METHODS

A representative sample of 14 704 women aged between 18 and 44 years was randomly selected from the telephone directory, stratified by region. If multiple eligible women were living in the household that was contacted, one woman was randomly selected. Some 3162 women refused to participate and 567 were otherwise excluded (unreachable women, incomplete questionnaire). The response rate was 75%. Women who in the past five years had an abortion or whose last pregnancy was unintended were all selected (sampling fraction=100%, n=1034) while a fraction of the other women were randomly selected (sampling fraction =19%, n=1829). Altogether, 2863 women answered the questionnaire.

Sample for this study

Of the 2863 women in the sample, 2235 women had had at least one pregnancy in their lifetime. The analysis was limited to the 722 women whose last pregnancy was unintended, whether this last pregnancy happened in the past five years or before. Pregnancies with outcomes other than birth or induced abortion were excluded (n=77), leaving 645 pregnancies that ended either in live birth or in induced abortion.

Questionnaire

The questionnaire was designed to explore the sociodemographic, economic, and other circumstances around the most recent pregnancy. It was developed on the basis of the results of a qualitative study on unintended pregnancies,¹⁸ and has been tested on a sample of 800 women.¹⁹

Independently of the question about “unintended pregnancy in the past five years” asked at the beginning of the questionnaire, pregnancy intention was asked about for each of the pregnancies: “Did you intend to get pregnant (then)?”, with the response options (1) You didn’t think about it, (2) You don’t remember anymore, (3) Not at all, (4) You intended to get pregnant later, (5) You intended to get pregnant earlier, (6) You intended to get pregnant at that moment. Closer examination of the data suggested that women who responded “did not think about it”, “not at all”, or “later” differed significantly from each other for both sociodemographic and reproductive characteristics. Therefore only women who gave “not at all” responses were treated as having had an unintended pregnancy.

The characteristics that were looked at for the women included age and marital status at the time of the studied pregnancy, education, income, country of birth, religion, financial and material difficulties at the time of pregnancy, and financial solvency without the partner. Factors related to work were work status at the time of the studied pregnancy, type of contract, how well the pregnancy fitted her work situation, how long she had worked before pregnancy, and her interest in her work. Reproductive characteristics examined were number of children before last pregnancy, number of years since previous birth, earlier induced abortions, and use of contraception when getting pregnant.

Characteristics of the partner were age, education, and work situation. Factors related to the relationship included duration of the relationship, situation in the relationship (stable, new, uncertain, no future, about to break up), and agreement with the partner about whether to continue or terminate the pregnancy.

Analysis

To take the sampling design into account in the analysis, each woman was given a sampling weight, equal to the product of the number of eligible women times the inverse of the sampling fraction. Furthermore, to obtain representative estimators for the whole French population, the weight of each woman was modified according to the under-representation (or over-representation) of the categories to which she belonged compared with the structure of the general population.²⁰ The individual weights were adjusted in such a way that the marginal distributions for some important characteristics (age, marital status, occupation, level of education) were identical in the sample (with the final weights) and in the French population. This process is based on the hypothesis that the “missing women” in a category (such as “20–24 years old”, “single”) behave in a way that is closer to the average of respondents of the same category than to the average of the whole sample.

Stratified analysis according to age (<25, 25–34, ⩾35 years) was used in order to study the impact of different factors in different stages of women’s life cycle. The χ² test was used to analyse the relation between the different factors related to decision (abortion compared with continuation of pregnancy) separately within each age group and for the study sample as a whole. Separate multivariate logistic regression models were run for each of the three age groups. All factors were entered simultaneously in the models. Variables selected for inclusion in the models were those statistically significant in the bivariate analysis, but time from previous birth could not be included because of small numbers in some cells. Based on earlier literature and the target of the study, two variables (level of education of the woman and age of the partner) were included although they were not statistically significant. A variable was created to combine women’s work status and how the pregnancy fitted the work situation.

Weighting was used in all analyses because of the complex sampling design (here inequal probability of inclusion) as it influences estimators of both parameters and their variances.²¹ STATA software was used to compute representative estimators for the general population and to perform statistical tests adequately.²²

RESULTS

Nearly half (48%) of the unintended pregnancies ended in birth, 38% in induced abortion and the rest on other pregnancy outcomes.

Bivariate analysis

Several factors were associated with the decision to terminate or continue an unintended pregnancy in the separate bivariate analysis for each variable. Singlehood, higher education of the partner, student status among the women and their partners, unemployment among women, work situation, problems in the relationship, and number of children were all statistically significant factors in the decision to terminate the pregnancy (table 1). Among younger women, unintended pregnancies were more commonly terminated when the educational level of a woman and her partner was high, or either one of them was a student. Pregnancies were more likely to end in abortion if the woman wanted an abortion but her partner did not. Among 25–34 year old women, having an abortion was linked to reproductive characteristics; the proportion of women having an abortion was lowest for women with one child. Furthermore, 9 of 10 pregnancies were terminated among women who had a baby during the previous year.

Among older women (⩾35 years), having an abortion was most clearly related to the woman’s work situation and to the relationship with the partner. Abortion was chosen in 9 of 10 cases when the work situation was not suitable to becoming pregnant. When the relationship had lasted for less than five years or when it was unstable, most of the pregnancies were terminated. In the case of disagreement between the couple, the pregnancy ended in abortion nearly always if the woman wanted an abortion even if her partner did not agree. Abortion was more common among older women who were single, had at least a high school education, and whose partner was highly educated.

Globally, no statistically significant differences were found between women terminating or continuing an unintended pregnancy for the following factors: nationality, religion, income, financial difficulties, financial solvency without the partner, type of work contract, how long she had worked before pregnancy, and interest in her work.

Multivariate analysis

The multivariate models confirmed that factors influencing the decision to have an abortion in unintended pregnancy varied by age. Younger (⩽25 years) women’s abortion decision were mainly related to being a single (table 2). Higher educational level of herself and/or the partner, being a student, or having an unsuitable time to get pregnant because of work situation also increased the likelihood of abortion among youngest women.

Among 25–34 year old women, the abortion was more likely to occur when the woman had already two children; the odds ratios were 0.62 (95% CI 0.25 to 1.56) if the woman had one child and 4.4 (95%CI 1.65 to 11.77) for women who already had two or more children, in comparison to women with no children. Similarly, unstable relationship and partner’s higher education increased the likelihood of abortion.

Among older women (⩾35 years), unstable relationship also increased the likelihood of abortion as well as unsuitable work situation, higher education, and being single. The confidence intervals of single women and women with unstable relationships were, however, wide because of the small numbers of older women in these groups.

The results were not affected by period effects and they remained the same when “years from last unintended pregnancy” variable was added into the models.

DISCUSSION

The impact of socioeconomic characteristics, work, reproductive history, and factors related to the relationship on a woman’s abortion decision varies according to the different phases of the life cycle of women. For younger women the decision to have an abortion is mainly influenced by a wish to continue her studies or to start work. In other research, women have stated unfinished studies to be a reason for abortion^8,9 although it is seldom given as the most important reason.⁶ The situation of women in the 25–34 year age group is different as this age range encompasses the maternity stage in women’s life cycle. In 2001, the average age of French women at the time of giving birth was 29.4 years.²³ Statistics show that family size in France has been very stable during the past 25 years, and today the average number of children is still around two.²⁴ Those women among 25–34 year olds who had only one child had the lowest likelihood of choosing abortion. Many women in this age group, however, already had the desired number of children and wanted to end their childbearing and therefore chose to have an abortion.

An important finding is the strong influence of the work situation. Although some studies have suggested that work is an important reason for having an abortion^8,9 this has not been sufficiently studied. Kero et al showed that work, studies, and existing children were together the most important reason for an abortion among Swedish women, but this study did not examine these factors separately.⁹ Our results show that when a pregnancy is perceived as not suitable for the work situation, it is an important factor that influences the decision to have an abortion, especially for older women.

Responses to direct questions about financial and material circumstances showed that these factors were not associated with the abortion decision. But among unemployed women, 7 of 10 pregnancies were terminated. Single parenthood seems not to be a desirable option, as earlier studies have shown.^6,25,26

Most studies have not paid attention to the characteristics of the partner. Among younger women, high education level of the partner was even more strongly associated with abortion than was education level of the woman. A study among teenage women also showed a positive association between the partner’s education and the likelihood of abortion.¹⁵ It may be that higher education levels of the woman and her partner are markers of social class, and that it is more acceptable for them to “plan” their family and terminate a pregnancy when the time is not right to have children and the priority is to establish their careers. On the other hand, for young women with no plans for a career requiring an education, having a child is a way of gaining a social identity.

Earlier studies have shown how a partner’s attitudes and support greatly influence the decision to continue or interrupt the pregnancy.^8,12–14 Our results are consistent with the literature¹³: if the woman felt that the relationship did not have a future the pregnancy was more likely to be terminated. But pregnancies were also likely to be terminated when the relationship was just beginning. Half of the pregnancies that occurred in uncertain or new relationships were terminated.

There was more agreement between the couple when an unintended pregnancy was continued than when it ended in abortion. When the woman stated that she wanted an abortion but her partner did not, the pregnancy was terminated in most cases. The group of 25 to 34 year old women were more often influenced by their partner’s wishes about terminating or continuing the pregnancy than were younger or older women. According to a Swedish study between 6% and 10% of women were influenced by someone (mostly by their partners) when deciding about abortion.²⁷ Another Swedish study gave some evidence that counselling before abortion can lead women in all age groups to continue their pregnancies when the initial reason for seeking an abortion was that the partner did not want a baby.⁶

Because of the retrospective nature of this study, recall biases are a possible issue. Our definition of unintended pregnancy was strict compared with other studies^2,28 and factors related to the decision to have an abortion may be different for mistimed pregnancies. Our decision to exclude women with mistimed pregnancies was justified as the sociodemographic and reproductive characteristics of the former were closer to those of women with intended than unintended pregnancies (unpublished data). Another limitation is under-reporting of abortions, which can vary in surveys between 29% and 59%.²⁹ Evidence from the pilot survey¹⁹ suggested that under-reporting of abortions with this method is around 40%, which may affect the conclusions drawn about the circumstances around abortion. Under reporting has been found to be significantly more common among young and unmarried women, where it can be around 60%.^29,30

This population based study shows that there are several factors that influence the decision to have an abortion in case of unintended pregnancy, and that the impact of these factors varies according to age. It is evident that in addition to all these socioeconomic, reproductive, and other personal factors, the decision to continue or terminate pregnancy is influenced by the sociocultural context of the woman or couple. Studies trying to integrate macro contexts in analyses are very rare.³¹ In our study the role of social norms related to maternity appear clearly for the 25 to 34 year-old women, the group in which abortion was chosen less often, whose abortion decisions were mainly influenced either by the fact that they were in the most reproductive stage of their life cycle or by the fact that they already had the wanted and acceptable family size, two children.²⁴ Younger women are also influenced by social norms at a stage where studies and professional issues are prioritised over parental life for those who wish to build careers. Whether there is a generation effect present that influences women’s decision on abortion at different ages remains unanswered. More will be known when the youngest cohort now emphasising education and career come to the end of their reproductive life cycle.