We used cross-sectional nationally representative data on individuals’ support for SRHR collected for the first time as part of the 7th global WVS wave. While the WVS has been conducted in most countries in the world, data collection in SSA has remained limited. For the current study, we used data from a new WVS module on attitudes toward SRHR developed by our team, implemented between February 2020–June 2021 in three sub-Saharan African countries where such information has been less available: Ethiopia, Kenya, and Zimbabwe. Despite great progress to ensure SRHR for all over the past decades, these countries carry a prevailing high burden of adverse outcomes such as maternal mortality and morbidity, complications from unsafe abortion, gender-based violence, adolescent childbearing, and limited sexual rights (Supplementary Table A1) [18]. The three countries also differ in terms of their abortion legislation, prevalence of HIV, and harmful practices, as well as their population size, health, and political systems [19]. They are all signatories of key SRHR documents such as the Protocol to the African Charter on Human and People’s Rights on the Rights of Women in Africa [20], which provides a policy framework to ensure SRHR, including ending harmful practices and ensuring many, although not all, reproductive rights.

The WVS has collected data on sociocultural values and beliefs through standardized face-to-face interviews with representative population-based samples of adults since 1981, available open-access. An in-depth explanation of the WVS data collection procedures to minimize bias as well as a full methodological report for each country can be retrieved from https://www.worldvaluessurvey.org. For the present study, the full WVS sample in the three included countries comprised 3,711 males and females aged 18 years or above (Ethiopia n = 1,230, Kenya n = 1,266, Zimbabwe n = 1,215). Data were collected following WVS standards including mechanisms to ensure the safety of data via direct uploading and storage of data on a highly secure password protected server. No identifying data from the participants were collected, removing the requirement for a written consent form. However, all participants were requested to provide oral informed consent, witnessed by the interviewer. The research was conducted in compliance with the principles laid out in the Declaration of Helsinki. An ethical permit was granted from the Swedish Ethical Review Authority to analyze data that were collected abroad in Sweden (Dnr 2020–05314).

Data used in the present study are based on a new SRHR module, which was first developed and piloted by our team in the Nigerian WVS wave 7 in 2018 [21]. The new module was further adapted and expanded with additional questions drawing on the Guttmacher-Lancet SRHR definition for the three countries in this study [19], which is why we did not include the Nigerian sample here.

The standard WVS questionnaire includes 14 items covering some aspects of SRHR, such as women’s role in society, subjective health status, empowerment, life satisfaction, as well as attitudes to, e.g., homosexuality, abortion, premarital sex, and divorce. In the new module, we added 44 measures of attitudes related to different domains of SRHR as per the Guttmacher-Lancet Commission definition, including child marriage, early childbearing, comprehensive sexuality education, contraceptive use, skilled birth attendance, gender-equitable relationships and gender norms, premarital sex, infertility, abortion, and sexual and gender minority rights. Supplementary Table A2 presents an overview of the complete 58-item battery and their response options. Most questions asked respondents to indicate their agreement with statements on a Likert-type response scale, such as “Please tell me if you strongly agree, agree, neither agree nor disagree, disagree, or strongly disagree with the following statements: A man should always have the final say about decisions in his relationship or marriage.” Some questions asked: “How frequently do the following things occur in your neighborhood – very frequently, quite frequently, not frequently, or not at all? Sexual assault/rape.” Finally, a third set of questions asked, “Please tell me for each of the following actions whether you think it can always be justified, never be justified, or something in between, using this card: Abortion.” The latter set of items was based on a 10-graded scale. Details on the development of the new SRHR module have been described elsewhere [19, 21].

Beyond country, we also included five sociodemographic characteristics as covariates in the current analysis: age groups (18–24; 25–29; 30–39; 40–49; 50 +), sex (we use the terms male/men or female/women interchangeably), place of residence (urban; rural), highest educational level (primary or lower; secondary; tertiary), and relationship status (married or cohabiting; divorced, separated, or widowed; single).

We used deidentified secondary data publicly available on the WVS website. Patients or the public were not involved in the design, conduct, reporting, or dissemination plans of our research.

We began with a descriptive analysis, which indicated that 25% of the full sample (N = 927) were missing responses for up to 46 items in the SRHR module. Non-response on these items varied by country, relationship status, and education, but not as much by age, sex, or place of residence. We excluded respondents with less than a 25% response rate on the SRHR items, i.e., those who did not respond to 14 or more of the total 58 items (N = 45, < 2% of the original WVS sample). Non-response rates on sociodemographic variables were low (< 3%) and deemed unproblematic for our analysis. The initial analytical sample thus included 3,666 respondents (Ethiopia n = 1,223, Kenya n = 1,228, Zimbabwe n = 1,215).

While the survey items were developed using a deductive approach, to capture the comprehensive nature of the Guttmacher-Lancet definition of SRHR, we applied an inductive, data-driven approach to develop the actual index, rather than “forcing” items into specific domains. We did this by using Exploratory Factor Analysis (EFA) to identify the most parsimonious number of hypothetical dimensions that could explain covariation among the 58 included items. EFA is useful to identify the factor structure for a set of variables inductively, without constraining items to load on specific factors [22]. Bartlett’s test of sphericity and the Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy showed that EFA was feasible.

All items were coded so that higher values indicate more supportive attitudes towards SRHR. For example, responses to the statement “Sexual education helps people make informed decisions” with response options 1 = agree completely, 4 = disagree completely, were subsequently reverse coded so that higher scores represented greater agreement, and thus more support for SRHR interventions [2]. We did not reverse code negative statements where disagreement indicated more support for SRHR, such as “A man shouldn't have to do household chores”.

Drawing on previous studies [23] we did an initial assessment of how many factors to retain based on Polychoric correlations (given the Likert-type variables), principal component analysis (PCA), scree plots, and parallel analysis (PA). Then, iterated principal factors (IPF) with oblique rotation were used to determine the appropriate number of factors.

Criteria for determining factor adequacy were established a priori: parsimony was preferred over complex loadings that were salient on more than one factor. We retained factors with a minimum of three coefficients loading > 0.40, an item uniqueness of < 0.70, and that were conceptually meaningful according to the Guttmacher definition. We used Cronbach’s α to test the internal consistency between factor items, with ≥ 0.7 considered acceptable reliability [21].

We tested the resulting factor solution with retained items: for the full pooled sample, each of the three countries, and on subsamples disaggregated by the five sociodemographic characteristics.

We started the EFA with complete cases (n = 2,722) on 58 variables. Results from the initial PCA and scree plots (Supplementary Figure A1) suggested a 9-factor solution, but some factors did not fulfill the criteria. We thus reduced the number of factors and excluded irrelevant items iteratively, until a solution was reached that met all the criteria outlined above. By excluding some items, the number of complete cases increased (n = 3,135).

The factor items were combined into subindices by extracting latent scores from each factor using regression scoring. Mean scores for the subindices were combined with equal weighting into an overall index by adding the scores and then dividing it by the number of subindices. The overall index and subindex scores were standardized to a 100-graded scale for interpretability. Higher scores represented more agreement with the achievement of sexual and reproductive health and rights.

We further conducted a sensitivity analysis to assess the proposed factor solution using multiple imputations based on 10 samples with standardized scales to fill in missing data on the SRHR items [22].

Finally, we conducted multivariable linear regression models to assess the association between the index scores with sociodemographic characteristics. Both pooled and stratified models by country and sex were conducted. This final step served both to test the construct validity of the index as a potential source of bias and as an empirical evaluation of characteristics associated with support for SRHR in the study settings.

In the regression models, we included only complete cases on the retained index items and the five sociodemographic variables (n = 3,113 or 84% of the original WVS sample). The covariates included in the regression models are displayed in Table 1. These sociodemographic characteristics did not differ notably from the original WVS sample (Supplementary Table A5). A sample flowchart is available in Supplementary Figure A3.