The study adopted a prospective, repeated measures design. Participants completed questionnaires and provided demographic information during the third trimester of pregnancy (Time 1) and at 4-6 weeks following childbirth (Time 2).

Women in their third trimester of pregnancy were eligible to take part in the study, regardless of age or parity. They were recruited through antenatal classes in a large National Health Service Hospital in Manchester, United Kingdom. As all questionnaires were presented in English, participants who indicated that their understanding of English was insufficient were excluded.

Postpartum Depression. Two measures were used to assess postpartum depression. The 10-item Edinburgh Postnatal Depression Scale34 (EPDS) is a screening tool, rated on a 4-point scale. Although a score of 12 indicates depression, a cut-off score of 10 has been used in community samples to prevent false negatives.35 The EPDS has a Cronbach's alpha reliability of .87 (.85 for this sample), specificity of 78% and sensitivity of 86%34 and good face validity36. This is a widely used measure which has been validated for use during pregnancy.3738 Some EPDS scores are presented to allow for comparison with other studies.

The Beck Depression Inventory II39 (BDI-II) is a revision of the original BDI40 designed to correlate more with DSM-IV criteria for depressive disorders. It is a 21-item-measure using a 4-point rating scale (0-3). Scores can be categories into minimal depression (0-13), mild depression (14-19), moderate depression (20-28) and severe depression (29-63).

The BDI-II has good internal consistency (alpha .92 for outpatients) and good test-retest reliability (.93; .82 for this sample). The BDI-II has been used with postpartum populations and has good concurrent validity with measures of postpartum depression41. The BDI-II provides a depression severity score and its total score was therefore used as the main indicator of depression.

Maternal-specific cognitions. The Pregnancy Related Beliefs Questionnaire42(PRBQ) was developed to identify the underlying beliefs of pregnant women, which may be associated with the development of postpartum depression. It was piloted on 42 women between 6 to 40 weeks gestation. It has 54 items rated on a 7-point scale from 'totally agree' to 'totally disagree'. Scores range from 128 to 245; a mean score of 191.1 (SD=29.6) was reported by Moorhead and colleagues42. High scores denote unrealistic expectations. The PRBQ has good internal reliability (.85; .93 for this sample). Examples of items included are 'motherhood is an instinctive and natural state for a woman' and 'if my baby is unhappy I will feel that it is my fault'. The PRBQ was chosen because this scale comprehensively assesses maternal beliefs about pregnancy and motherhood.

The Multidimensional Perfectionism Scale43 (MPS) consists of 35 items, scored on a 5-point scale from 'strongly disagree' to 'strongly agree'. It has six subscales: concern over mistakes; personal standards; parental expectations; parental criticism, doubts about actions and organisation. A total perfectionism score can also be obtained by totalling all responses except those in the organisation subscale. Therefore total scores range from 29 to 145. The MPS has good internal reliability (.91; .96 for this sample). It has previously been used with postpartum populations and moderate associations have been found between the concern over mistakes subscale and doubts about actions subscale and the EPDS (.34 and .31 respectively).44

Developed with the aim of providing a meaningful measure of women's perception of their social support, the 6-item Maternity Social Support Scale45 (MSSS) was used. This scale is rated on a 5-point likert rating scale (always, most of the time, some of the time, rarely, never). The MSSS has good reliability with Cronbach's alpha scores of .69 (.80 for this sample) in the antenatal period and .78 (.77) postpartum.

Demographic information including age, parity, current or previous psychiatric disorder, marital and employment status and their emotional response to being pregnant were also collected.

Ethical approval was granted by relevant committees. Participants were approached in the antenatal clinic waiting room of a local Manchester hospital and given an information sheet outlining the study. Once they provided written consent, they were given the initial questionnaire pack (demographic information sheet, EPDS, BDI-II, MSSS, PRBQ and MPS), which they could choose to complete in the clinic or return in a stamped, addressed envelope. Participants also provided information on their estimated date of delivery and a contact health professional (e.g., GP, health visitor or midwife), who was contacted following the due date to ensure that it would be appropriate for the mother to continue with the study. The Time 2 packs were sent out to the women 4-6 weeks following the birth of their baby.

Data were analysed using SPSS for Windows. Sample characteristics were examined using descriptive statistics. Missing data were minimal but excluded. One-sample Kolmogorov-Smirnov tests indicated that the use of parametric tests was permitted for most variables. As log transformations were unsuccessful for two MPS subscales and the MSSS at Time 2, equivalent non-parametric tests were used. Univariate and multivariate analyses, including hierarchical multiple regression analysis, were used to test hypotheses. For this purpose, some variables (e.g., marital status) were recoded as dichotomous variables. Pearson correlation coefficients were used to examine bivariate comparisons. In order to reduce the risk of Type 1 error, a stricter p-value of .01 was adopted as the criterion for statistical significance, because of the number of multiple comparisons.

One hundred and sixty-four women were approached to take part in the study, of those 130 agreed and were given questionnaires. Seventy-three returned Time 1 questionnaires and of those women, 61 also completed Time 2 questionnaires. At the end of the study period, two women had not yet had their babies and the remaining 10 women were lost to the study. This produced an overall response rate of 46.92% and a retention rate of 83.56%.

The mean age of the sample was 28.11 years (the youngest participant was 17 and the oldest was 41) (see Table 1). Twenty-six women were married, 28 were in a long-term relationship and 7 women were single. Thirty-three women were employed full time, 17 part-time and 11 were unemployed. All of the women who responded had some qualifications, 14 had GCSE's or equivalent and the remaining 37 had gone on to further and higher education with 5 women being educated to doctoral level. Overall, women in the sample were reasonably satisfied with their current financial situation (see Table 1). Thirty-seven of the women were primiparious, 14 had 1 child and 10 participants had two or more children already.

Whilst their feelings about pregnancy were fairly positive, participants varied in respect of whether pregnancy was expected or planned. In terms of their mental wellbeing, 14.8% of participants reported having a current emotional difficulty; 4.9% indicated experiencing anxiety and 9.8% experienced depression. Forty-three % reported past emotional difficulties: 11.5% reported prior experience of anxiety, 24.6% reported past depression and 6.6% experienced anxiety comorbid with depression in the past.

Table 2 presents an overview of mean scores for antenatal and postnatal depression, perfectionism beliefs, perceived social support, maternal beliefs about motherhood. Paired samples t-tests indicated significant differences between antenatal and postnatal scores for EPDS (t=-4.72, df=60, p<0.001) and BDI-II (t=-8.32, df=60, p<0.001). Although participants endorsed significantly more items on both depression measures after childbirth, their mean scores were not indicative of severe depression. For example, using the EPDS cut-off score of 10 for community samples, 14.5% participants screened positive for antenatal depression and 17.2% for postpartum depression. When the stricter cut-off of 12 was used, only 7.2% women appeared to be antenatally depressed and 6.9% postnatally. Using the suggested categories for BDI-II scores, 95.1% presented with minimal symptoms and 4.9% with mild antenatal depression. After childbirth 77.0% presented with minimum depression, 21.3% with mild and 1.6% with moderate symptoms.

No significant differences were observed for participants’ perceptions of their social support, their beliefs about motherhood and their perfectionistic beliefs. These variables remained relatively stable from late pregnancy until after childbirth (see Table 2).

The relationships between antenatal maternal beliefs about motherhood, beliefs about perfectionism, perceived social support, labour experience and postpartum depression were examined by carrying out Pearson correlation coefficients (Table 3). No significant associations were found between antenatal maternal beliefs about motherhood and antenatal depression as well as postpartum depression (as measured by either EPDS or BDI-II at Time 2). However, postpartum depressive symptoms as measured by the BDI-II were positively correlated with overall antenatal perfectionistic beliefs (r = 0.339, p<0.01 for total MPS score) and with its subscale of concerns over mistakes (r=0.338, p<0.01). With the exception of perfectionistic beliefs about organisation, associations were also noted between postpartum depression and personal standards (r=0.323, p<0.05), parental expectations (r=0.268, p<0.05), parental criticism (r=0.256, p<0.05) and doubts about actions (r=0.291, p<0,05), albeit at a lesser significance level. Weaker correlations were also noted for postpartum depression as measured by the EPDS at Time 2 and antenatal personal standards (r=0.267, p<0.05) and doubts about actions (r=0.259, p<0.05).

A highly significant negative association was noted between perceived antenatal social support (MSSS) and postpartum depression (r=-0.417 for BDI-II and r=0.610 for EPDS, p<0.01). These findings suggest that participants with elevated postpartum depression scores endorsed more beliefs about needing to be perfect and perceived their partner or family as less supportive during pregnancy.

Postpartum depression scores were not associated with age, marital status or whether the pregnancy was expected or planned in this sample. However, postpartum depression scores were associated with participants feeling less happy about their pregnancy (on the BDI-II- r = -.27, p<.05, on the EPDS -.33, p<.01). Participants with elevated postnatal depression scores were also more likely to have had a past history of emotional difficulties (t(59) = -2.78, p<.01) but this was only the case for the BDI-II.

Linear regression analyses revealed that postpartum depression (using BDI-II) was not predicted by maternal age, marital status, educational status, employment status, financial status, number of children or whether the pregnancy was planned but feelings about the pregnancy and a history of emotional difficulties were predictive (as dichotomous variable). Of the psychological variables PRBQ scores did not predict postpartum depression, but perceptions of social support did so. Although all MPS subscales, except for organisation, predicted postpartum depression, only the MPS total scores were used in the hierarchical regression analysis because the sample size would only allow for a limited of variables to be entered.

Using the enter method at step one to control for the influence of antenatal depression (as measured by BDI-II at Time 1), a multiple hierarchical regression analysis was undertaken to explore the contribution of maternal beliefs (PRBQ), perfectionistic beliefs (MPS) and perceived social support (MSSS), to postpartum depression (BDI-II at Time 2). All other variables were entered using stepwise and these included participants’ happiness about the pregnancy and their past history of emotional difficulties.

As anticipated, at stage 1 antenatal depression contributed significantly to the regression model, F (1,53)=20.93, p<0.001, and accounted for 27% of the variance (see Table 4). The introduction of PRBQ explained an additional 15% of variation and this change in R²was significant, F (2,52)=20.71, p<0.001. The addition of MSSS explained a further 11%, F (3,51)=21.01, p<0.001. Finally, the addition of past history of emotional difficulties to the regression model explained an additional 4% and this change in R² was also significant, F (4,50)=18.38, p<0.001). Perfectionism and feelings about the pregnancy (i.e., happiness) were not retained in the model. The final model accounted for 57% of the variance in postpartum depression scores.