Partner involvement in abortion trajectories and subsequent abortion safety in Nigeria and Côte d’Ivoire

Background: Unsafe abortions contribute to maternal mortality and morbidity worldwide, with disproportionate impacts in lower-income countries. Identifying factors associated with an elevated risk of experiencing an abortion under the most unsafe conditions is an important component of addressing this burden. The partner’s role in obtaining a safe or unsafe abortion is not well understood. This study provides a quantitative assessment of the relationship between partner involvement and subsequent abortion safety. Methods: The data are drawn from the PMA2020 female surveys and abortion follow-up surveys, fielded in Nigeria and Côte d’Ivoire between 2018 and 2020. The sample includes 1144 women in Nigeria and 347 women in Côte d’Ivoire who reported having ever experienced an abortion. We assess partner involvement in discussing the abortion decision and/or in selecting the method or source and evaluate the relationship between partner involvement and most unsafe abortion (using non-recommended methods from a non-clinical source) versus safe or less safe abortion, adjusting for sociodemographic characteristics. Results: We find a strong association between experiencing any partner involvement and decreased odds of experiencing a most unsafe abortion (Nigeria: aOR = 0.34, 95% CI 0.26–0.45; Côte d’Ivoire: aOR = 0.27, 95% CI 0.16–0.47). Analyzing the two types of partner involvement separately, we find that partner involvement in the decision is associated with lower odds of most unsafe abortion in both countries (Nigeria: aOR = 0.48, 95% CI 0.39–0.72; Côte d’Ivoire: aOR = 0.34, 95% CI 0.19–0.60); partner involvement in selecting the method and/or source was only significantly associated with lower odds of most unsafe abortion in Nigeria (Nigeria: aOR = 0.53, 95% CI 0.39–0.72; Côte d’Ivoire: aOR = 0.65, 95% CI 0.32–1.32). Conclusion: In Nigeria and in Côte d’Ivoire, respondents whose partners were involved in their abortion trajectory experienced safer abortions than those whose partners were not involved. These findings suggest the potential importance of including men in education on safe abortion care and persistent need to make safe abortion accessible to all, regardless of partner support. Nigerian federal law only allows for legal abortion in order to save a woman’s life, and in Côte d’Ivoire, abortion is only legal in order to save a woman’s life or in cases of pregnancy resulting from rape [29]. Despite limited grounds for legal abortion, annual abortion incidences are estimated at 45.8 abortions per 1000 women of reproductive age in Nigeria (95% CI 41.0–50.6) [30] and 40.7 abortions per 1000 women in Côte d’Ivoire (95% CI 33.3–48.1) [31]. Over 60% of abortions in both countries are classified as “most unsafe”, involving non-recommended methods from a non-clinical source [30, 31]. This study uses data from PMA2020 in Nigeria and in Côte d’Ivoire. PMA2020 uses a three-stage cluster sampling design in Nigeria (geopolitical zone, state, geographic cluster) and a two-stage cluster sampling design in Côte d’Ivoire (region, geographic cluster), stratified by urban/rural residence to achieve a nationally representative sample of reproductive-aged women in each country (for further sampling information, see: https://www.pmadata.org/data/survey-methodology). We used Nigeria Round 5 and Cote d’Ivoire Round 2 household and female survey data, which were conducted in April–May 2018 and July–August 2018, respectively. These rounds included abortion modules and a subsequent abortion follow-up study. First, the PMA2020 female survey collected demographic information, and data related to reproductive and family planning history. Embedded within the female survey, the abortion module included questions about community norms related to abortion, and the respondent’s abortion experience. The module used two sets of terminology to ask about abortion history: ever doing something to remove a pregnancy or ever doing something to bring back a late period when respondents were worried that they were pregnant, both of which we considered to be abortions. All respondents who reported ever having had a completed abortion in the female survey, and who consented to be recontacted, were eligible for the abortion follow-up survey. The follow-up data collection occurred from November 2019–February 2020 in Nigeria and October–November 2020 in Cote d’Ivoire. The survey was not fielded in Kano State in Nigeria due to the scarcity of eligible respondents (44 respondents reported ever having had an abortion). The abortion follow-up surveys in both countries asked respondents additional questions about their only or most recent completed (i.e., successful) abortion. The follow-up survey included questions confirming information about the abortion that was reported in the baseline survey, and about abortion complications, measures of patient-centered quality of care, post-abortion contraceptive use, abortion preferences, how respondents understood the concepts of pregnancy removal vs. period regulation, and awareness of legal status of abortion. It also included questions regarding partner involvement in relation to the first or only abortion method/source sought by the respondent. The full abortion follow-up questionnaires and datasets can be requested via the PMA website (Nigeria: https://datalab.pmadata.org/dataset/doi%3A10349767ty2-va92; Côte d’Ivoire: https://datalab.pmadata.org/dataset/doi:1034976xqy6-nf94). Out of the 1790 respondents who reported ever having had an abortion in the Nigeria Round 5 female survey (excluding respondents in Kano State), 80.2% consented to be recontacted, among whom 79.7% participated in the abortion follow-up survey, resulting in a final sample size of 1144 respondents. In Côte d’Ivoire, 666 women initially reported ever having had an abortion, 70.7% of whom consented to follow-up, and 73.7% of those who consented to be recontacted participated in the abortion follow-up survey for a final sample of 347 respondents (Fig. 1). Out of these final samples, 55 (4.8%) and 14 (4.0%) respondents in Nigeria and Côte d’Ivoire, respectively, were excluded from the regression analyses due to missing data on variables included in the regression models. We use Chi-square tests to assess differences along key study variables between our final analytic sample and those respondents who were eligible but did not complete the abortion follow-up survey. Sample selection flowchart. * This count does not include respondents in Kano State. Due to the scarcity of eligible respondents based on self-report of abortion, Kano State was excluded from recruitment for the abortion follow-up survey. 1759 Kano respondents completed the female survey and 44 reported pregnancy removal or period regulation. ** Nigeria: 14 refused, 2 incomplete, 8 died, 18 did not confirm ever having had an abortion, 88 moved, 45 recruited for a separate study, 117 unknown/other. Côte d’Ivoire: 19 refused, 1 incomplete, 5 died, 16 did not confirm ever having had an abortion, 11 moved, 50 recruited for a separate study, 22 unknown/other The primary outcome examined is a “most unsafe” abortion, defined as using a non-WHO-recommended method (any method other than surgical abortion or medication abortion using mifepristone and/or misoprostol) that was provided by a non-clinical source (any source outside of the public or private medical sector) [28]. Non-clinical sources include pharmacies, other shops, friends or relatives, faith-based organizations, traditional healers, or markets/hawkers. Abortions were categorized dichotomously as most unsafe or not, based on method and source used.2 Abortions using recommended methods provided by non-clinical sources, using non-recommended methods from clinical sources, and using recommended methods from clinical sources are all coded as abortions that are not most unsafe. For women who reported multiple attempts to terminate their pregnancy, we categorized abortion safety based only on the first method/source, as we anticipated the decision surrounding which method/source to use initially would be more likely to be influenced by partner involvement compared to decisions regarding follow-up care for treatment of complications or incomplete abortion. The primary independent variable is partner involvement leading up to the abortion. Respondents were asked about multiple actors’ involvement in making the decision to have an abortion (“Did you talk to any of the following people about the decision to [remove the pregnancy/bring back your period]?”). Respondents were also asked about multiple actors’ involvement in method and/or source selection in two ways: (1) “Before deciding to use [FIRST / ONLY METHOD] from [FIRST/ ONLY PROVIDER], did you seek input or information from any of the following sources?”, and (2) “Who recommended you use this source?” (only for respondents who reported they used a source because it was recommended to them). Multiple select response options for each of these questions included: partner, various family members, friend, health provider, traditional healer, other, or none of the above. For primary analyses, these items were combined into a dichotomous variable, with respondents either being coded as having no partner involvement (if they reported no involvement in any of these items), or having their partner involved (if they reported partner involvement in at least one of these items). In the secondary analysis, we included an indicator for partner involvement in the decision to terminate the pregnancy and a separate indicator for partner involvement in selecting method and/or source (combining the two questions about method and/or source selection) to assess whether specific types of partner involvement play different roles in shaping the abortion trajectory. We considered several sociodemographic background characteristics in this analysis that, based on existing literature, plausibly shape both partner involvement and access to safer abortions. These include: age at abortion (categorized as 15–19, 20–29, and 30–49 years)3, marital status at abortion (married, not married), residence at abortion (rural, urban), schooling (level attending at time of abortion, or highest attended at interview if not attending at time of abortion; none, primary, secondary, or higher), parity at abortion (no children, any children), wealth tertile at interview (based on results from principal components analysis from information on household assets, building materials, water, and sanitation), and in Nigeria, state at interview (the 6 out of 36 states in which PMA2020 conducted the follow-up interview). We first conducted univariate and bivariate analyses, and report bivariate distributions of background characteristics and abortion characteristics, by partner involvement (any vs. none) and by abortion safety, per country. Next, we implemented country-specific multiple logistic regression models regressing most unsafe abortion on the binary measure of partner involvement exposure. We subsequently used country-specific unadjusted and adjusted multiple logistic regression models assessing the separate association between most unsafe abortion and the two types of partner involvement (discussing abortion decision and selecting abortion source/method). This model allowed us to explore whether one type of partner involvement may be driving any observed associations between the binary partner involvement indicator and the outcome, accounting for the possibility of partner involvement in both ways.4 We also tested for an interaction effect by fitting additional multiple logistic regression models including these same involvement type-specific indicators and including an interaction term for the two types of partner involvement to explore a potential non-additive relationship between experiencing both types of involvement and abortion safety. We further considered the role of gestational age, which may influence both partner involvement and the availability of safe methods (though the causal relationship may work in either direction), by including gestational age categories (< 12 weeks vs. ≥12 weeks) in a multiple logistic regression of most unsafe abortion on binary partner involvement.5 Given that the relationship between partner involvement and abortion safety was not qualitatively different from the analyses without gestational age, we present these results in supplemental materials (Additional file 1: Table S1). All analyses were conducted in STATA version 17 [32], with statistical significance set a priori at p < 0.05.