Measures

Primary exposures were four partner-related autonomy constraints: 1) RC, 2) IPV, 3) lack of partner encouragement for ANC, and 4) lack of partner encouragement for PNC. Partner-related autonomy constraint items were pilot tested using cognitive interviewing prior to survey implementation—no issues with comprehension and interpretation of items were indicated.

RC was measured via a four-item pregnancy coercion sub-scale modified from the RC Scale that was developed, validated, and refined in the United States.21, 22, 23 Psychometric analyses indicated one latent construct (eigenvalue = 1.84; factor loadings>0.4) and moderate reliability (Cronbach's alpha = 0.69). Past-year RC items comprised “said he would leave you if you didn't get pregnant”; “told you he would have a baby with someone else if you didn't get pregnant”; “took away your family planning or kept you from going to the clinic”; and “hurt you physically because you did not agree to get pregnant”. A fifth item was measured within PMA Ethiopia (told you not to use family planning), however, previous analyses indicated that item wording lacked clear motivation and thus functioned differently from other RC items; therefore, it was not utilized for the present analysis. All items were assessed at baseline interview when women were pregnant (i.e., examined experience prior to pregnancy). RC was analyzed as a binary variable, with affirmative response to any of the four RC items indicative of any past-year RC experience.

IPV during pregnancy was measured via the 10-item Revised Conflict and Tactics Scale,²⁴ which asks about specific violence behaviors per best practices for violence research.²⁰ IPV during pregnancy was measured at the six-week postpartum interview to ensure the most thorough measurement of violence over the course of the entire pregnancy. IPV was analyzed as binary, with affirmative response to any of the ten items indicative of IPV experience during pregnancy.

Lack of partner encouragement of ANC was measured via a single item assessed at the six-week postpartum interview: “Did your partner encourage you to go to the clinic for antenatal care?” Response options included: “Yes”; “No, did not encourage”; and “No, actively discouraged”, where “No, did not encourage” and “No, actively discouraged” were combined into a single binary response to indicate lack of encouragement due to small cell sizes. Notably, <1% of participants indicated “actively discouraged”, with the majority of affirmative responses indicating “did not encourage.”

Similarly, lack of partner encouragement of PNC was measured via a single item assessed at the six-week postpartum interview: “Did your partner encourage you to go to the clinic for postnatal care?” Response options and distributions mirrored those of lack of partner encouragement of ANC. Similar to lack of encouragement of ANC, this dichotomized variable largely reflects a lack of encouragement from partners, with overt discouragement being much less common (<1% of women).

Additionally, given substantial overlap between lack of partner encouragement of ANC and lack of partner encouragement of PNC, a categorical variable was created to understand potential dose–response associations of lack of encouragement with care-seeking throughout the MNH continuum of care. The categorical variable was defined as encouragement of ANC and PNC (coded as 0); partner encouragement of ANC or PNC (coded as 1); and lack of partner encouragement for both ANC and PNC (coded as 2).

Primary outcomes examine a range of care-seeking behaviors across the MNH continuum of care, specifically, 1) complete ANC contact (four or more visits), 2) facility-based delivery, 3) receipt of any PNC by six-weeks postpartum, and 4) infant immunization (receipt of both Bacillus Calmette–Guérin (BCG) and polio vaccination); all outcomes were assessed dichotomously and utilized standard assessments.^17,19 Additionally, a binary variable was created to examine completion of the MNH continuum of care (completion of all four care-seeking behaviors vs. three or fewer care-seeking behaviors). All outcome data collection occurred at the six-week postpartum interview.

A number of sociodemographic covariates that could confound the associations between partner-related agency constraints and care-seeking behaviors were considered based on theory and prior literature, including, residence (urban/rural); age (15–19, 20–29, 30–39, 40–49); marital status (married/living with a partner); education (never attended, primary, secondary or higher); parity (nulliparous, 1, 2–3, 4+), religion (Orthodox, Muslim, Protestant/Other); and polygynous marriage (yes/no). Additionally, given split of the six-week interview into pre-COVID and during-COVID periods, a pre/post variable was created to assess potential confounding by six-week interview period.