful to policymakers, it is imperative that the models from which "findings" are derived are relevant to the programs where they will be applied. Len1gth o£ Bith Int5r2al Scientific interest in birth spacing is recent compared to interest in birth order and maternal age. Early studies that included birth interval as a variable were conducted by the U.S. Children's Bureau in the 1920s, and there were a few other studies over the following decades [Hughes,1923; Woodburyr 1925; Eastman, 1944; Morris and Heady, 19553. But it wasn't until the 1970s that a concerted effort was made to gather data on this factor. Only in the last ten years did researchers begin to discern that birth interval can be a more powerful influence on infant mortality than either birth order or maternal age [Swenson, 1976, 1977]. The World Fertility Survey, by providing quantities and kinds of data never before available, has lead to confirmation of this finding. 21 Prec2dina Blith Interval The WFS data confirm that a short birth interval adversely P-ffects both the child born at the start of the interval and the one born at its end. Figure 6 shows the relationship between infant mortality in nine countries and the length of the preceding birth interval. This relationship is expressed as the relative risk of death among infants born at the end of a short birth interval (less than 24 months). Infants born at the end oL a medium-length interval (24-47 months) serve as the baseline group. The countries are placed in order of their overall infant and child mortality levels. During the decade 5-14 years before the survey, in Senegal there were 271 deaths per 1,000 among children younger than five, compared to 129 in Ghana. An increase in infant mortality is associated with short birth interval in each of the nine countries. Ihe relative risk ranges from 1.30 in Senegal (i.e., an excess of 30 percent) to 2.10 in Ghana, as Figure 6 shows.1 Thus, among these African countries there is a rough inverse relationship between the excess in infant deaths and overall infant mortality. However, it is not clear what meaning, if any, this observation may have, since no such relationship was found in an analysis of WFS data from 40 countries [Rutsteinr 19841. Furthermore, and perhaps more 1 The WFS data tape for Nigeria is not yet generaly available, but Shea Rutstein kindly provided the results of some preliminary analyses (personal communication, June 1985). These show that in Nigeria, which contains one-fifth of the population of SubSaharan Africa, short birth intervals carry substantial excess risk of childhood mortality. Relative risk of infant mortality at inter- vals less than 24 months? compared with 24-47 months, was 1.6. When the interval was longer than 47 months, the relative risk fLell to .7. 22 important, while infant mortality has declined in all of these countries in recent decades, the relative risk associated with short intervals has not changed in any systematic way over the 30 years of birth histories covered by the WFS surveys. Figure 6 also shows that being born at the end of a long interval (48 or mnore months) confers substantially more protection against infant mortality than does being born at the end of a medium interval in all nine countries. Working with WFS data on births 0-15 years before the survey in Cameroon, Ghana and Kenya, we broke down birth intervals into finer categories. This analysis shows that infants with preceding birth intervals of three years (36-47 months) have substantially lower death rates than infants with preceding intervals of two years (24-35 months). (See Figure 7.) Taken together, Figures 6 and 7 indicate no uthreshold" in birth inte.vals above which risk of death stabilizes. On the contrary, the beneficial effect of lengthening birth interval is incremental, at least up to six years. Figure 8 shows deaths among "tocldlers" (children 12-23 months old) in the nine SSA countries. In Senegal, Mauritania and Ivory Coast, short preceding birth interval was not associated with increased toddler mortality. In the other six countries, toddlers born at the end of short intervals were 24-58 percent more likely to die than those born at the end of a medium interval. The meaning of the inconsistencies that appear among these nine SSA countries is not clear. Three community studies of child spacing in SSA have not found evidence of excess mortality 23 associated with short intervals (Cantrelle and Leridon, 1971; Doyle et. al.t 1978; Boerma and van Vienene 19841. The different study designs and absence of controls for potentially confounding factors make comparison impossible, but these anomalous findings strongly suggest the need for more community-level studies with common methods and definitions of intervals and other variables. The additional benefit of a long interval (48 or more months) is more pronounced among toddlers than among infantsr as Figure 8 shows. In eight of the nine African countries, intervals of at least four years were associated with the lowest mortality. in a number of countries, mortality among toddlers born after such long intervals was about E.alf that among those born after a medium interval; in NigeriaF it was about one-third (see footnote p. 22)e As for deaths at 2 to 5 years, there is no consistent excess of deaths among children born after short intervals (not shown). Subae Th Tel The widespread and strong traditions of child spacing in Africa generally have, as their stated purposer the survival and development of the child already born -- i.e., the first born of the pair [Williams, 1938]. Becoming pregnant again before that child is on its feet (literally and figuratively) is a reason for shame in many parts of SSA (Lesthaeghe, et al, 1981]. Premature weaning is obviouSly hazardous to children's health [Cantrelle and Leridon, 1971]. After the birth of the second child, the first may receive much less maternal attention, and is commonly given into the care of an older child. 24 In spite of the recognition of these hazards in many African cultures, there has been little information on the effects of the subsequent interval on child health. Only WFS data permit quantification of the effects of birth spacing on both children of a pair o. adjacent siblings in many countries. Using WFS data, Hobcraft et. al., analyzed the effect of subsequent birth interval on deaths from ages 1-5 in Lesotho, Kenya, Sudan and Ghana [Hobcraft, et al, 19831. The results are shown in Figure 9, where the baseline group is children for whom both the preceding and subsequent birth intervals were short (Group A). These children are compared to children who had a short preceding birth interval (B). or a short subsequent interval (C or D). This allows us to compare the relative importance of preceding and subsequent intervals. (Note that the WFS researchers made the cutoff point for "short" succeeding intervals at less than 18 months for toddlers, and less than 30 mjnths for children 2-5 years old. This may account for some of the apparent differences shown in Figure, 9.) The first three columns of Figure 9 show that, for toddler mortality, a short subsequent interval is more harmful than a short preceding interval. This seems to confirm folk wisdom about the importance of subsequent interval to the wellbeing of the weanling. The relative risks for short preceding interval are 1.23-1.46, while those for short subsequent interval are 1.49- 2.72. We examined the WFS data for Cameroon, Ghana and Kenya for the effects on toddler mortality of 18-30 month intervals compared to longer and found no elevation of risk. In a later analysisr Hobcraft et al. analysed the risks of toddler death associated with very short subsequent birth intervals - i.er less than 12 months LHobcraft, et al, 19843. Relative risks following such short intervals were extremely high, even after the effects of birth order and maternal age and education were controlled. As for mortality in the second to the fifth year, Figure 9 shows weaker and more varied results, with relative risks of 1.01 to 1.48 associated with short preceding intervals, and of 0.95- 1.46 associated with short subsequent intervals. Furthermore, when multivariate analysis was used to control the effects of birth orderr maternal age and education, the effects of subsequent interval on mortality at 2 to 5 years disappeared in most of these African countries [Hobcraft, et. al., 19841. To summarizer it appears that in SubSaharan Africa, as in other parts of the world, the longer the birth interval, the greater are the child's chances of survival. This is true at least for the first five years of life, although the effect diminishes with age. Comparison of the effects of preceding and subsequent intervals indicates that (at least among toddlers) subsequent interval is the more powerful. While there is considerable variation among countries in relative risks associated with birth spacing, this variation does not invalidate the general conclusions stated above. Further research may produce some insight into the reasons for these differences among countries. 26 Possib Cofled a atz Explanat_ons Ever sinCe studies of the relationship between birth interval and child mortality have been publishede questions about its validity have been raised, and theories about mechanisms of action have been proposed. The wealth of new data and recent studies allows us to address some of these issues. In discussing the risks associated with short birth interval, some investigators have raised the possibility that in many cases the birth interval may be short e _ the first child of the pair died. Death of a young child may hasten the next conception by interrupting lactation amenorrhea, postpartum abstinence or use of contraceptives. It is likely that children born in such circumstances are at higher risk because of conditions particular to the family, either biological or environmental. We tested this hypothesis using the WFS data for Cameroon, Ghana and Kenya. As expected, mortality rates are much higher for siblings of non-survivors, and in the multivariate analysis the death of an immediately adjacent sibling is a powerful predictor of mortality. However, the relative risk of mortality associated with birth spacing is quite similar whether the older sibling survives or dies. The following table shows the relative risk of death associated with a preceding interval less than 24 months, with and without control for survival of the child born at the beginning of the interval (at least until age two). Differences in relative risk are apparent only during the first year of life, and they are not large. When we adjust for the survival of the older child 27 of the 1pair, the relative risk of death in the first month of life is reduced from 3.5 to 3.0 in Cameroon. In the other countries, differences due to this adjustment are ',maller. Relative risk of death associated with preceding birth interval less than 2 years compared with 3-4 years: Unadjusted and adjusted for survival of the older sibling of the pair. Age at Death Cameroon Ghana Kenya 1st Month Unadjusted 3.5 2.6 2.3 Adjusted 3.0 2.2 2.1 1-11 Months Unadjusted 2.3 3.1 2.1 Adjusted 2.0 2.7 1.9 12-23 Months Unadjusted 1.8 2.2 2.3 Adjusted 1.8 2.2 2.3 24-59 Months Unadjusted 1.3 le0 1.4 Adjusted 1.3 10 1.4 Other factors that might confound the relationship between birth interval and mortality are maternal ager birth order and socioeconomic status. The reason is that short-interval births might be concentrated among high-risk women -- i.e., poor women, women at either extreme of their reproductive years, and high parity women. In this case, the apparent excess of deaths among children with short birth intervals might be due to these other factors. As will be shown below in the section on "Exposure to Risk," concentration of short-interval births among disadvantaged women 28 does not explain the association of birth interval and child mortality. Secondlyr by using logistic regression we can discern the relative contribution of these various factors to mortality. The effects of birth interval on mortality are clearly independent of maternal education (an indicator of socioeconomic status), maternal age and birth order. When these three factors are introduc-ed into a logistic regression, relative risks of death are virtually unaffected, as the following table. shows: Relative risk of death associated with preceding birth intervals less than 2 years compared to 3-4 years: Unadjusted and adjusted for maternal education, maternal age and birth order. Age at Death Cameroon Ghana Kenya 0-11 Months Unadjusted 2.8 3.0 2.2 Adjusted 2.8 2.9 2.2 12-23 Months Unadjusted 1.8 2.2 2.3 Adjusted 1.8 2.2 2.3 24-48 Months Unadjusted 1.4 1.1 1.5 Adjusted 1.3 1.0 1.4 One popular theory about the reason for the birth spacing effect (both in folk wisdom and among researchers) is that when children in a family are born too close together, they compete for the family's resources, such as food and parental attention. Despite the common-sense appeal of this theoryr the WFS data indicate that competition within the family is not a major factor in the birth spacing effect. First of all, it would seem that families of higher socioeconomic status would be able to provide 29 enough food and care for several young children. But, as we have seen, controlling for maternal education does not materially change the relative risks of death associated with short birth interval. Secondly, if the older child of a closely-spaced pair di.es in infancy, and the competition is thus ended, the risk of death for the younger child associated with short interval should be reduced. Again, this is not the case. Another theory about the hazards of short birth intervals is that they are a characteristic of "high-risk" families, and that this intra-familial risk (rather than short interval) is the real reason for the apparent effect of birth spacing on child survival. But the death of an older sibling seems a reasonable marker for high-risk families, and yet the effects of birth spacing are not appreciably greater in such families than in those where the older sibling survived. Finally, there is the "maternal depletion" theory, in which it is the physical strain on the mother exerted by closely-spaced pregnancies that adversely affects the children. While this might explaiTffsome of the effect on newborns, it is a less likely explanation for risk to older children. In truth, each of the theories discussed above may play a part in the effect of birth spacing. However, even when taken togethert they do not appear to account for the entire effect -- from birth through the first five years of life, on both ch,,ildren in the pair. Furthermore, the persistence of tlie birth spacing effect after adjustment for maternal education, maternal age and birth order, and across countries and decades, indicates that we 30 have much to learn about its causes. ?xposure to Risk of Infant and Child Death As was noted in the section on maternal mortality, very large families (e.g., seven or more children) are common in SSA. Even in SSAr many women do not want such large families, Births at unfavorably young and old ages are also common. Among the women aged 20-34 at time of interview, 39 percent had their first birth before they were 18 in Kenya, and 29 percent in Ghana. As Table 3 has shown, childbearing among women in their 40s is still relatively frequent in SSA. E= .9 AfffeZins Blh Sjpacing Anthropological studies show that people in traditional societies all over the world have long perceived the advantages of lng birth intervals. This was certainly the case in scores of African societies, where prolonged postpartum sexual abstinence and breastfeeding were the rule. In a recent review of anthropologic data from SubSaharan Africa, 167 societies were reported to have traditions of postpartum abstinence. (Schoenmaeckers, et al., 1981) The duration of abstinence varied, being less than 40 days in 17 percent of cultures. 40 days to one year in 28 percent, and longer than one year (often 2-3 years) in 55 percent. Generallyr the justifications for abstinence referred directly or indirectly to child health. Often the duration was linked to developmental landmarks for the child, i.e.r abstinence was continued until the child could walk or was completely weaned. This review also indicated that patterns of postpartum 31 abstinence are changing and, in fact, probably have been for many years. The authors observed, for example, that since Islam imposes abstinence only during the first 40 days postpartum, this standard has been adopted in a number of SubSaharan societies where it was probably much longer in the past. Urbanization and education of women beyond the primary level will also probably shorten breastfeeding and postpartum abstinence. Unless these traditional practices are replaced by use of modern contraceptives, the result will be shorter birth intervals. This expectation is based on experience in other countries, and on differences that are now appearing between urban and rural residents and among women of different educational levels. In Nigeria, for example, the ideal birth interval in rural areas is three years, compared to two years in urban areas [Acsadi and Johnson-Acsadi, 1984]. The Caldwells' work in Nigeria also produced evidence of change [Caldwell and Caldwell, 1981]. Women with no schooling continued breastfeeding for an average of 24 months and abstinence for 30 months. Among women with at least secondary schooling the average period of breastfeeding was 10 months and that for abstinence was 11 months. The duration of abstinence and lactation in the intermediate eduational levels was between these two extremes. These findings of shorter periods of lactation and abstinence at higher educational levels are confirmed by WFS data from Ghana, Kenyar Lesotho, and Sudan. [Acsadi and Johnson- Acsadi, 1984). Table 8 shows differences by educational level 32 in duration of breastfeeding and postpartum abstinence. From the lowest educational level to the highest, mean duration of breastfeeding declined by 5-7 months, while abstinence declined by 0-4 months. (Note: Because these are national datar they do not reflect regional and ethnic differences.) Birth IntervLa Table 9 shows the proportion of births preceded by intervals of less than two years in 26 developing countries [Hobcraft, et al, 1983). The five African countries for which this information is available fall into two distinct groups. Among the 26 countries, there are only four countries in which less than 20 percent of births follow short intervals. Three of these are African -- Ghanar Lesotho, and Senegal. In Kenya and Sudant on the other hand, at least 35 percent of the births are preceded by short intervals. Given the importance to child survival of birth spacing, it is of interest to know what groups of women are most likely to have closely spaced births. Of perhaps even more interest is any evidence of trends in birth spacing. With the WFS data for three SSA countries -- Cameroon, Ghana and Kenya -- we were able to address these issues. The declines in breastfeeding and postpartum abstinence in SSA has been noted by a number of researchers. Such trends, it i s feared, may increase the frequency of short-interval births. We were surprised to find that in these three countries short birth intervals seem to be getting less common, rather than more so. When the fertility histories reported by women are divided into births that took place 10-14 years and 0-5 years before the 33 survey, the proportion of births with short preceding intervals decreased: from 34 to 29 percent in Cameroon; 25 to 20 percent in Ghana; and 44 to 38 percent in Kenya. These declines were apparent in each maternal age group. One possible explanation is that, since infant deaths have become less common, lactation amenorrhea is less often interrupted. This is, however, pure conjecture. More research is needed on trends in birth spacing and the reasons for such trends. Examination of the WFS data clearly shows that women younger than 20 are particularly likely to have short birth intervals. As Figure 10 showst in Ghana short birth intervals were almost twice as common among adolescents as among older women. There is no clear pattern in the relationship of birth spacing in these countries to either maternal education or area of residence, as Figures 11 and 12 show. Detailed local studies contribute to our understanding of birth spacing patterns and trends in SSA, and some of the evidence is hopeful. For example, the Caldwells found that the upper educational groups in Nigeria have begun to substitute modern contraceptives for postpartum abstinence. Data from Ghana on recent births indicate longer intervals among younger women and among those with secondary education, and especially long intervals among women in the largest urban centers. [Gaisie, 1984], This is probably evidence of effective use of contraception. Differences were also apparent in the age at stopping reproduction. In Ghana, women with some education stopped, on the averager about one year before those with no 34 education. Women in large urban areas stopped having children about two years earlier than did rural women. The Potential for Averting Infant and Clhild Deaths The information presented above allows us to estimate the effect on infant mortality of various changes in fertility patterns in Africa. Using WFS country-specific data on the proportions of births that follow short intervals, and on the risks associated with such intervals, we estimate that if all preceding intervals less than 24 months long were extended to 24-47 months, the proportion of infant deaths averted would be 12 percent in Lesotho, 19 percent in the Sudan, 5 percent in Senegal, 17 percent in Ghana, 22 percent in Cameroon, and 20 percent in Kenya. (See Appendix 1 for methodology.) If intervals were extended to four or more years, the reductions would be even greater. A number of points should be made about the above estimates. First of all, they refer only to infant mortality. whereas the WFS data clearly indicate that the deleterious effects of short birth interval persist for at least the first five years of life. Second, they represent only the effects of the preceding birth interval, whereas it seems that in the second year of life the subsequent interval is more important. Third, in Africa. where women generally bear children as long as they are abler extension of birth intervals would undoubtedly result in a decline in fertil- ity. This would, in turn, reduce the number of high-order births, further reducing infant mortality. Unfortunatelyr we have no way of knowing how big this reduction would be. In short, there are a number of ways in which our estimates of infant deaths prevented 35 by improved birth spacing are probably too low. The proportion of women who say they want no more children is smaller in SSA than in other parts of the world. Even sor if such women had no more births, we estimate that there would be smaller, but still substantial, reducticns in infant deaths: 8-11 percent in Ghana, Kenya, Lesotho and the Sudan. These estimates are based on the proportion of women of various parities who say they want no more childrenF and on country-specific relative risks of infant death by birth order. DISCUSSION The previous sections have delineated the risks to maternal and child health posed by unfavorable family formation patterns in SubSaharan Africa. In addition, we have constructed rough estimates of maternal and infant deaths which might be averted if more favorable patterns were adopted. For example, even though the proportion of women who say they want no more children is relatively small, if only these women had no more birthsr about one-sixth of maternal deaths would be averted. Improvement of birth spacing patterns, according to our estimates, would dramatically reduce infant deaths. If all birth intervals were at least two years long, we estimate that at least 12-20 percent of infant deaths in SubSaharan Africa would be averted. This is a truly remarkable pro*spect, and an attractive one also, given the favorable cultural background for birth spacing. The above estimate does not take into account the fact that fertility would almost certainly decline if birth spacing were increased, which would avert even more infant deaths. 36 Discussions of health and population policy in developing countries often get bogged down in arguments about the reciprocity of fertility and mortality. The ostensible object of these arguments is to establish priorities among various kinds of programs. We believe that consideration of family planning as a health intervention obviates most of these arguments. For example, one such argument is that reducing mortality without first reducing fertility will, in the long run, aggravate health problems through uncontrolled population growth. This argument simply does not apply to family planning, which reduces mortality and fertility at the same time. According to the available evidence, there are few initiatives that have the potential to decrease both fertility and mortality. Increases in urbanization, income, education and employment of women are all associated with reductions in infant mortality. Studies in various parts of the world have associated these same factors with declines in fertility, but (at least so far) the data from Africa are conflicting. For exampler while urbanization is generally associated with smaller family sizer in Africa it is also linked to the deterioration of traditional birth spacing practi ces, resulting in shorter birth intervals. It may be that this is a temporary effect. But even if this transitional phase lasts only a generationt any rise in the already hiah levels of fertility in Africa has serious implications for health. Thus, it is our opinion that whether the objective is prompt improvement of family health or long-term balancing of fertility 37 and mortality in order to bolster development efforts, an essential component of development effor';s in SubSaharan Africa should be programs that: a) educate families about the health risks associated with unfavorable reproductive patterns; and b) provide them with access to modern methods of contraception so that they can put into practice their knowledge and desires. 38 Table 1. Indices of Women's Health. Latin North Index Africa Asia America America Life Expectancy (years) 50 60 64 74 Maternal Deaths per 100,000 Live Births (range) l60-11100 7-l,0O0 16-458 7-15 Deliveries Attended by Trained Personnel (%) 28 51 62 100 Sources: Kent, 1983; WHOP 1983. Table 2. Maternal Mortality Rates* in SubSaharan Africar by Country and Type of Study. MMR per 100O0oo Country Period Births HosX-ial (Booked Patients) Kenya 1977 280 Nigeria 1966-1972 225 1966-1975 98 1970 159 1970-1971 530 1971-1972 88 1970-1974 142 230 1976-1979 130 Zimbabwe 1976 59 Malawi 1977 103 Niger 1980 135 Nigeria 1961-1969 .172 Uganda 1967 126 Ethiopia 1981-1983 566 Ghana 1972 400 Senegal 1962-1982 700 * Rates are usually per 100000 total birthst rather than live births. Sources: Hgadtii Series: Kenya, 1977 -- Aggarwalt 1980; Nigeria, 1966-1972 -- Akingbat 1977; Nigeria, 1966-1975 , Ayangade, 1981; Nigeria, 1970 -- Hartfield, 1980; Nigeria, 1970-1971, Waboso, 1973; Nigeria, 1971-1972 -- Megafut 1975; Nigeria, 1970-1974 -- Okoisor, 1978; Nigeria, 1976-1979 -- Harrison and Rossiter, 1980; Zimbabwer 1976 -- Brown, 1981. Oficial Records: Malawi, 1977 -- Bullough, 1981; Niger, 1980 -- Thuriaux and Lamotter 1984; Nigeriar 1961-1969 -- Hartfield and Woodland, 1980; Uganda, 1967 -- Grech, Galea and Trussell, 1969. jg)mmunily Studija: Ethiopia, 1981-1983 -- Kwast, et al, 1985; Ghana, 1972 -- Department of Community Health, 1979; Senegalr 1962-1982 -- Family Health International, 1983. Table 3. Percent of All Births to Women in Various Age Groups, WFS. 15-19 + Country 15-19 20-24 25-29 30-34 35-39 40-49 35-49 Benin 10 28 29 18 10 5 25 Cameroon 23 28 21 14 8 6 37 Egypt 14 31 27 16 9 3 26 Ghana 15 27 23 16 11 9 34 Ivory Coast 19 30 23 13 8 6 34 Kenya 10 21 22 18 15 13 38 Lesotho 14 30 23 16 10 7 31 Nigeria 18 25 25 18 8 6 33 Senegal 19 25 24 15 11 6 36 Sudan 11 32 16 11 5 27 Tunisia 6 29 27 17 13 9 28 Source: WFS First Country Reports Table 4. Women Aged 45-49 Who Have Borne a Given Number of Childrer) WFS. Nunber of Children Country 7-8 9+ 7+ Cameroon + 17 20 37 Egypt * 25 33 58 Ghana * 28 29 57 Ivory Coast + 20 37 57 Kenya + 26 47 73 Lesotho + 19 18 37 Sudan * 23 26 49 Tunisia + 26 34 60 * Percent of ever-mrried women. + Percent of all women. Source: WFS First Country Reports Tale 5. Percent of Women Who Want No More Children, and Percent of Those Who Want No More Currently Using an Efficient -Method of Contraception, WFS. Want No More Using Efficient Country Children * Contraceptive + Cameroon 10 6 gypt 53 46 Ghana 12 17 Ivory Coast 4 4 Kenya 17 17 Lesotho 15 16 Nigeria 5 4 Sudan 17 16 lunisia 47 48 * Currently marriedo fecund women + "Exposed" women who want no more children. Source: WFS First Country Reports Table 6. EstimateA Proportion of Maternal Deaths that Would be Averted If Women 15-49 Who Want No More Children Had mb More * Country % Egypt 28 Ghana 15 Ivory Coast 5 Kenya 15 Lesotho 18 Nigeria 6 Sudan 18 Tunisia 26 * Corrected for women who are already using an efficient method of contraception. Note: See Appendix A for methodology. Table 7. Estimated infant and, child mortality rates Ln 1982, ard percent decrease since 1960, SubSaharan Africa. Region & Infant Mortality Child Mortality Ccuntry Rate % Decrease Rate % Decrease Wte st Aer Benin 117 32 23 45 Ghana 86 35 15 44 Ivory Coast 119 29 23 42 Liberia 91 47 16 62 Mali 132 26 27 40 Mauritania 132 26 27 40 Niger 132 26 27 40 Nigeria 109 43 20 60 Senegal 155 13 34 24 Sierra Leone 190 39 50 30 Togo 122 39 25 54 Upper Volta 157 33 36 49 (Bourkina Fasso) Middle Afriga Angola 165 24 39 38 Camneroon 92 31 16 43 Central African Republic 119 30 23 44 Chad 161 23 37 38 Congo 68 42 10 56 Zaire 106 29 20 38 Lesotho 94 29 23 42 Sudan 119 29 23 42 Burundi 123 14 24 22 Ethiopia 122 29 25 40 Kenya 77 31 13 38 Madagascar 116 34 23 49 Malawi 137 33 29 52 Mozanbique 105 32 20 41 Rwanda 126 24 25 38 Somalia 184 14 47 23 Tanzania 98 32 18 45 Uganda 120 14 22 21 Zambia 105 36 20 47 Zimbabwe 83 17 14 26 Note: Infant mortality rate is deaths under one year of age per 1,000 live births. Child mortality is deaths at ages 2-5 per lsOOO children of these ages, per year. Source: World Bank, 1984 Table 8. Mean duration (in months) of breastfeeding and postpartum abstinence in last closed pregnancy interval, by educational level, W'S. Lesotho Sudan Kenya Ghana No Schooling 21 17 14 16 Loer Primary 20 14 14 14 Upper Primary 20 11 12 13 Secondary or 15 na 9 9 Higher Difference -6 -6 -5 -7 No Schooling 14 6 4 10 Lcwer Primary 14 5 4 7 Upper Primary 14 4 3 7 Secondary or 11 na 4 6 Higher Difference -3 -2 0 -4 Source: Acsadi and Johnson-Pcsadi, 1984 Table 9. Percent of All Births Following an Interval of Less than 24 Months, WFS. Region 16-19 20-24 25-29 30-34 35-39 40+ Africa Ghana Caneroon Kenya Lesotho Sudan Senegal K a Korea Nepal Indonesia Bangladesh Malaysia Jordan Pakistan Sri Lanka Philippines Syria Thailand Lati A Haiti Ecuador Costa Peru Rica Colcmbia Guyana Jamaica Mexico Panama Source: Hocraft et. al., 1983; Cameroon WFS Tape Figure 1. Maternal Mortality, by Age, "High Mortality" Countries 5.0 44. 8 4.0 3.0 CZ 2.0 1.5 - 1.3 1.0 - .- < 20 20-24 25-29 30-34 35-39 40-44 45-49 Age Source: Nortman, 1974 Figure 2. Maternal Mortality, by Age, Zaria Hospital, Nigeria, 1976-1 9 79 7.0 6.8 6.0 5.0 4.0 3.0 2.5 2.6 2.0 1. 3 1 .3 1.0 F- L - L 15 16 17- 20-24 25-29 30-34 35-39 40-44 45-49 19 Age Sources: Harrison and Rossiter, 1980. Figure 3. Maternal Mortality, by Parity, Dakar, Senegal, 19 71-1 9 75 5.0 4.0 4.0 - 3.0 2.9 2.9 2.2 CJ 2.0 1 2 3 4 5 6 7 8 9 10 11 1 2+ Parity Source: Correa, 1978. Figure 4 Infant Mortality by Birth Order, 9 Sub-Saharan Countries 1 8 O , 170 160 ,-Lesotho 1 50 140 /-Benin 130 / 7 , -Ivory Coast 1 2 0 \ 110 negl Y"~ ICameroon >> llO \<~ enegal . , 4, 100 / X-Kenya 0 9 0 -Mauritania \ ' -Ghana 8 0 \> - Sudan 70 Ixb ,. 60 1 2-3 4-6 7+ 1 2-3 4-6 .+ 1 2-3 4-6 7+ Birth Order Source: Rutstein, 1984. Figure 5 . Infant Mortality by Maternal Age, 9 Sub-Saharan Countri es 160 0 ;-Ivory Coast 150 140 \ 1 \~ '-Lesotho 130 \/ 1 20 \ /-B enin 110 -Senegal 100 \ ,-Cameroot \-Kenya 90 \ 0 80 -Mauritania ' Sudan 70 '-Ghana 60 20 20-29 30-39 .20 20-29 30-39 4'20 20-29 30-39 Maternal Age Source: Rutstein, 1984. Figure 6. Infant Mortality by Preceding Birth Interval (in Months) Senegal Benin Mauritania .1.8 1.3 .01.5 8 1. L'O c 24 24-47 w47 .24 24-47 ;47 2 24 74-47 >47 Cameroon Lesotho Ivory Coast 1.8 c1i4 4 c 24 24-47 o47 G24 24-47w S '24 24-47 -47 Sudan Kenya Ghana 2.1 1.6 1.7 < 24 24-47 '47 < 24 24-47 -47 c 24 24-47 w47 Source: Derived from Rutstein, 1984. Figure 7. Infant Mortality rates among children with preceding birth intervals of less than 2 yearst 2 years, and 3 years* Cameroon Ghana D D e e a a t 150 140 t 150 h h s 125 S 125 114 P100 82 P 100 e e r 75 r 75 5 56 1 50 1 50 41 0 0 0 25 0 25 0 0 0 0 1 7----------------D----------------- <2 2 3 <2 2 3 l1e-a Ya ears ryla er Kenya D e 200 a t 150 h 106 P 100 e r 75 49- 1 50 0 0 25 0 0 (2 2 3 * Probability of death during the first year of life per lOOO live births. Interval categories are <24, 24-35r and 36-47 months. Source: WFS data tapes Figure 3. Toddler Mortality by Preceding Birth Interval (in Months) Senegal Benin Mauritania 1. 1.I2 1. O lo 1.0 c24 24-47 >47 G 24 24-47 7 47 < 24 24-47 -47 Cameroon Lesotho Ivory Coast 1.6 1.5 1.0 1.0 1e0 1.0 1.0 . 24 24-47 >47 424 24-47 >47 424 24-47 4 7 J Sudan Kenya Ghana 1.4 1.4 1.6 - 1.0 -. 1eO I.O 1 7 5 6 < 4 24 24-47 t47 ' 24 24-47 >47 < 24 24-47 -47 Source: Derived from Rutstein, 1984. Figure 9. Relative Risk of Mortality by Preceding and Subsequent Birth Interval, ages 1-2 and 2-5 AGE A;E AGE AGE 1-2 years 2-5 years 1-2 years 2-5 years I 8 1 .4-1. 4'1 .2 5 Lesotho 1.0 1.0 Sudan 1.0 1- A B C B D A B C B D 145 1 . 4 LE. 5 Kenya a95 Ghana 1.0 X- I A B C B D A z C B D A - Base, no close births B - Preceding birth within 24 months C - Subsequent birth within 18 months D - Subsequent birth within 30 months Source: Derived from Hobcraft, McDonald and Rutstein, 1983, Tables 5 and 6. Figure 10. Percent of children with preceding birth intervals less than 2 years long, by maternal age Cameroon Ghana 50 45 50 39 P 40 P 40 e 32 27- e r 3 0 r 30 c c 22 21 e 20 e 20 n n t 10 t 10 0 0 <20 20-29 30+ <20 20-29 30+ maenal Aatenal Auh Kenya 60 5 2 50 42 P 40 36 e r 3 0 C e 20 n t 10 <20 20-29 3 0+ Source: WFS data tapes Figure 11. Percent of children with preceding birth intervals of less than 2 yearst by maternal education Cameroon Ghana 50 50 P 40 P 40 e 31 31 e r 30 r 30 24 c c 23 --- 22 e 20 e 20 n n t 10 t 10 NA* 0 0 0 1-3 4+ 0 1-3 4+ aS Edtation Years Ediaration Kenya 50 44 39 39 - P 40 e r 30 C e 20 n t 10 0 0 1-3 4+ Years E tiz * Not applicable Source: WFS data tapes Figure 12. Percent of children with preceding birth intervals of less than 2 years, by area of residence Cameroon Ghana 50 50 P 40 P 40 e 30 32e r 30 r 30 23 24 C c 19 e 20 e 20 n n t 10 t 10 0 0 Rural Urban Rural Small Large Urban Q&g Rdenceienc Kenya 47 50 50 39 P 40 e r 30 C e 20 n t 10 Rural Small Large Urban e 1 E Source: WFS data tapes References G.T. Acsadi and G. Johnson-Acsadi, D-e fp.rd r! and Spa in agbkahgr. ArfXjiga Rev. text, Unpublished, February 1984. S.O. Ayangader "Maternal Mortality in a Semi-Urban Nigerian Communityr" joqUjn.1 of tha Esiti-Onsl Madical Ascai 73(2):137-140, 1981. V.P. Aggarwalv "Obstetric Emergency Referrals to Kenyatta National Hospital," Te EXat African Medicgl JTourn.i 57(2):144- 149, February 1980. J.B. Akingba, "Abortion, Maternity and other Health Problems in Nigeria," Nig ri a aedical Journal, 7(4):465-471, October 1977. D.A. Ampofo, "Causes of Maternal Death and Comments, Maternity Hospital, Accra 19631967 " WeXt rig-an Medical JQouralM , 18(3):75-81, June 1969. W.ZO Billewicz and I.A. McGregor, "The Demography of Two West African (Gambian) Villagest l951-75," ougnal B i oscial rjaflQj 13(l):219-240Y January 1981. J.T. Boerma and H.A. van Viened, "Birth Interval, Mortality and Growth of Children in a Rural Area in Kenya," JoQ1Lnaj o Baiosocial sienget 16(4):475-486, 1984. I.McL. Brown, "Perspectives in Obstetric Care for Zimbabwe," The Central ican J nl - Meiga 27(3):37-41, March 1981. C.H.W. Bullough, "Analysis of Maternal Deaths in the Central Region of Malawi," Thje Egst rgLi£n MIedical, Journal,' 58(1):25-36, January 1981. J.C. Caldwell and P. Caldwell, "Cause and Sequence in the Reduction of Postnatal Abstinence in Ibadan City, Nigeriaia in, H.J. Page and R. Lesthaeghe, eds., CId-Spin TLopigal Aica: :raditionge A Academic Press, London, 1981, pp. 181-199. P. Caldwell and J.C. Caldwell, "The Function of Child-Spacing in Traditional Societies and the Direction of Change," in, H.J. Page and R. Lesthaeghe, eds.e £hijdSIp._a_ng in TrQpical Ariga: drxition= and Change,_ Academic Press , London, 1981, pp. 73-92. Cameroon. Ministere de L'Economie et du Plan, Direction de la Statistique et de la Comptabilite Nationaler EnguteNtoap la ndi& -1idu clmfrlBon 192i8 Ralmo2t P Linc Volum& 1E 3Er . S ..tigi guez (World Fertility Survey), Yaounder Cameroon, April 1983. Cameroon. Ministere de L'Economie et du Plan, Direction de la Statistique et de la Comptabilite Nationalee, qget Niationla 5UL, 2£ FeQdi. u Camerouna 1978, EappQr- t Princ l olme 11:L- 2± T Statistiguesi (World Fertility Survey), Yaounde, Cameroon, April 1983. P. Cantrelle and H. Leridon, "Breastfeeding, Mortality in Childhood and Fertility in a Rural Zone of Senegal.." popJ ion Stui-± 25(3):505-533 t 1971. L.C. Chent M.C. Gesche, S. Ahmed, A.I. Chowdhury and W.H. Mosley, "Maternal Mortality in Rural Bangladesh," Stdiea in amily Rlalnning. 5(11):334-341P 1974. P. Correa, F. Diadhiou, J. Lauroy, M.D. Bah et F.A. Adanleter "Mortalite maternelle au cours de la parturition a la clinique gynecologique et obstetricale de Dakar," Bujjpt.i n D_ La Soi Medical& j21AjLi. 2N7i_e _D_ Lanlgu Franucais-e, 23 (3): 262-269, 1978. Department of Community Health, University of Ghana Medical School and UCLA School of Public Health, Division of Population, Family and International Health, University of California, "Health Status at Baseline," in The PAnfa C-omRr hnural Helth = Familg nnig Prole±tL Ghana Einal Eq_ort. Accra, 1979r pp. (5-9)-(5 12) P. Doyle, D. Morley, M. Woodland and J. Cole, "Birth Intervals, Survival and Growth in a Nigerian Village, "JQournl 9 fliogrial Sgieigi, LO(8):81-94, 1978. N.J. Eastman, "The Effect of the Interval Between Births on Maternal and Fetal Outlook," Amerigan Journal of Qbtctrcs and Qynecoloag.- 47(4):445-465, 1944. Egypt. Central Agency for Public Mobilisation and Statistics, Th. EaDian Fertility Sugrvey J i 9 0 gVlume 1IIL Pert ili±a .d EAm kl=nings (World Fertility Survey), Cairo, 1983. Egypt. Central Agency for Public Mobilisation and Statistics, Th&. yijn Fertility Su1vey- 1.2 Q 3olum f III, So iQ-E on Q Differentials n C-Qmatia D frgm ksnsa x w i.A (World Fertility Survey), Cairot 1983. Egypt. Central Agency for Public Mobilisation and Statistics, The Egypti-an Fertlit mLey 90 3LOlM IY itgl Tablas' (World Fertility Survey) t Cairo, 1983. Family Health International, "Study of Maternal Mortality in Senegal Planned," Network, Q 5(l):2, 1983. J.A. Fortney and S.M. Rogers, personal communicationt October 1984. J.A. Fortney, S. Saleh' S. Gadalla and S.M. Rogers, "Causes of Death to Women of Reproductive Age in Egypt,r paper presented at the annual meeting of the Population Association of America, Pittsburghi April 14-16, 1983. S.K. Gaisie, "The Proximate Determinants of Fertility in Ghana." WFS Scientific Report No. 53r International Statistical Institute, Voorburg, Netherlands, March 1984. Ghana. Central Bureau of Statistics, Gbara etili±y Burey 1279- 1980Q LoIqXma 1L BaQ1m, o r Methodo logy and indinagf E.irt ReportL (World Fertility Survey)r Accra, Ghana, 1983. Ghana. Central Bureau of Statistics, Oam3 Fertility- 'lusy 129279. 1--. Volgme IL StatiU&il Tkblea First 1L__ (World Fertility Survey), Accra, Ghana, 1983. R.H-. Gray, "Maternal Reproduction and Child Survival," Am.!xJgAn J0garal DI Public palhLL 74(10) :1080-1081, 1984. E.S. Grech, J. Galea and R.R. Trussell, "Maternal Mortality in Uganda," Internional J_u1nA. t Gvnaecology and ObQtetric-s 7(6) :263-278, December 1969. E.S. Grech, "Obstetric Deaths in Lusaka.." M1jical Journal P-L mZnbia, 12(2):45-53p, April-May 1978. D.R. Gwatkinr "Signs of Change in Developing-Country Mortality Trends: The End of an Era ?" Development Paper 30, Overseas Development Council, Washingtonf D.C., February 1981. K.A. Harrison and C.E. Rossiter, "Zaria Materni ty Survey 1976-79. VII. Maternal Mortality," Unpublishedt E1980]. V.J. Hartfield, "Maternal Mortality in Nigeria Compared with Earlier International Experience," International Joural f alog and O=teicst 18(1) :70-75, 1980 n V.J. Hartfield and M. Woodland, "Prevention of Maternal Death in a Nigerian Village," In - LterAI Joural SL Affgology and Qbget:=cs t 18:150-152 , 1980 . C. Haubt D.W. Heisler and M.A. Condron, World's Wmen Data Sheet, Population Reference Bureau (in collaboration with UNICEF)f Washington, D.C., April 1980. M.U. Hickey and J.M. Kasondei, "Maternal Mortality at University Teaching Hospital, Lusaka," Medical Jogrn.al p- Zambia, 11(3):74- 78t June-July 1977. J. Hobcraft, J.W. McDonald and S.0. Rutstein, "Child-Spacing Effects on Infant and Early Child Mortality," PopgalAt-on Index., 49(4):585-618, Winter 1983. J. Hobcraftr J.W. McDonald and 5.0. Rutsteinr "Demographic Determinants of Infant and Early Childhood Mortality: A Comparative Analysisr" World Fertility Survey Technical Report 2370w June 1984. E. Hughes, In3tryAl et wn LthLa U.S. Children's Bureau, No. 112, Washington. D.C., 1923. Ivory Coast. Ministere de L'Economie et des Financest Direction de la Statistiquey Mete iszirn gul la. Ferondite 1980-81 ERport 12rinclpal- Q1mrn IL Analya des P caUX Resul 9s Abidjans 1984. Ivory Coast. Ministere de L'Economie et des Financest Direction de la Statistique, agqete lyoir ienne sur 1AE.cod:p 1980-81, Bapo- 1Erciaa Y-ma1. Tableaux i egr Abidjant 1984. D. B. Jelliffe sem be Nutritional L Qf the .Qmmanitjiy UiWi pegial Refera Lo Field Sryeys in 1alg Counriaes& World Health Organization, Genevat 1966. M.M. Kent, 1983 Ii o--1i.nd PopglaJion Data 5hg Population Reference Bureau, Washingtonrt D.C.e April 1983. Kenya. Central Bureau of Statistics, Kenya Eer rey 12227- 12--li. YZoJEaiIIa.m I. F ortr (World Fertility Survey)r Nairobi Kenya, Februaryp 1980. Kenya. Central Bureau of Statistics, Emya Fei a Lurvev 19227- j12-. Yolume 2± Uij.. Report, (World Fertility Survey), Nairobi, Kenyar February, 1980. B.E. Kwast, W. Kidare-Mariam, E.M. Saed, and F.G.R. Fowkes, "Epidemiology of Maternal Mortality in Addis Ababa: A Community- Based Study," EithiQian di£al Journal, 23(l):7-16, Jan. 1985. W.H. Lamb, C.M, Lamb, F.A. Foord, R.G. Whitehead, "Changes int Maternal and Child Mortality Rates in Three Isolated Gambian Villages Over Ten Years," -tnha Langet, October 20t 1984. J.V. Larsen, C.L. Msaner M.C. Monkher "The Fate of Women Who Deliver at Home in Rural Kwazulu," S-ouLth Afri£An Medical J.ouLnali 63(2) :543-545 April 1983. Lesotho. Central Bureau of Statistics, esotho Frti su2ey 1 977 iQme 2. Eirst o (World Fertility Survey), Maseru, Lesotho, 1981. Lesotho. Central Bureau of Statistics, Lesotho eUility Suarvy 192277 Yolume. I Eirat Baqe=pnf (World Fertility Survey), Maseru, Lesotho, 1981. R. Lesthaeghet P.O. Ohadike and H.J. Pages "Child-Spacing and Fertility in Sub-Saharan Africa: An Overview of Issues," in, H.J. Page and R. Lesthaeghe, eds., ii in rpirQal. i . a. Tr°dtins A Chngage Academic Press, London, 198l pp. 73-92. L. Liskin, "Complications of Abortion in Developing Countries," opulation BRepoLrs Series F, No. 7, 1980. D. Maine, Family Planning: It. £Impact nn t h. Hie2th o1 Womex =d hildz=n' Columbia University, Center for Population and Family Healtht New York, 1981. R. McNamara, B.ix.ta Intervzya and MortaliY in Ifanc w Ch idhood:_ A 5tuLdy of tiye Xi-k in Inesia n M£Q2 Doctoral Dissertation, Columbia University, New York, 1985. U. Megafur "Causes of Maternal Deaths at the University of Nigeria Teaching Hospital Enugu 1971-1972," The klegt AfriL-q Iigi agi.urn L, April, 1975. J.N. Morris and J.A. Heady, "Social and Biological Factors in Infant Mortality, V_r Mortality in Relation to Father's Occupation, n atr 1:554-560, 1955. National Center for Health Statistics, VZital Statistics 91 t 5. 1dSt agIVl IIL M-LtalityL Eart Ai. Public Health Service, Washington, D.C., 1984, pp. 242, 243, 284. New York Academy of Mediciner Committee on Public Health Relations, "Maternal Mortality in New York City: A Study of all 'Puerperal Deaths 1930-1932r" Commonwealth Fund, New York, 1933P pp. 4-5. Nigeria. National Population Bureau, higrid. Erilit y IL92z.. va PorZt Y.Llap. L Statistical T.bleg, (World Fertility Survey), Lagos, 1984. D. Nortman, "Parental Age as a Factor in Pregnancy Outcome and Child Development," Report Qa = pujatIianZFamiy R jngj, No. 16r The Population Council, New York, August 1974. O.A. Ojo and V.Y. Savage, "A Ten-Year Review of Maternal MZortality Rates in the University College Hospital, Ibadan, Nigeria," Amn gi&Can JQnal 2L . 91 : an nQdQIz 118(4):517-522, February 15, 1974. A.'T Okoisor, "Maternal Mortality in the Lagos University Teaching Hospital: A 5-Year Survey-1970-1974,r" Nige.ria.nn Mg&a ,IsLnaIr 8(4):349-354, July 1978. A.R. Omran and C.C. Standley, edsc, EfimiiL EQ mg Patterns1La . Ithi± ha natin QhnA Collaborative -tns ia Indiar Iran h ii2egti.. World Health Organizationp Geneva, 1976. A.R. Omran and C.C. Standley, eds., E Qmily aqtio o n Heal U Furthe tie. World Health Organization, Genevar 1981. H.W. Ory, J.D. Forrest and R. Lincoln, Making ce Evalutin the Hgalth RisE a,nd S it. 0. Bh CQntrol M hoQs Alan Guttmacher Institute, New York, 1983. H.W. Ory, A. Rosenfield and L.C. Landman, "The Pill at 20: An Assessment, International Family nnninn Perspegtives£ 6(4):125- 130, 1980. R.W. Rochet, D. Kramer, P. Senanayake and C. Howellp "Induced Abortion and Health Problems in Developing Countries," The Langetr 2(8192):484, 1980. S.O. Rutstein, "Infant and Child Mortality: Levels, Trends and Demographic Differentials," WFS Comparative Studies: Cross- National Summaries, No. 24, International Statistical Institute, Voorburg, Netherlands, September 1983, pp. 5-36. S.O. Rutstein, "Demographic Differentials in Infant and Child Mortality," WFS Comparative Studies Cross-National Summaries: Preliminary Tables, Unpublished, 1984. R. Schoenmaeckers, I.H. Shah, R. Lesthaeghe and 0. Tambashe, "The Child-Spacing Tradition and Postpartum Taboo in Tropical Africa: Anthropological Evidence," inr H.J. Page and R. I4esthaeghe, eds., Child-Spacn in iaOP . 1Afi Traditions Changer Academic Press, London, 1981 pp. 25-71. Senegal. Ministere de L'Economie et des Finances, Direction de La Statistique, Division des Enquete et de la Demographier ^; DL su laet gl ip r la Fe%onditg 121EL MRVpXt L 1na 'nalng1eL 3lA1ume. Lt (World Fertility Survey), Dakar, Senegal, 1981. J.C. Smith, J.M. Hughes, P.S. Pekow and R.R. Rochatt "An Assessment of the Incidence of Maternal Mortality in the United States," Amerjcgn Jgurnal f Publig Healt 74(8) :780-3, 1984. Sudan. Ministry of National Planning Department of Statistics, Th -gdanl Fetlv fiurvey .122!L YQ-um.QI- Pr Xi a I Rg--pQ (World Fertility Survey), Khartoum, November 5, 1981. Sudan. Ministry of National Planning Department of Statistics, The 5_Lg FEr lit Suygy .1 olume I-L I 2 e R, 2Xr (World Fertiltiy Survey), Khartoum, November 5r 1981. I * Sw en son, The f o.- P r aang. A Qt2i a e I e t&d Mterna Fagtoa n 1 e±.al.L Infant =Arly Chilhood ortgaiity in Rraal Doctoral dissertation, Johns Hopkins University, Baltimore, 1976. I. Swensont "Expected Reductions in Fetal and Infant Mortality by Prolonged Pregnancy Spacing in Rural Bangladesh," BAn.ladph Devlopmnt Sieg-. 5(10:1-16, 1977. M.C. Thuriaux and J-M. Lamotte, "Maternal Mortality in Developing Countries: A Note on the Choice of Denominator," 1nje.Qrnationaj jogranal 21. Epidemi ology, 13(2):246-247f 1984. J. Trussell and A.R. Pebley, "The Potential Impact of Changes in Fertility on Infant, Child and Maternal Mortality," . in Family Plannina, 15(6r Pt. 1) :267-280, 1984. Tunisia. Office National du Planning Familial et de la Population, EngUC.e Tnjia Q afla. in l eron 1978 Rappoht jrL inial-VZl=ma If (World Fertility Survey) f Tunis, Tunisia, December, 1982. Tunisia. Office National du Planning Familial et de la Population, Se T sur jA Beaoodit. 127XR klincipal-Vol-ume 2' (World Fertility Survey), Tunis, Tunisia, December 1982. United Nations, Department of International Economic and Social Affairs, Statistical Office, _ma £ Yearbook, United Nations, New York, 1984t p.341. United States Bureau of the Census, "Infant, Maternal, and Neonatal Death Rates, and Fetal Death Ratios, by Race: 1940 to 1977t" [Table No. 108], in Statist ica Ajstract d1 IInied z.te 100th edition, Washington, D.C., Government Printing Office, 1979, p. 75. M.F. Waboso, "The Causes of Maternal Mortality in the Eastern States of Nigeriar" Nigerianen dical Journal, 3(2), April 1983. B. Williams, "Maternal Mortality in Sierra Leone,"l Xini;stry_ 1t Hial-th, Freetown, Sierra Leone, 1979, pp. 1-21. C.D. Williams, "Child Health in the Gold Coast," L909Pt, 1:97- 102, 1938. R.M. Woodbury, Causal Factr in Infa_nt. Morality: a Statistical dBased o Invs!-tons in Eight Ci.ties, U.S. Children's Bureau, Not 142, Washingtont D.C.r 1925. World Bank, WoclA Deyelom2nt R Oxford University Press, New York, 1984. World Fertility Survey. E:l uete Fecondite A BaniL 192i Resume da Rgultats, No. 48, Voorburg, Netherlands, International Statistical Institute, June 1984. World Fertility Survey. Thl Nrian iertilit > I a1z2.L A Samm._o 91 E1aa_a, No. 49, Voorburg, Netherlands, International Statistical Institute, September 1984. World Health Organization, Maternal and Child Health Programme, "The Health of Mothers and Children: A Brief Overview," Unpublished, November 1983. J.D. Wray, "Population Pressure on Families: Family Size and Child Spacingr" lepogt_a pn Population/Family Pl.anninaa No. 9, The Population Council, New York, August 1971. N. Younis, A. Ibrahim, G. Surour and A. El Hemaly, "Analysis of Maternal Deaths in Two Egyptian Maternity Hospitals," -Pp.!aatign gcene, (1): 59-65 P 1979. APPENDIX I: METHODOLOGY Proportion of Births to Women of Various Ages 1. For each country, the number of women in each age group was multiplied by the age specific fertility rate for that group, which yields the number of births to women in each age group. Added together these produce the total births. WFS First Country Reports were used. 2. The number of births to women in each age group were divided by the total number of births, giving the proportion of births to women in that age group. Potential for Averting Maternal Deaths 1. We constructed a synthetic cohort of childbearing w;omen, using country-specific data from WFS First Country Reports. First we ascertained the distribution of all women by five-year age group. For women of reproductive age (15 -49), we multiplied this proportion by the age-specific fertility rate (ASFR) for the appropriate age-group. In order to work with sufficient numbers, the product was multiplied by one million. This gives us the number of births per year in each age group for a cohort of women conforming to the age and fertility patterns of the country. 2. The number of births to women in each age group were then multiplied by an arbitrary maternal mortality rate (.002). Note that the rate of maternal mortality used here does not affect the results, as we are determining the, proportion of deaths averted, not the number. The product is multiplied by the relative risk of maternal mortality in that age groupt derived from Nortman's data for "high mortality" countries [Nortman, 1974, Table 21. This produces the number of maternal deaths expected, given current fertility patterns. 3. To estimate deaths avertedt we began with the proportion of women in each age group who said that they want no more children [Table 3.1.1.]. This we discounted by the proportion of such women who were already using an efficient method of contraception. To do this, we multiplied the percent of women who want no more children by l-(percent of these women using efficient contraceptives) [ Table 5.2.3.A]. The product is the percent of women who want no more children and are not using efficient contraceptives. This is multiplied by the number of maternal deaths expected in that age group in our synthetic cohort, producing the number of deaths averted under these conditions and assumptions. This process is repeated for each age group and the deaths averted summed. The total is simply divided by the total number of expected deaths to derive the proportion 4. If women in an age group are assumed to have no more children (i.e.r tiers 2 and 3 of Table 6) then all deaths occuring in the age group (rather. than just those among unprotected women who want no more children) are added to the numerator. E WJ aged 15-39 in Lesoth .048 = proportion of all women 35-39 at time of survey e169 = their age-specific fertility rate .048 x .169 x 14000,000 - 8,112 births per year e002 = arbitrary overall maternal mortality rate 3.17 = relative risk of maternal death at 35-39 8,112 x .002 x 3.17 = 51 maternal deaths .275 = proportion of women 35-39 in Lesotho who want no more children .203 = proportion of these women currently using an efficient method of contraception .275 x 1 - (.203) = .219 = proportion of women 35-39 who want no more children and are not using an efficient method of contraception .219 x 51 = 11.2 = deaths averted if these women had no more births. This process is repeated for-each age groupt the deaths averted summed and divided by the total deaths, to give the proportion of deaths averted. Multivariate Analysis of Factors Affecting Infant Mortality Logistic regression (SAS: Proc Logist) was selected as an appropriate method for analysis of mortality as a binary dependent variable. The procedure employs maximum likelihood estimation to calculate the log of the ratio of the odds of the probability of dying and the probability of not dying, in the presence of selected conditions. It approximates the relative mortality risk associated with a particular value of some factor (e.g., relative risk of death at a specific birth interval or birth order). The multiple regression equation takes the form: P (Yu ) In 80 + aIXi 82X2 .X. + SkXkI where Bo = constant Bi = regression coefficients Xi = predictors In the multivariate analysis, the odds of dying when the birth interval is less than 24 months and 24-35 months was com-. pared to those odds when it is 36-47 months. This permitted observation of relative risk associated with both short and medium length intervals. Reference categories for the other variables were a) birth orders 2 and 3, b) maternal age 20-29, c) some maternal education, and d) survival of the preceding sibling to age two. Potential for Averting Infant Deaths Deaths Averted By Birth Spacing We assume that no woman really wants to have a short interval birth, or that, converselyr no woman would mind waiting a few extra months to become pregnant again. Consequently, our calculations in this case assume that nQ Mou short-interval births take place. This simplifies calculations considerably. 1. For each country, we determined the proportion of all births that followed short intervals (less than 24 months) [Hobcraft, et aly 1983]. 2. We determined the infant mortality rates among infants born after short and long intervals [Rutstein, 1983, 1984]. 3. Then we calculated the percent attributable riskr using the following formula: Io - In Io Where: Io = infant mortality rate in the total population = WnIn + WsIs In = infant mortality rate after long interval (24-48 months) Is = infant mortality rate after short intervals (<24 months) Wn = proportion of births at long intervals Ws = proportion of births at short intervals Wn = .65 Ws = 635 In = 68 Is = 117 Io = .65(68) + .35(117) = 85.15 % deaths averted = 85 - 68 = 2 0 % 85 Deaths Averted By Reducing Birth Order To estimate the proportion of infant deaths that would be averted by reducing birth order, we constructed a synthetic cohort similar to that used to estimate maternal mortalityr except that women are grouped by parity, not age. Unfortunately, the published WFS data are not arranged to make this a simple process. 1. The distribution of ever-married women by number of children ever born was used to determine the order of their last birth. From a cross-tabulatiLon of age by parity, we derived the parity distribution of (last) births in each age group [Table 2.2.11. This was multiplied by the total number of births to women in that age group in our synthetic cohort (see step 1 of maternal mortality), which yielded the number of births by age and parity. When summed across parity, we had the parity distribution in our synthetic cohort, having taken into account the age distribution and ASFR of the country. 2. The number of births in each parity group are then multiplied by the overall infant mortality rate and the relative risk associated with parity in that country [Kent, 1983; Rutstein, 1983, 19841. This gives us the number of infant deaths by birth order. 3. The remaining steps are similar to steps 3 and 4 of the maternal mortality calculations. The proportion of women at various parities who want no more children (Table 3.1.2) was multiplied by l-(the proportions of those women already using efficient methods of contraception) (Table 5.2.3.A). One slight difference is that the proportion of women who want no more children was figured using the parity at last birth minus one. For exampler we multiplied the proportion of women whose last birth was their seventh by the proportion of women whor at parity 6, said they did not want another child. The product was then multiplied by the number of infant deaths in that parity group, derived above. APPENDIX 2 Notes to Appendix Tables 1-4: In the logistic regression analysis, four models were used: Model I included preceding interval onlyl Model II included the interval and maternal age, Model III included interval and birth order, Model IV, the "full model," included interval, maternal age? birth order, and maternal education. The relative risks associated with any one variable in Model IV are therefore controlled for the influence of the other vari- ables. Risks are relative to those in the following reference categories: Preceding birth interval 36-47 months. Maternal age 20-29. Birth order 2-3 Maternal education 1+ years (i.e., any formal education) * indicates .05 significance ** indicates .01 significance Appendix Table 1. Relative risk of neonatal mortality Model I Model II Model III Model IV CAMEROON Preceding interval < 24 mo 3.3 ** 3.5 24-35 mo 1.7 ** 1.8 48+ mo 1.4 1.4 Maternal age < 20 1.0 1.0 35-39 1.4 ** 1.3 Birth order 7-8 1.1 1.2 9+ 1.5 ** 1.1 Maternal education none 1.8 GHANA Preceding interval < 24 mo 2.7 ** 2.6 24-35 mo 1.5 * 1.4 48+ mo .9 .9 Maternal age < 20 1.0 1.2 35-39 191 .9 Birth order 7-8 1.3 1.2 9+ 2.3 ** 2.0 Maternal education none 1.0 KENYA Preceding interval < 24 to 2.1 ** 2.3 24-35 mo 1.2 1.4 48+ 1.1 1.1 Maternal age K 20 .9 1.0 35-39 1.0 6 Oe Birth order 7-8 1.2 1.4 9+ 1.9 ** 2.3 Maternal education none 1.2 * Appendix Table 2. Relative risk of postneonatal mortality Model I Model II Model III Model IV CAMEROON Preceding interval < 24 mo 2.3 ** 2.3 * 24-35 mo 1.4 * 1.4 48+ .9 .8 Maternal age < 20 11 .1 35-39 1,1 1.1 Birth order 7-8 . 1.0 9+ 1.4 * 1.2 Maternal education none 1.8 GHANA Preceding interval < 24 mo 3.2 ** 3.1 24-35 mo 1.5 * 1.5 48+ 1.0 .9 Maternal age < 20 1.0 1.1 35-39 1.0 .9 Birth order 7-8 .9 .8 9+ 1.5 1.4 Maternal education none 1.6 KENYA Preceding interval < 24 mo 2.2 ** 2.1 24-35 mo 1.3 1.3 48+ 1.0 1.0 Maternal age < 20 1.6 ** 1.6 35-39 1.0 .8 Birth order 7-8 1.1 1.2 9+ 1.0 1.1 Maternal education none 1.4 Appendix Table 3. Relative risk of toddler mortality Model I Model II Model III Model IV CAMEROON Preceding interval < 24 mo 1.8 ** 1.8 24-35 mo 1.3 1.3 48+ .6* .6 Maternal age < 20 1.5 ** 1.4 35-39 .7 .8 Birth order 7-8 .5 ** .6 9+ .7 .8 Maternal education none 1.6 * GHANA Preceding interval < 24 mo 2.2 ** 2.2 .24-35 mo 1.6 * 1.6 4.. + .8 .7 Maternal age < 20 .8 .9 35-39 .7 .6 Birth order 7-8 1.2 1.3 9+ .6 .6 Maternal education none 1.6 KENYA Preceding interval < 24 mo 2.3 ** 2.3 * 24-35 mo 1.8 ** 1.8 * 48+ . 9 .8 Maternal age < 20 1.3 1.3 35-39 1.2 1.1 Birth order 7-8 .8 .8 9+ 1.1 1.1 Maternal education none 1.8 Appendix Table 4. Relative risk of child mortality Model I Model II Model III Model IV CAMEROON Preceding interval < 24 mo 1.4 * 1.3 24-35 mo 1.0 1.0 48+ .7 .7 Maternal age < 20 1.0 1.1 35-39 1.1 .8 Birth order 7-8 1.4 * 1.5 9+ 1.6 ** 1.9 Maternal education none 1.7 e GHANA Preceding interval < 24 mo 1.1 1.0 24-35 mo .9 1.0 48+ .4 ** .3 Maternal age < 20 2.1 ** 2.2 * 35-39 1.2 1.0 Birth order 7-8 .8 .9 9+ 1.6 1.7 Maternal education none 4.2 KENYA Preceding interval < 24 mo 1.5 1.4 24-35 mo 1.4 1.4 48+ .7 .7 Maternal age < 20 1.4 * 1.5 35-39 2.0 ** 1.6 e Birth order 7-8 1.4 * 1.3 9+ 1.8 ** 1.6 Maternal education none 1.8 Appendix Table 5. Relative risk of toddler mortality when births are closely spaced, demographic and socioeconomic factors controlled CAMEROON GHANA KENYA Relative risk Preceding int. 1.7 2.1 2.2 Subsequent int. 2.2 2.3 1.6 Combined 3.7 4.8 3.5 Preceding interval <24 months Subsequent interval <18 months Note: These relative risks differ slightly from those shown in other tables because the estimates are based on the mortality rates of those children who have both preceding and succeeding intervals, i.e., it excludes first born and last born children. Subsequent intervals of this very short length characterize 12 percent of these births in Cameroont 6 percent in Ghanar and 16 percent in Kenya. Appendix Table 6. Relative risk of neonatal, postneonatalt and toddler mortalitybymaternalageandbirthorder, unadjusted and adjusted for birth interval and maternal education A: CAMEROON Unadjusteda Adjusted Neonatal Maternal age <20 1.1 1.0 30-34 1.0 1.l 35-39 1.1 1.3 40+ 1.1 1.3 Birth order 4-6 1.0 .9 7-8 1.0 1.2 9+ 1.5 * 1.1 Postneonatal Maternal age <20 1.2 1.1 30-34 .8 .9 35-39 .8 1.1 40+ 1.3 1.3 Birth order 4-6 1.1 1.0 7-8 1.3 * 1.0 9+ 1.6 ** 1.2 Toddler MatGrnal age <20 13 * 1.4 * 30-34 .9 1.0 35'-39 .7 .8 4 + .9 .9 Birth order 4-6 .9 .8 7-8 .7 6 9+ 1.0 .8 a Age controlled for order, order controlled for age Appendix Table 6 (continued). Relative risk of neonatal, post- neonatal and toddler mortality by maternal age and birth order, unadjusted and adjusted for birth interval and maternal education B: GHANA Unadjusteda Adjusted Neonatal Maternal age <20 1.2 1e2 30-34 1.2 1.4 35-39 .6* 9 40+ 1.0 1.6 Birth order 4-6 1.4 ** 1.2 7-8 1.6 * 1.2 9+ 2.5 ** 2.0 ** Postneonatal Maternal age <20 1.4 * 1.1 30-34 .7 * .8 35-39 .8 .9 40+ 1.3 1.4 Birth order 4-6 1.1 1.0 7-8 1.1 .8 9+ 2.0 **1.4 Toddler Maternal age <20 1.4 .9 30-34 .9 .9 35-39 .6 .6 40+ 1.0 1.1 Birth order 4-6 1.3 1.1 7-8 1.6 1.3 9+ .8 .6 Appendix Table 6 (continued). Relative risk of neonatal, post- neonatal and toddler mortality by maternal age and birth order, unadjusted and adjusted for birth interval and maternal 'education C: KENYA Unadjusteda Adjusted Neonatal Maternal age <20 1.1 1.0 30-34 s7** 8* 35-39 .6 6 40+ .8 1.1 Birth order 4-6 1.2 1.2 7-8 1.7 ** .1.4 ** 9+ 2.7 ** 2.3 ** Postneonatal Maternal age (20 1.7 ** 1.6 ** 30-34 1.0 1.0 35-39 .8 .8 40+ 1.2 1.4 * Birth order 4-6 1.1 1.0 7-8 1.3 * 1.2 9+ 1.4 * 1.1 Toddler Maternal age <20 1.5 ** 1.3 30-34 .9 1.0 35-39 1.0 1.1 40+ .6 .8 Birth order 4-6 1.1 1.0 7-8 1.0 .8 9+ 1.4 1.1 * .05 significance ** .01 significance Note: Reference group: maternal age 20-29' birth order 2-3p some maternal education Appendix Table 7. Summary table of infant (1-11 months) and child (12-59 months) survival by preceding birth interval (3 categories),0-15 years before survey Preceding interval (in months) <24 24-47 48+ CAMEROON Infant* Survived 3,449 6,601 1,392 Died 598 564 89 Total 4,047 7,165 l,481 Child Survived 1*981 3,775 827 Died 285 367 42 Total 2,266 4,142 869 GHANA Infant* Survived 1t919 5,426 l, 502 Died 256 323 57 Total 2,175 5,749 l1 559 Child * Survived lr220 3,314 907 Died 100 229 28 Total l1320 3,543 935 KENYA Infant* Survived 6,332 7,896 1,554 Died 792 546 92 Total 7,124 8,442 1,646 Child Survived 3,864 4,759 923 Died 326 323 30 Total 4,190 5, 082 953 * Born at least one year before survey date Born at least five years before survey date Appendix Table 8. Proportion of births, following intervals less than 24 months, by demographic and socioeconomic characteristics CAMEROON GHANA KENYA Total 31.4 22.4 40.6 Maternal age <20 45.4 38.6 52.5 20-29 32.2 21.6 42.2 30-34 28.1 21.8 38.7 35-39 24.8 21.1 35.3 40+ 28.0 18.1 27.6 Birth order 2-3 30.9 20.8 41.7 4-6 31.3 22.7 39e9 7-8 31.0 25.1 40.1 9+ 36.7 28.8 40.5 Cohort* 1 33.6 25.0 44.3 2 32.5 23 .5 41.0 3 29.3 19.9 38.0 Residence Large urban 19.4 50.4 Urban 30.5 24.2 46.8 Rural 31.7 22.6 39.3 Maternal education none 31.5 22.6 39.3 1-3 years 31.0** 24.1 39.4 4-6 years 241.3 39.2 7+ years 20.6 49.5 1 - 10-15 years before survey ** primary+ 2 = 5-10 years before survey 3 < (5 years before survey