Unintended pregnancy is an important public health concern in both the developing and developed world because of its association with adverse social and health outcomes, for both mothers and children. These include the higher likelihood of unsafe abortion, of late initiation and underutilization of prenatal care, and of low birth weight.1 The level of unintended pregnancy also can serve as an indicator of the state of women's reproductive health, and of the degree of autonomy women have in determining whether and when to bear children. It is important to identify factors associated with unintended pregnancy, to enable policymakers and program planners to design legislation and services specifically for the women who are most likely to experience this problem.
This article examines the determinants of unintended pregnancy in the Andean country of Ecuador, where fertility has declined markedly over the past two decades. For example, the total fertility rate fell steadily from 5.4 lifetime births per woman in 1975 to 3.8 in 1984-1989, and stood at 3.6 births as of 1989-1994.2 To a large degree, this decrease in fertility reflects the increased availability and use of modern contraceptive methods: Prevalence among married women of reproductive age grew from 34% in 1979 to 53% in 1989, and reached 57% in 1994.3
Despite the decrease in overall fertility and the increase in contraceptive prevalence, however, many women in Ecuador, where induced abortion is illegal, still experience unintended pregnancies. In fact, while the level of unintended pregnancy has fallen in every other Latin American country but Paraguay, it has risen markedly in Ecuador in recent years.4
For example, in Ecuador's 1994 Encuesta Demográfica y de Salud Materna e Infantil, or ENDEMAIN (Demographic and Maternal-Child Health Survey), 36% of women in union reported that their most recent pregnancy (in the five years before the survey) was unintended5—19% who said their pregnancy had been unwanted (because they had not wanted any more children at the time of the pregnancy) and 16% who said the pregnancy was mistimed (because it occurred earlier than desired). Comparable proportions in the 1989 ENDEMAIN were 7% unwanted and 6% mistimed, respectively. 6
Background
While research has identified a number of predictors of unwanted and mistimed pregnancy, the majority of these multivariate analyses have been conducted in the United States, rather than in Latin America. Moreover, in both regions, unwanted pregnancy has been examined more often than has mistimed pregnancy. The available information on unintended pregnancy in Latin America is based largely on data collected in the Demographic and Health Surveys (DHS) series. While these data are nationally representative, the survey reports contain bivariate data only.
Nonetheless, these surveys collected data on pregnancy intention status, regardless of outcome, and many of the characteristics identified in the bivariate data as determinants of unwanted pregnancy are consistent with results of multivariate analyses conducted with U.S. data. These analyses have shown that U.S. women who experienced an unwanted birth were more likely than those who did not to be unmarried, to be older (in their 30s), to have already had several children or to have not graduated from high school.7 The association between poverty and unwanted childbearing among U.S. women is less clear, however, as conflicting studies have indicated both a statistically significant association and the lack of one.8
Bivariate data from DHS surveys conducted in four South American countries other than Ecuador (Bolivia, Colombia, Paraguay and Peru) indicate that the likelihood of unwantedness increases with a woman's age and parity. 9 Moreover, rural woman are more likely than urban women to have more children than they consider ideal. In three of these countries (Bolivia, Colombia and Paraguay), the less educated a woman is, the greater her likelihood of having more children than she considers ideal. In Paraguay, women with low levels of education and of low socioeconomic status are more likely to have unwanted pregnancies than are better educated, more affluent women.
Moreover, one multinational multivariate analysis, which included two Latin American countries, found that in Colombia and Peru, unwanted pregnancy decreased with years of education, but that no significant relationship emerged between unwanted pregnancy and socioeconomic status (as measured by a composite index of household characteristics and possessions).10 In a study conducted with Chilean data, women aged 15-24 living in households of low socioeconomic status (as measured by the father's level of education) were more likely than the daughters of better educated men to experience unintended (unwanted or mistimed) pregnancies.11
Research suggests that a slightly different set of factors is associated with mistimed rather than unwanted births and pregnancies. A literature search uncovered only one published multivariate analysis of the determinants of mistimed (as opposed to planned) births in the United States. That study showed that teenage mothers were far more likely than older women to have a mistimed rather than a planned pregnancy. Mistimed pregnancy was also significantly more likely to occur among unmarried than among married women. Poverty also affected the likelihood of mistimed pregnancy; the odds that a birth was mistimed (versus planned) decreased as poverty status improved, but only among married women.12
Turning to Latin American data, the DHS bivariate data for Bolivia, Paraguay and Peru indicate that the proportions of pregnancies that are mistimed are highest among second pregnancies (since first births tend to be planned) and among younger women (who likely desire a child at some time in the future). In Peru, the proportion of pregnancies that are mistimed peaks among 15-19-year-olds, and in Bolivia and Paraguay, that proportion is highest among women in their 20s.13 The multivariate analysis conducted with Peruvian and Colombian data mentioned earlier found that the likelihood of mistimed pregnancy in these countries (compared with both planned and unwanted pregnancies) increased with a woman's years of education, but that there was no significant relationship between mistimed pregnancy and socioeconomic status.14
In general, the overriding factor affecting fertility, and thus the likelihood of unintended pregnancy, is the practice of family planning.15 It follows that knowledge of methods and access to services are important determinants of effective use. If a woman is unaware that she can regulate her fertility, does not know how to do so or is unable to obtain services, she is largely incapable of avoiding unwanted and mistimed pregnancy. However, knowledge about and access to contraception by no means guarantee its use or efficacy.16
In this article, which presents results of a multivariate analysis of the determinants of unwanted and mistimed pregnancy in Ecuador, all pregnancies, regardless of outcome, are included, unlike previous studies that examined the intention status of pregnancies resulting in live births only. Moreover, the analysis detailed here jointly examines the factors associated with all three categories of pregnancy intention status—planned, mistimed and unwanted. This approach differs from those of other studies that considered only two categories or that divided the three into separate groupings for purposes of analysis, blurring the distinctions between them.
Methodology
Survey Data
Data are drawn from a nationally representative sample of women of all marital statuses—the 1994 ENDEMAIN, fielded from May through August 1994. For this survey, which used a three-stage cluster design sample, 20,000 households were randomly selected for interview, and 72% (14,084) had at least one woman of reproductive age (15-49). Female interviewers completed 13,582 personal interviews, achieving a 96% coverage rate.
At the regional level, the survey was representative of the country's two major regions, the Sierra (highlands) and the Costa (coast), where approximately 97% of the population lives.* The Sierra region, while comprising less than one-fourth of the country's land mass, is home to more than one-half the total population. Quito, Ecuador's capital, is located in this region, which is dominated by the Andean mountains and has a largely cold and arid climate. Guayaquil, the country's largest city, is in the Costa region, which enjoys a year-round tropical climate.
This article focuses on the 33% of the ENDEMAIN sample who were pregnant between January 1992 and the date of their interview—a total of 4,534 women. This relatively recent time frame was chosen to limit problems of recall and rationalization associated with retrospective data.17 To control for multiple pregnancy outcomes to the same woman, only the most recent pregnancy is considered in the analysis. Seventy-eight percent of the women in the study had a pregnancy ending in a live birth, 15% were still pregnant at the time of the interview and 7% had lost the pregnancy, either through spontaneous or induced abortion or through stillbirth.
The women in the study sample are representative of Ecuadoran women who experienced a recent pregnancy; they differ slightly from all women who were interviewed, however, in age, parity and area of residence. Compared with women who had not had a recent pregnancy, those who had tended to be slightly younger (mean age 27.1 vs. 30.2) and to have had fewer children, on average (3.2 vs. 3.5); women with a recent pregnancy were more likely than the others to live in rural areas (53% vs. 46%).
Analytic Approach
Multinomial logistic regression was used to assess the factors associated with the odds of a pregnancy being unwanted, mistimed or planned. It is unclear whether pregnancy intention should be measured on a continuum—i.e., whether such categories as "unwanted," "mistimed" and "planned" represent sequential levels of intention status (so that "wantedness" is measured on a continuous scale) or are discrete, unordered categories. Given this uncertainty, responses on pregnancy intention were treated as distinct motivations, and the data were analyzed using multinomial logistic regression (as opposed to ordinal logit).18 This approach was also supported by the bivariate analysis results, which suggested that many of the factors associated with mistimed pregnancy in the study population were distinct from those associated with unwanted pregnancy. In bivariate analyses, chi-square tests were used to test the significance of differences between nominal variables, and one-way analysis of variance was used for continuous variables.
The SAS software package was used for univariate and bivariate analyses, and Stata was used for the multivariate analyses. In both analyses, a weighting factor was applied to all observations to compensate for the ENDEMAIN sampling design and for differing regional response rates. (The weighting factor was inversely proportional to a woman's probability of selection, which was dependent on the number of women in her household and on the population in her region, province and census sector.) Thus, all results presented are weighted.
In the multivariate analyses, no corrections were made for the multistage cluster sample design because of statistical software limitations. Thus, while the coefficients in the multinomial regression models remain unbiased, the standard errors and associated z-statistics may be biased. This may lead to incorrect inferences about statistical significance. The likelihood of this happening, however, is mitigated by the inclusion of sample selection design variables, such as area and region of residence. When bivariate analyses were run unweighted, the results were only marginally different from the weighted ones. The weighted multivariate analyses are likewise unlikely to differ significantly from those using unweighted data.
Dependent Variable
The main outcome of the analysis, intention status, is a retrospective measure of a woman's reproductive intentions and was determined by asking respondents to recall their feelings at the time they became pregnant. If a woman was pregnant at the time of the survey or if her pregnancy had ended in a spontaneous abortion, induced abortion or stillbirth, the interviewer asked, "Did you want to become pregnant at that time, did you want to wait more time, or did you not want this pregnancy?" If a woman's most recent pregnancy ended in a live birth, the question was worded, "At the time you became pregnant with [child's name], did you want the child at that time [planned], did you want to wait more time [mistimed], or did you not want more children [unwanted]?" If a woman reported that her pregnancy occurred when it was desired or later (rather than sooner), it was categorized as a planned pregnancy.
Such a retrospective measure can be biased, since women may not remember how they felt at the time of conception or may not wish to report a conception as unwanted or mistimed, once the child born from that pregnancy has become a loved family member. In addition, women probably underreport unintended pregnancies that do not end in a live birth (i.e., in induced abortion or some other outcome). In this analysis, as in other studies,19 a small proportion of women are assumed to incorrectly report their reproductive intentions, and thus to assert that an unintended pregnancy had been "intended." This measurement error is likely to bias the analysis conservatively, resulting in an underestimate of the proportion of unintended pregnancies.
Explanatory Variables
The explanatory variables considered in the multivariate analysis of the predictors of unintended pregnancy included the woman's area and region of residence (within the broad Sierra and Costa regions, the metropolitan areas of Quito and Guayaquil, other urban areas and rural areas); socioeconomic status (high, middle or low, based on an index of up to 10 household amenities); parity (the number of live births before the respondent's most recent or current pregnancy); marital status (a dichotomous variable indicating whether a woman was in a union at the time her pregnancy ended); age-group (15-19, 20-29 and 30-49); education (a dichotomous variable measuring whether a woman had completed primary school); knowledge of family planning (number of methods with which a woman was familiar); and use of a modern method before the pregnancy under study.
Results
Background Characteristics
Among Ecuadoran women who had recently been pregnant, 25% lived in a major metropolitan area (Guayaquil or Quito), 28% resided in other urban areas and 47% lived in rural areas (Table 1). The sample was evenly divided between the two major regions. While slightly more than one-half (51%) lived in middle-income households, 20% were from lower income households and 29% from high-income ones. Fifty-three percent were in their 20s, 32% were aged 30-49 and 15% were aged 15-19. A large majority (87%) were married or in a consensual union, and most (78%) had completed primary school. Women had had, on average, 2.2 births prior to the pregnancy under study (not shown).
About one-third (35%) of the sample had used a modern method prior to their most recent pregnancy (Table 1). Moreover, women were familiar with an average of 4.4 methods (not shown).
Pregnancy Intention Status
Approximately 21% of the women reported that their most recent pregnancy had been unwanted, 18% characterized the pregnancy as mistimed and 62% classified it as planned (Table 2). Pregnancy intention varied significantly by all seven background characteristics considered; moreover, the women most likely to have had an unwanted pregnancy differed from those most likely to have experienced a mistimed pregnancy.
By area of residence, rural women were more likely to have an unwanted pregnancy (24%) than were either metropolitan women (20%) or those living in other urban areas (16%). Mistimed pregnancy, on the other hand, was relatively more common among metropolitan women (21%) than among nonmetropolitan urban or rural women (16-18%). Women living in the highlands were more likely than those living on the coast to report that their most recent pregnancy had been unwanted.
By socioeconomic status, women living in relatively poor households were most likely to report their pregnancy as unwanted (26%), while those in the highest-income households were least likely to do so (14%). Women in middle-income households were more likely than those of other socioeconomic backgrounds, however, to classify their most recent pregnancy as mistimed.
In the bivariate analysis, age and parity were both significantly associated with pregnancy intention. The oldest women (those aged 30-49) were more likely than younger women to say their pregnancy had been unwanted (33% vs. 10-16%), but women in their 20s were more likely than both younger and older women to have classified their pregnancy as mistimed (22% vs. 12-17%).
High parity and unwanted pregnancy were clearly linked. Women with unwanted pregnancies had had an average of 3.7 previous births, while women with mistimed or planned pregnancies had had 1.9 and 1.7 previous births, respectively (not shown).
Table 2 also shows that single women were more likely than women in union to say their pregnancy had been unwanted (25% vs. 20%), while women in union were more likely than single women to have experienced a mistimed pregnancy (19% vs. 12%). Women with no formal education or who had not completed primary school were more likely to have had an unwanted pregnancy than women with a primary schooling. However, mistimed pregnancy was more common among women with a primary schooling than among less-educated women.
Unwanted pregnancy was also linked with knowledge of family planning: Women with planned or mistimed pregnancies knew of slightly more modern methods than did women with unwanted pregnancies (means of 4.5 methods vs. 4.1 methods). Both unwanted and mistimed pregnancy were more common among women who had used a modern method of family planning (25% and 20%, respectively) than among those who had not used a method before their most recent pregnancy (18% and 17%, respectively).
Multivariate Analysis
The results of the multinomial logistic regression indicate that once all explanatory variables had been controlled for, both area and region of residence were strongly associated with pregnancy intention status (Table 3). In contrast to the pattern suggested in the bivariate analysis, rural women were 37% less likely than women from Quito or Guayaquil to have had an unwanted pregnancy, relative to a planned one. Rural women were 29% less likely than metropolitan women to have classified their most recent pregnancy as mistimed rather than as planned. Moreover, women from nonmetropolitan urban areas were 30% less likely than residents of Quito or Guayaquil to have had an unwanted pregnancy, and they were 23% less likely to have a mistimed pregnancy, relative to a planned one.
Women from the Sierra region were at a particularly high risk of unintended pregnancy. They were more than twice as likely as women from the coast to have viewed their most recent pregnancy as unwanted (odds ratio of 2.2), and they were 1.8 times as likely to have had a mistimed pregnancy, relative to a planned one.
Women living in a high-income household were about one-third less likely than those living in a middle-income household to have experienced either an unwanted or a mistimed pregnancy, compared with a planned one. However, those residing in a low-income household were no more or less likely than middle-class women to have done so.
Parity was significantly associated with both unwanted and mistimed pregnancy. For each previous child that a woman had had, the odds that her most recent pregnancy was unwanted rather than planned increased by a factor of 1.4, and the odds that the pregnancy was mistimed rather than planned increased by a factor of 1.2. Further, among women who had had an unintended pregnancy, each additional previous child increased the odds that the most recent pregnancy had been unwanted rather than mistimed, by a factor of 1.2.
Older Ecuadoran women were not at any overall higher risk of unintended pregnancy than were younger women. Although older women in general were less likely to have been pregnant than were younger women, an older woman who became pregnant was more likely to have planned to do so. For example, women in their 30s and 40s were 56% less likely than those in their 20s to have had a mistimed pregnancy rather than a planned pregnancy, and were 21% less likely to have had an unwanted pregnancy. Ecuadoran teenagers, however, were no more likely than women in their 20s to say that their current or most recent pregnancy was unintended.
Among older women who experienced an unintended pregnancy, the pregnancy was more likely to be unwanted than to be mistimed: Women in their 30s and 40s were 80% more likely than those in their 20s to have had an unwanted pregnancy rather than a mistimed pregnancy.
Women in union were 52% more likely than single women to classify their most recent pregnancy as mistimed rather than as planned, but they were 56% less likely than single women to have had an unwanted rather than a planned pregnancy. Among women with unintended pregnancies, women in union were 71% less likely than single women to have considered their most recent pregnancy to be unwanted rather than mistimed.
Education increased the odds that a pregnancy was mistimed rather than planned. Women who had completed primary school were 57% more likely than those who had not to have experienced a mistimed rather than a planned pregnancy. Educational attainment, however, had no significant effect on the likelihood of an unwanted rather than a planned pregnancy. Among women who had had an unintended pregnancy, those who had finished primary school were 36% less likely than those who had not to have had an unwanted rather than a mistimed pregnancy.
Finally, women who had used a modern contraceptive were more likely than those who had not to have classified their most recent pregnancy as unwanted or mistimed. Prior contraceptive users who had used a modern method were 59% more likely than nonusers to report that their most recent pregnancy had been unwanted rather than planned, and they were 30% more likely to say it had been mistimed rather than planned. The number of family planning methods known by a woman was not significantly associated with pregnancy intention status, however.
Discussion and Conclusions
These results indicate that particular groups of Ecuadoran women are at significantly elevated risk of unintended pregnancy and thus would benefit from quality family planning services that are tailored to their needs. For example, the significant differences in intendedness that emerged by region and area of residence suggest that family planning services need to be expanded or improved in metropolitan areas and in the Sierra region. While bivariate data from DHS surveys conducted in other South American countries suggest that rural women are more likely than urban women to experience unintended pregnancies, this analysis found that residence in rural and nonmetropolitan urban areas independently lowered the likelihood of both unwanted and mistimed pregnancy relative to residence in Ecuador's two largest cities, Quito and Guayaquil.
The recent heavy migration from rural areas to these large metropolitan areas might have overburdened family planning services there. In addition, rural women's ideal family sizes tend to shift downward when they move to large cities, where living space is more limited and the cost of living is higher. Even if family planning services are available, women who have recently migrated may lack the knowledge and skills needed to achieve their modified reproductive preferences, or they may consider a formerly desired pregnancy to be unwanted or mistimed.
Unwanted pregnancy was far more likely to occur among women from the Sierra region than among those from the coast. Such regional disparities may be due to cultural factors (i.e., that the bulk of the country's indigenous Quichua population lives in the Sierra region), as well as to differences in the availability and quality of family planning services. Unfortunately, the ENDEMAIN did not collect information on these factors, other than to ask women who had used family planning how long they had traveled to obtain their method.
Not unexpectedly, women from high-income households were less likely than those from middle-income backgrounds to have had an unwanted or mistimed pregnancy. However, somewhat surprisingly, women who lived in the poorest households were no more likely than those from a middle-income background to have had an unwanted or mistimed pregnancy, compared with a planned one. This suggests that middle-income and low-income women are similar to one another in not being able to afford the family planning services that are probably more accessible to high-income women.
Both parity and age were associated with the likelihood of unwanted and mistimed pregnancy. The more children a woman already had, the more likely she was to report that her current pregnancy was unwanted or mistimed, regardless of her age or other factors; this corroborates findings from previous research. Women who have had many children may differ in meaningful ways from women who have not; high-parity women, for example, may have limited access to services or may experience particular difficulty in practicing contraception.
Ecuadoran teenagers, unlike U.S. teenagers, were no more likely than women in their 20s to classify their most recent pregnancy as mistimed. However, the finding that women in their 30s and 40s were less likely than those in their 20s to have had an unwanted or a mistimed pregnancy rather than a planned one contrasts with data from U.S. studies and DHS surveys conducted in other South American countries.
This finding may reflect more deferred childbearing among Ecuadoran women, or it may indicate that older women want larger families than younger women do. Indeed, the 1994 ENDEMAIN indicated that women aged 30-49 considered a family of 3.3 children to be ideal, compared with one of 2.6 children among women in their 20s and one of 2.2 among 15-19-year-olds.§ Nonetheless, unwanted pregnancy remains a problem for some older women in Ecuador; when women in their 30s and 40s did have an unintended pregnancy, it was usually unwanted rather than mistimed.
The analysis also suggests that single women and women in union have different family planning needs. Single women (13% of the sample) were much more likely than women in union to say that their most recent pregnancy had been unwanted rather than planned or mistimed; this finding is similar to conclusions from U.S. research. Women in union, on the other hand, were more likely than single women to have had a mistimed pregnancy rather than a planned one. These findings suggest that single women need assistance obtaining appropriate methods for limiting, more than for spacing, their births. This finding is somewhat puzzling and probably reflects the fact that more than one-half (52%) of women who were not in a union were separated, divorced or widowed. These women may have already achieved their desired family size.
Contrary to expectation, women who had completed primary school were more likely than those who had not to have had a mistimed rather than a planned pregnancy. Moreover, among women with unintended pregnancies, those pregnancies were more likely to be mistimed among women who completed primary school, and they were more likely to be unwanted among women with no formal schooling or who did not complete primary school. Poorly educated women might be less likely than better educated women to have mistimed pregnancies because they might have more modest expectations of their ability to control the timing of their pregnancies.
Experience using a modern family planning method, but not knowledge of contraception in general, was independently associated with the likelihood of unintended pregnancy in this analysis. Past users of a modern method were more likely than nonusers to say their pregnancy had been unwanted or mistimed than to report that it was planned. Users of a method might have higher expectations about limiting or spacing their pregnancies, and thus be more likely to view a pregnancy as unplanned.
The lack of an association between contraceptive knowledge and pregnancy intention status may indicate that awareness does not always indicate an ability to obtain methods or to use them correctly and effectively. More refined measures of both family planning knowledge and experience may shed more light on these relationships.
Thus, the results point to the need for further research in several areas. Current measures of pregnancy intention status are probably inadequate and need to be refined to be more relevant to different social and cultural groups. In addition, the biases inherent in reporting retrospective attitudes toward pregnancy need to be addressed; feelings about pregnancy may change throughout the gestation, as well as after the birth. Further research is also needed to identify the cultural and psychosocial factors that differentiate women at high risk of unintended pregnancy from those who are able to plan their pregnancies. More information is needed on the role of family planning methods and services in preventing unintended pregnancy.
Program planners and policymakers in Ecuador would benefit from knowing what proportion of unintended pregnancies are caused by nonuse of contraception and what proportion stem from contraceptive failure or inconsistent or inaccurate use. Unfortunately, the data source for this analysis, the 1994 ENDEMAIN, did not ask women such questions.
The role of the quality of care in improving women's ability to achieve their reproductive goals and reduce their number of unintended pregnancies should also be given special attention, since disrespectful treatment, inadequate information or limited method choice might lead women to underutilize services. Indeed, a recent study carried out in Peru found that quality of care significantly affected women's ability to avoid unintended pregnancy.20
Finally, it should be noted that some women with unintended pregnancies—those who respond to mistimed or unwanted ones by resorting to induced abortion—probably did not report such pregnancies, and thus were not reflected in this study. Qualitative research methods, such as in-depth interviews, may prove more effective than survey research in exploring the characteristics of these women, particularly in societies where abortion is illegal or highly restricted, as well as where it is allowed but socially condemned.