Participants
This study is part of the longitudinal cohort Generations
2, which follows first-time pregnant mothers during pregnancy and after they give birth to their child to study the transition to parenthood. Participants were recruited through midwifery practices in the area of Amsterdam, the Netherlands. All women who visited cooperating midwifery practices for the first time received a recruitment letter and informed consent form. Women were eligible for inclusion when they were pregnant with their first child, did not receive a prenatal diagnosis for a congenital abnormality of the fetus after ultrasounds at 12 and 20 weeks gestation and were fluent in Dutch. When they filled out the questionnaires at the first assessments, they were approximately 12 weeks pregnant (target sample,
N = 1355). For this study, we used the data from women who filled out questionnaires at three time points during pregnancy (at approximately 12, 22, and 32 weeks pregnant), and at one time point after giving birth (±3 months after birth). Written informed consent was obtained from all participants. These women were on average 29.6 years old (SD = 4.18) at first assessment. Sixty-five percent of the women had a higher educational level (higher professional education or university level), which is higher than women in the general Dutch population between age 25–35 (42–44% had a higher education in 2009–2011; (Statistics Netherlands
2015)). Most women had a partner (98.1%). The large majority (86.8%) had a Dutch background (i.e., both their parents and the women themselves were born in the Netherlands). Whenever a woman and/or her parents were not born in the Netherlands, the majority (±60%) was born in a Western country (e.g., Europe or North America; Statistics Netherlands
2015). The large majority of women (96.5%) reported that their baby was born healthy.
Analyses were performed on a sample of 1073 women (79% percent of the target sample) for whom pregnancy-specific anxiety scores at the third time point (t3, M = 32.4 weeks of pregnancy, SD = 3.89) and at least one of the other two earlier time points during pregnancy (i.e., t1, M = 13.7 weeks of pregnancy, SD = 3.17; t2, M = 22.2 weeks of pregnancy, SD = 2.77), parenting stress scores at t4 after child birth (t4, M = 14.0 weeks after child birth, SD = 3.19), as well as the control variables included in this study (described later in this section) were available.
Measures
Parenting stress constructs were assessed at 3 months after birth (t4) with the Nijmeegse Ouderlijke Stress Index (NOSI; de Brock et al.
1992), the Dutch version of the Parenting Stress Index (PSI; Abidin
1995). The PSI assesses perceived parenting stress due to different sources, which can be parent- or child-specific domains. Only the parenting domain of the PSI—containing 58 self-report items—was assessed in this study because the child-specific domains could not be applied 3 months after birth. Mothers could respond on a 6-point Likert scale with scores ranging from 0 (totally disagree) to 5 (totally agree). The parenting domain consists of seven subscales: sense of competence (13 items; Cronbach’s alpha = .84; item example, I cannot take decisions without help). A high score on this scale reflects a poor sense of competence; role restriction (seven items; alpha = .83; example, I feel restricted because of my duties as a parent; attachment (seven items; alpha = .60; example, My child and I always have a good relationship, reverse-coded item). As with sense of competence, a high score on this scale also reflects poor attachment of the parent to the child; depression (12 items; alpha = .81; example, There are quite some things in my life that bother me); experience of health (six items; alpha = .75; example, Lately, I feel fine physically, reverse-coded item), with high scores reflecting poor experienced health; social isolation (six items; alpha = .69; example, I feel alone and without friends); and relationship with spouse (seven items; alpha = .74; example, Since the birth of this child, my partner gives me less support and help than I expected), with high scores reflecting poor relationship with spouse. Before calculating mean scores of the subscales, all items (six items) where a high score represented low parenting stress were reversed, so that high scores indicated high parenting stress. For each subscale, mean scores were then calculated by dividing the sum score across items with valid data by the number of valid items in the subscale. These mean scores ranged from 0 (low parenting stress) to 5 (high parenting stress) Note that the relationship with spouse subscale was not available for single mothers without a partner (1.5% of the final sample).
Pregnancy-specific anxiety was assessed at approximately 12 (t1), 22 (t2), and 32 (t3) weeks gestation with the Dutch version of the self-report Pregnancy-Related Anxieties Questionnaire-Revised (PRAQ-R; Huizink
2000; Huizink et al.
2004). The PRAQ-R is a shortened version of the PRAQ (Van den Bergh
1990), containing the 34 items of the original PRAQ with the highest factor loadings on each of the five subscales (i.e., fear of giving birth, fear of bearing a physically or mentally handicapped child, fear of changes and disillusion in partner relationship, fear of changes, and concern about one’s mental well-being and mother child-relationship). The PRAQ-R contains items such as “I am afraid of pain during the contractions and the child-bearing,” “I have thoughts our child will be infirm or weak,” “I am afraid my partner is unfaithful to me,” “I worry about my unattractive appearance,” and “I worry about the sudden changes of my mood.” Scores on each item ranged from 0 (absolutely not applicable) to 4 (very well applicable). Cronbach’s alpha for the total PRAQ scale ranged from .89 to .90 across the three time points. Total mean item scores for pregnancy-specific anxiety were computed, with a minimum of 0 (not anxious at all) and a maximum of 4.
State anxiety was measured with the Dutch adaptation of the State-Trait Anxiety Inventory (STAI; Spielberger et al.
1970) at t1-t3, which has been shown to be a reliable and valid measure in general populations but also in pregnancy (Gunning et al.
2010; Meades and Ayers
2011). The state anxiety scale of the STAI contains 20 items that ask the participant to describe how she feels at the moment, conceptualized as a transitory emotional state (in contrast to trait-anxiety, asking how she generally feels). Items in this scale were for example “I am tense,” “I am confused,” “I am worried,” “I feel calm,” and “I feel self-confident.” Women could respond on a 4-point Likert scale ranging from 1 (not at all) to 4 (very much). Some items were stated as an opposite of anxiety and scores were reversed, so that higher scores represented more anxiety. Cronbach’s alpha for the scale ranged from .93 to .94 across the three time points. Mean item scores with a minimum score of 0 (not anxious at all) and a maximum of 3 were calculated for the main analyses.
Trait anxiety was also measured with the STAI (Spielberger et al.
1970). The trait anxiety scale of the STAI contains 20 items that ask the participant to describe how she feels generally, referring to a relatively stable proneness to anxiety. Items in this scale were for example “I feel exhausted,” “I lack self-confidence,” “I feel nervous and restless,” “I feel calm and cool,” and “I feel at ease.” Women could respond on a 4-point Likert scale ranging from 1 (not at all) to 4 (very much). Like state anxiety, some item scores needed to be reversed, so that higher scores represented more anxiety. Cronbach’s alpha for the scale ranged from .92 to .93 across the three time points. Mean item scores with a minimum score of 0 (not anxious at all) and a maximum of 3 were calculated for the main analyses.
Depression was assessed with the Dutch version of the Beck Depression Index (BDI; Beck et al.
1961
) at t1-t3. The BDI consists of 21 items that assess the intensity of (symptoms of) depression, including “Sadness,” “Guilty feelings,” “Loss of energy,” “Irritability,” and “Indecisiveness.” Women could select one of out of four options, ranging from 0 (representing low intensity of depression; e.g., I do not feel sad, or I make decisions about as well as ever) to 3 (representing high intensity; e.g., I feel so sad or unhappy I can’t stand it, or I have trouble making any decisions). The BDI has also been validated for use in pregnancy (Holcomb et al.
1996). Cronbach’s alpha ranged from .82 to .84 across the three time points. For the analyses, the mean was computed of the 21 depression items, ranging from 0 (low intensity of depression) to 3.
Higher educational level, ethnicity, and having a partner were dummy-coded as, respectively, 1 = higher professional education or university degree, 0 = all other educational levels; 1 = both parents and participant born in the Netherlands, 0 = at least one parent and/or participant born elsewhere; and 1 = having a partner, 0 = no partner.
Life events were also assessed at 3 months after birth with the PSI (de Brock et al.
1992). The life events subscale consists of 40 items, in which mothers were asked about the presence of major/stressful life events (e.g., accident, marriage, death of a close family member, financial problems) in the last 12 months. When any of the life events occurred, the score was 1 for that life event, and 0 when the event did not occur. Because all women were pregnant and gave birth, these two items were excluded. The remaining 38 scores were summed.
Birth-related covariates were also assessed at 3 months after birth, including gestational age at birth, and interventions during labor (interventions, incision, surgical evacuation of placenta, artificial rupture of the membranes, labor induction in the hospital, artificial stimulation of contractions, suturing, vacuum extraction, cesarean section, planned caesarian section; labor interventions score = number of “yes” on these nine items).
Statistical approach
Before performing the main analyses, non-response analyses, means, and standard deviations for the study variables at each time point were calculated, as well as correlations between the variables. The main analyses were performed in three steps. In the first step, latent scores were computed for the t1-t3 scores of each of the four mood variables separately (i.e., pregnancy-specific anxiety, state anxiety, trait anxiety, and depression), and tested in four separate models as predictors of the seven t4 parenting stress constructs. In the second step, the seven t4 parenting stress constructs were simultaneously regressed on the four latent t1-t3 mood scores. In the final step, difference scores between the observed t3 and the t1 mood scores were examined as predictors of the seven t4 parenting stress constructs. In each of these models, cross-sectional correlations were estimated between the t4 parenting stress constructs, and between the prenatal mood variables (in steps 2 and 3). To control for life events, duration of pregnancy, and interventions during labor, the t4 parenting stress constructs were also regressed on these variables. All parenting stress and mood variables were regressed on age, higher educational level, ethnicity, and having a partner.
All models were fitted in Mplus 7.3 (Muthén and Muthén
2010). Model fit was evaluated using the comparative fit index (CFI), the Tucker-Lewis index (TLI) with values >.90 indicating acceptable fit and values >.95 indicating close fit (Bentler and Bonett
1980), and the root mean square error of approximation (RMSEA) with values ≤.08 indicating acceptable fit and values ≤.05 indicating good fit (see Marsh et al.
2004). A robust maximum likelihood estimator (MLR), which produces robust standard errors, was used to account for the non-normal distribution of mood and parenting stress scores. Due to the large sample size and the number of tests performed in this study, alpha was set at .01.