Demographic variables
Demographic data included marital status, household income, educational attainment, and pregnancy and delivery characteristics (e.g., delivery date and method, infant sex, birth weight, NICU admission). Participants also reported the daily hours spent by themselves and their partners on childcare, housework, and paid employment.
The Edinburgh postnatal depression scale (EPDS)
The EPDS is a 10-item self-report scale developed by Cox et al. in 1987 to screen for postpartum depression, with each item rated from 0 to 3 and total scores ranging from 0 to 30. In the original validation study, the cut-off score of 12/13 yielded a sensitivity of 86%, specificity of 78%, and positive predictive value of 73% [14].
The Japanese version, developed by Okano et al. in 1996, demonstrated good reliability and validity, with a Cronbach’s alpha of 0.78 and significant differences in scores between postpartum depressed and non-depressed women [15]. The suggested cut-off score for postpartum use was 8/9 (sensitivity = 75%, specificity = 93%, PPV = 50%). For use during pregnancy, a study involving 210 women recommended a cut-off of 12/13, with sensitivity and specificity of 90.0% and 92.1%, respectively [16].
In the present study, Cronbach’s alpha (α) during pregnancy was 0.84.
Generalized anxiety disorder 7-item scale (GAD-7)
The GAD-7, developed by Spitzer et al. in 2006, includes seven items rated on a 4-point scale, with total scores ranging from 0 to 21. The original cut-off of 10 yielded a sensitivity of 89% and specificity of 82% [17].
The Japanese version, validated by Doi et al., showed high internal consistency (α = 0.92 in psychiatric patients; α = 0.91 in general adults). Using a cut-off of 10, sensitivity and specificity were 82.1% and 79.2%, respectively, in psychiatric samples [18].
Internal consistency in this study was good (α = 0.88).
Quality marriage index (QMI)
The QMI, developed by Norton in 1983, consists of six items assessing overall marital satisfaction [19]. Five items use a 7-point scale, and one uses a 10-point scale.
In Japan, the scale was translated by Moroi, with each item rated on a 4-point scale from 1 to 4, resulting in a total score ranging from 6 to 24 points [20]. In Japanese perinatal samples, the QMI has demonstrated good to excellent internal consistency (α = 0.87–0.94 for fathers; α = 0.90–0.96 for mothers) [21].
In the present study, internal consistency during pregnancy was excellent (α = 0.97).
Multidimensional scale of perceived social support (MSPSS)
The MSPSS, developed by Zimet et al. in 1988, assesses perceived support from family, friends, and significant others [22]. The subscales consist of three factors: significant other, family, and friends. Higher scores indicate a higher level of perceived social support.
Iwasa created a seven-item version in Japan in 2007 as a shortened version [23]. In a study targeting middle-aged and older adults, the Japanese version of the 12-item MSPSS showed a high positive correlation with the original scale and high consistency in score distribution and gender and age differences, suggesting its reliability and validity.
In the present study, internal consistency during pregnancy was excellent (α = 0.96).
Sense of authenticity scale (SOAS)
Ito and Kodama developed the SOAS in 2008 to measure the sense of authenticity, defined as the feeling of being true to oneself [24]. It consists of seven items rated on a five-point scale, with higher total scores indicating a stronger sense of authenticity. Regarding reliability, the internal consistency among college students was acceptable (α = 0.79). The test-retest reliability was also adequate (r = 0.70, p < 0.01). They confirmed a one-factor structure, supporting the scale’s construct validity. They also reported acceptable reliability (r = 0.70, p < 0.01). The scale correlated positively with self-esteem (r = 0.73, p < 0.01).
Given its reported associations with reduced anxiety and depression, enhanced purpose in life, personal growth, positive social connections, and autonomy, we considered the sense of authenticity measured by this scale to represent a core facet of self-esteem in our analysis.
In this study, internal consistency for data from pregnant women was good (α = 0.86).
Statistical methods
To assess potential attrition bias, we compared baseline characteristics between the 768 participants who completed the 52-week postpartum survey and the 971 participants who dropped out after the pregnancy survey.
We then conducted a multinomial logistic regression analysis to examine factors associated with changes in marital satisfaction (QMI scores) from pregnancy to 52 weeks postpartum. Of the 768 participants who responded at both time points, we excluded 16 individuals (three who started a new relationship, six who experienced separation, and seven who consistently had no partner), as tracking changes in marital satisfaction was not feasible for these cases. As a result, 752 participants with continuous relationships were included in the final analysis.
All questions were mandatory within each survey wave; thus, no item-level missing values were missing. The dependent variable was categorized into three groups based on the difference in total QMI scores between pregnancy and 52 weeks postpartum: a deterioration group (≥ 1-point decrease), an unchanged group (0-point difference), and an improvement group (≥ 1-point increase). The unchanged group served as the reference category. We selected 52 weeks postpartum as the endpoint because, by that time, children’s daily routines are typically stabilized and maternal physical recovery is complete, making it a more valid indicator of sustainable marital satisfaction.
Independent variables were selected based on previous studies of perinatal marital dynamics, as referenced in the Introduction. Variables with variance inflation factor (VIF) values below five were included to avoid multicollinearity. We ensured an event-per-variable (EPV) ratio of ≥ 10 to maintain model stability.
Categorical predictors included parity (primiparous vs. multiparous), the respondent’s time spent on household chores and childcare (categorized as follows: <4 h [low], 4–<8 h [moderate], 8–<16 h [high], 16–<24 h [very high], and ≥ 24 h [extremely high/abnormal]), presence of cohabitants (individual and children, nuclear family, extended family), and child’s gender. Continuous variables included respondent age, household income, MSPSS subscale scores (family and friends), SOAS total score, the respondent’s and her partner’s working hours and educational attainment, and total scores on the EPDS and GAD-7. Partner-related variables (e.g., education and working hours) had slightly smaller sample sizes (N = 1719) than maternal variables (N = 1739), because some participants either did not have a cohabiting partner, were separated, or did not provide information on their partner.
Among these, log transformation was applied only to the EPDS and GAD total scores during pregnancy, as these variables showed positively skewed distributions (EPDS skewness = 1.23, kurtosis = 1.03; GAD skewness = 2.23, kurtosis = 5.53). Although the EPDS distribution was not as extreme as the GAD distribution, both were transformed to maintain analytic consistency and improve model stability. For other continuous measures, the distributions were not perfectly normal but were considered sufficiently close to normality, and thus, raw scores were used. Descriptive statistics, including skewness and kurtosis for all study variables, are presented in Supplementary Table S1.
Time spent on housework and childcare exceeding 24 h per day was treated categorically as an indicator of psychological distress. All analyses were conducted using IBM SPSS Statistics for Windows, version 30.0 (IBM Corp., Tokyo, Japan). Statistical significance was set at p < 0.05.
link