This paper is among the first to evaluate two widely used sets of household survey questions regarding pregnancy intention and to report on associations of these with stillbirth in any LMIC setting, and neonatal death in sub-Saharan Africa. Since survey data are crucial for monitoring maternal and newborn health for more than two-thirds of the world’s births, this paper answers an important question regarding assessment of pregnancy intention in LMICs.
Completeness and data quality of questions in measuring pregnancy intention
In our comparison of completeness and data quality, we found challenges regarding desired-versus-actual family size questions. A total of 12.1% of the responses were excluded due to ‘don’t know’/undecided responses. Desired family size is an abstract concept and in high-fertility populations, as in our study sites, there may be an inability or unwillingness to provide a numerical answer for desired number of children if reproductive choice and agency are not yet fully accepted [28]. However, analysis of the DHS data from 32 countries showed that women’s provision of non-numeric responses to the desired family size question declined over a 19-year period (1993–2011) [28]; additionally, this study reported that as fertility rates declined, the proportion of women giving non-numeric responses also decreased. Knowledge about, and use of, contraception as well as level of education were inversely related to providing non-numeric responses where a numerical response is required [28].
Where numeric responses are provided, it is possible that women may adjust their desired number of children to match their actual number of children. Though assessment of desired-versus-actual family size is widely used by demographers to calculate aggregate wanted fertility rates [29], its utility in identifying individual undesired pregnancies was weak in our study population, and there were inconsistencies between assessment of desired-versus-actual family size and future childbearing preferences. Reported desire for another child is generally acknowledged to be the most robust assessment of fertility preferences [30]. Within our study, nearly half (45%) of women who had achieved or exceeded their desired family size at the time of conception of their most recent pregnancy nevertheless reported wanting another child (Table 4D), indicating a mismatch between the sets of questions. Similarly, we found few differences in perceived importance of avoiding or delaying the pregnancy, in contraceptive use, or in consideration of termination, between women with calculated desired and undesired pregnancies, though this lack of difference may be partly because the questions were only asked for women who stated that their most recent pregnancy outcomes were unintended (Table 4B). Lastly, the percentage of respondents who reported desired family sizes of zero was high in in Dabat (28%) (Fig. 3), which suggests that the survey question may not have been as well understood as intended. Given these points, further research should consider the challenges of asking women to provide numeric responses regarding desired family size.
The pregnancy-specific intention questions also presented some challenges. Though there was higher agreement with future childbearing desires (Table 4D), ultimately women may be reluctant to report a pregnancy as unwanted. Only 4.1% of pregnancies were classified as unwanted, which is low in view of the appreciable proportions of women who reported wanting no more children in the future and the moderate level of contraceptive use in the study sites. In societies where childbearing is highly valued, however, this reluctance may be unsurprising and the high value traditionally placed on children in many sub-Saharan African settings may act to outweigh or offset any previous intentions to avoid or delay births [8, 10]. Longitudinal studies in Senegal, Nigeria, Malawi and Kenya found that only small proportions of women who wanted no more children at baseline retrospectively classified a subsequent birth as unwanted [31].
In our study, it was more common for pregnancies to be classified as mistimed than unwanted, and 15% overall were reported as mistimed. This tendency appears strongest in Bandim where 24% of pregnancy events were reported as mistimed compared with just 2.7% as unwanted (Table 3). Though mistimed pregnancies were less likely to be associated with contraceptive use than unwanted pregnancies, this difference disappears when looking at perceived importance of delaying or avoiding a pregnancy or evoking thoughts of termination (Table 4B). The proportions of women who attached great importance to the desire to delay or avoid were almost identical for those with mistimed (82.7%) and unwanted (84.8%) pregnancies and similar for consideration of termination (19.9% and 19.4%). Our results are consistent with a large body of evidence that spacing of children is a crucial consideration intricately linked to many aspects of social life in most of sub-Saharan Africa, and short spacing between births tends to invite social criticism [32]. More recently, it has been shown that postponement of births, to delay pregnancy for an indefinite time until conditions are conducive, is an important element of reproductive culture in Africa [33]. Under this perspective, the distinction between mistimed and unwanted pregnancies become blurred.
Pregnancy intention is such a complex concept that major advances in its measurement and understanding of its implications for health and welfare are unlikely to be achieved without the application of prospective survey designs and multi-item scales, such as the London Measure of Unplanned Pregnancy (LMUP), a six-item scale which measures the degree of pregnancy intention and which has been validated in a number of low-income settings [34,35,36,37]. Two severe barriers stand in the way of progress. First, evidence on pregnancy intention will continue to come primarily from cross-sectional surveys because large, representative prospective studies are so expensive and time-consuming. Second, these surveys are likely to be multi-purpose, such as the DHS, with many competing interests and questions, making it improbable that space will be found for more ambitious pregnancy intention measures such as LMUP.
The three additional questions added in this study had some value. Perceived importance of delay/avoidance may be the most promising question in terms of level of agreement (only less than 6% of women with ‘unwanted’ and mistimed pregnancies did not perceive delay or avoidance as important). Contraceptive use may likely be influenced by accessibility issues and consideration of termination by socio-cultural factors, personal values and reporting bias [38]. Given this, these additional questions are not the only possible succinct additions to the standard DHS approach. Our results from the additional questions demonstrate that lack of control of pregnancy timing and spacing is not a trivial matter for women and that pregnancy timing is an equal concern to limitation. The results from the question on contraception, which is a component of LMUP, revealed the substantial gap between desires and behaviour, and has obvious programme implications.
Association between unintended/unwanted/undesired pregnancies and maternal health care and adverse pregnancy outcomes
We showed an association between pregnancy intention and timing and frequency of ANC visits. Women who reported their pregnancy as intended had more frequent and earlier ANC visits. This is consistent with previous studies [39,40,41,42] and likely indicates positive health-seeking behaviour amongst mothers who had intended to become pregnant. Using the pregnancy-specific intention questions, we found some evidence of an association between pregnancy intention and stillbirths and neonatal deaths, with women with unintended births less likely to report these outcomes. No such association was found when using the desired-versus-actual family size assessment. Previous evidence on this topic is mixed. In line with our findings, Smith-Greenaway et al. found that pregnancies resulting in neonatal death were less likely to be reported as unintended, in their study of DHS data from 31 sub-Saharan African countries [43]. Longitudinal studies in Ghana and Malawi found no association between pregnancy wantedness and child survival [8, 10], while Hall et al. found some evidence of reduced risk of stillbirth for intended pregnancies, but no association with neonatal mortality, miscarriage and low birthweight nor with a composite measure comprising of all four outcomes [17]. However, Singh and Chalansani identified increased risk of neonatal mortality amongst unwanted pregnancies in Bangladesh and India [13, 44], while in their study of pregnancy intendedness in Ethiopia, Assefa et al. found increased odds of pregnancy loss, defined as miscarriage, induced abortion and stillbirth, amongst unintended pregnancies [16]. Other studies have also reported increased likelihood of other outcomes including premature rupture of membranes, preterm delivery and poor child outcomes, for unwanted pregnancies [45, 46].
Underreporting of neonatal deaths and stillbirths may have influenced our results to some extent. As such, generalisability of findings to other sites may be diminished [3]. As discussed, women may be reluctant to report a child, particularly one who has died, as unwanted or undesired, and may revise their intention or desired family size following conception or birth. Additionally, persistent stigma associated with neonatal deaths and, in particular, stillbirths across many LMIC and HIC settings can undermine collection and analysis of survey data [47]. Detailed qualitative analysis of barriers to reporting stillbirths and neonatal deaths within household surveys are reported elsewhere in this series [48], and we highlight this here again as an important gap in data collection for maternal and newborn health.
Strengths and limitations
This paper reports on pregnancy intention and related outcomes from five HDSS sites in sub-Saharan Africa and Asia, utilising data on almost 15,000 stillbirths, neonatal deaths and live births. This research provides a valuable contribution to assessing pregnancy intention in LMICs, including an in-depth analysis of utility for two commonly used methods of assessing pregnancy intention in surveys plus additional questions. We have also reported on associations between pregnancy intention, health care utilisation and adverse pregnancy outcomes, with important programmatic consequences. Through our analysis of pregnancy intention and stillbirths, which has received less attention in previous studies, this paper provides a valuable contribution to stillbirth research.
Limitations include a cross-sectional study design, retrospective data collection, challenges with representing a nuanced understanding of pregnancy intention, restrictions on some analysis of stillbirths and neonatal deaths, and a high proportion of missing birthweight data. Firstly, and importantly, our cross-sectional survey design, as in the DHS, makes it difficult to establish pregnancy intentions prior to birth and limits assessing of causal association between pre-birth intention and stillbirth and neonatal outcomes. It is widely reported that retrospectively assessing pregnancy intention may underestimate unintended pregnancy due to reluctance to report a child as unwanted or mistimed following pregnancy and birth [15, 43, 49,50,51,52]. In addition, some women in this study were reporting on intention for pregnancies conceived up to 7 years prior to the survey, and previous evidence has shown that the longer the time since conception, the less likely a pregnancy is to be reported as unintended or unwanted [43, 53]. However, median length of recall of intention in the study appeared to be quite short; 26 months overall, with a range of just a few months (22–29 months) across sites. Secondly, the use of binary and categorical variables in this study may place limits on our understanding of pregnancy intention, as the concept is socially contextualised and may be better captured on a range or continuum of feelings or intentions. Pregnancy intention is so closely linked to sociodemographic conditions that separating and categorising intentions in meaningful ways becomes difficult [9], and broadening the definition to include multiple social and cultural understandings of pregnancy intention is important. Thirdly, in examining the association between pregnancy intention and maternal health care utilisation, we restricted analyses to live births only, as most high-risk pregnancies ending in stillbirth or neonatal death are more likely to elicit greater obstetric care and thus increase maternal health care utilisation [24,25,26], which may have influenced our results. Lastly, 46.8% of birthweight data were missing, and it is unclear how this may have influenced the association between pregnancy intention and low birthweight.