Sample
We conducted a cross-sectional analysis using baseline dyadic data collected between September 2018 and June 2019 from the CHARM2 [Counseling Husbands and wives to Achieve Reproductive Health and Marital equity] intervention study of young women (18 to 29 years old) and their husbands in rural Junnar district, Maharashtra, India (N = 1201). Gender matched interviews were carried out in-person separately with husbands and wives using electronic tablets lasting for about 40 minutes. CHARM2 is a two-arm cluster randomized controlled trial (RCT) to evaluate a gender-synchronized, gender-transformative family planning intervention. CHARM2 aims to increase uptake of contraceptives, prevent unintended pregnancy, and decrease interpersonal violence. Couples who were not currently married or cohabiting, or who were using a permanent contraceptive method, were not eligible to participate in the study, in order to meaningfully measure study outcomes, including contraceptive use and unintended pregnancy at follow up. The detailed protocol for this cluster-RCT is published elsewhere [27]. Participants were recruited from households in each of the 20 geographic clusters, with each cluster based on its attachment to a single public sub-health centre catchment area. We then randomized to intervention or control condition at the cluster level; all clusters were identified and randomized prior to study recruitment. The analytic dataset for the current study excluded couples with currently pregnant wives (n = 199) and missing information on decision-making (n = 36). Additionally, one couple missing demographic information and couples using uncommon methods (injectable contraceptive (n = 3) and emergency contraceptive pill (n = 1) were excluded, for a final sample of 961 couples. The University of California San Diego, ICMR-National Institute for Research in Reproductive Health in India, and the Population Council obtained approval from their respective IRBs for the protocol.
Measures
The primary outcome of interest was women’s report of any current modern contraceptive method (dichotomized as yes/no) based on past three months use. Modern contraceptive methods included were oral contraceptive pills, Intrauterine Devices (IUDs), and male condoms [28]. Our survey included all types of contraceptives that were available in the study area as response options, though only condoms, pills, IUDs, and emergency contraceptive pill are modern spacing methods covered under the public health system. For assessing the association between couples' concordance on women’s contraceptive decision-making agency and women led contraceptive use, methods included were non-modern (withdrawal and rhythm), male condoms, pills and IUDs, where use of pills and IUDs can be considered as women led.
The primary exposure of interest was couples’ perceived women’s contraceptive decision-making agency, and included both wife and husband’s report of wife’s involvement in contraceptive decision-making. Both members were asked, “Would you say that using or not using contraception is: mainly your decision, your husband’s/wife’s, joint by both husband and wife, your mother, mother in law, elderly head of household, your sibling, your husband’s/wife’s sibling or someone else?” The responses were collapsed into four categories of decision-making including woman alone, husband alone, wife and husband jointly, or others. The final couples’ concordance/discordance on women’s involvement in contraceptive decision-making variable was constructed combining husband and wife reports into four categories of contraceptive decision-making:
-
Concordant 1 (women and men in agreement): Both agree women were involved (women only or joint decision-making)
-
Concordant 2: Both agree that women were uninvolved (men only or others decided).
-
Discordant 1: Women report women were involved and men report women were uninvolved
-
Discordant 2: Women report women were uninvolved and men report women were involved
Additional variables included a priori as confounders based on previous literature and author expertise were: wife’s age, wife’s education (none or primary, secondary or higher), husband’s age, husband’s education (none or primary, secondary or higher), caste (General, Scheduled Caste/Scheduled Tribe/Other Backward Castes), religion (Hindu, non-Hindu), parity (0, 1, 2–4), any living sons (Yes, No), fertility desires (Have a/another child, No more/none, Undecided/ Don’t know), Below Poverty Line card holder (Yes, No), and wife’s age at marriage. In addition, we included women’s report of ever experience of intimate partner violence (physical and/or sexual), wife’s knowledge of contraceptive methods (number of methods), husbands knowledge of contraceptive methods, and couple concordance of contraceptive discussion in the past 3 months (both yes, both no, Wife yes/Husband no, Wife no/Husband yes). For assessing intention to use, women and men were asked: “Will you use a contraceptive method or continue to use one in the next 3 months to avoid or delay pregnancy?” with a yes/no response.
Analysis
Descriptive frequencies and proportions were calculated. Multivariable Poisson regression was used to model the relationship between women’s involvement in contraceptive decision-making (reference group: Concordant 1) with modern contraception use for all women, in both an unadjusted and adjusted model for all potential confounders listed above. A Poisson regression with robust variance estimation for confidence intervals was carried out to limit possible inflation in the effect size relative to logistic regression, since the outcome is not rare (modern contraceptive use is greater than 10% in this sample) [29, 30]. All comparison contrasts (comparing Discordant 2 with Concordant 2, Discordant 2 to Concordant 1, and Concordant 1 to Concordant 2) in both unadjusted and adjusted models are reported in Additional file 1: Table S2.
An exploratory analysis to examine contraceptive use intention was carried out with the multivariable model, adjusting for women’s intention to use modern contraceptives (Additional file 1: Table S3 M2), and then men’s intention to use modern contraceptives (Additional file 1: Table S3 M3). Further, an equivalent multinomial logistic regression was carried out with categorical type of contraceptive use as the outcome.
As a sensitivity analysis (Additional file 1: Table S1), a propensity score adjusted Poisson regression was carried out to limit possible selection bias from the observational design of the study. All analyses were conducted using STATA version 14.0 [31].