Abstract
Recent meta-analyses suggest that the primary risk factors for postnatal depression (PND) include past history of psychopathology, psychological disturbance (mainly depression or anxiety) during pregnancy, poor marital relationship, poor social support, stressful life events prior to birth, low social status, child care stress and birth complications [1–3]. However, risk factors are not necessarily consistently identified across studies. For example, poor marital relationship has been consistently identified as a risk factor for PND, while obstetric complications were identified as a risk factor in four studies, but not in nine separate studies [4].
The majority of early research suggested that obstetric factors were unrelated to PND [5–9], with several exceptions [10–12]. Similarly equivocal findings have recently been reported. In one study, two obstetric factors, an unplanned pregnancy and not breast-feeding at 6 weeks, were reported to be associated with significantly higher odds of PND, on both a univariate and multivariate basis [13]. Warner et al.'s study [13], which assessed eight obstetric factors, considered the combined effect of demographic variables, but not other factors which have previously been shown to be associated with increased risk for PND, such as personality [14], psychiatric history [15] and life events [7]. In another study it was reported that women who experienced adverse obstetric histories prior to the current pregnancy were reported to be at greater risk, although this was not statistically significant, of developing PND; however, complications of the current pregnancy, labour or childbirth were not associated with increased risk [16]. This study was limited to primiparous women between the ages of 20 and 40 years who were married or cohabiting. Several other recent studies, which have shown associations between PND and obstetric factors, such as pregnancy complications [17], two or more antenatal procedures, caesarean/forceps delivery, mid-high obstetric procedure scores [18] and caesarean section [19], were limited by retrospective designs and, as a result, were susceptible to potential recall bias.
One recent study reported no association between obstetric and/or perinatal complications and the development of PND [20]. This study also examined sociodemographic factors, psychiatric history and life-event variables (several of which showed significant associations with PND), but did not carry out multivariate analysis, and did not give details of the obstetric factors examined.
In summary, limitations in methodology, design or analysis were evident in the majority of recent studies of obstetric risk factors for PND. The aim of the present study was to prospectively examine obstetric factors and risk for PND, while controlling for sociodemographic factors, personality and psychiatric (personal and familial) history in an unselected sample of women from urban and rural communities.
Method
Participants
The sample was drawn from the Wentworth (Nepean Hospital) and Central West (Cowra, Dudley and Orange Hospitals) regions of New South Wales, Australia. Women who delivered in one of the participating hospitals during the recruitment period were approached by postnatal ward staff and invited to participate in the study. For logistic reasons, not all women who delivered over this period were approached, making the recording of details on the total number of women approached, or those who declined to participate, impossible. From previous experience using a similar methodology we found that there were no differences in sociodemographic or psychosocial variables between participants and women who declined to participate [21]. Exclusion criteria for the study included a history of psychotic illness; drug or alcohol dependence; significant medical problems with the infant; insufficient English skills to complete the questionnaire; younger than 16 years of age; or a history suggesting severe personal dysfunction as evidenced by repeated deliberate self-harm. Recruitment occurred from September 1995 to January 1996 in the Wentworth region and from November 1995 to March 1996 in the Central West region. The Wentworth Area Health Service and the Central West Area Health Service granted ethics approval for this study.
Measures
Records from the New South Wales Midwives Data Collection, a population-based surveillance system which includes items on maternal health, the pregnancy, labour, delivery and perinatal outcomes, were obtained for each participant to examine obstetric factors. This form is completed by the attendant midwife at the time of births.
A postnatal questionnaire, which was purpose-designed for this study, was used to obtain sociodemographic information about participants and their partners and families, psychiatric history information from participants and their families and personality and life-events information from participants. The personality component was adapted from the Structured Clinical Interview for DSM-III-R (SCID) personality section [22], while the life-events component was based on the Tennant and Andrews life-events scale [23]. The Postnatal Questionnaire was completed within 1 week of delivery.
The Edinburgh Postnatal Depression Scale (EPDS) [24], a measure of postnatal morbidity, was completed by each participant at 8 weeks postpartum. The EPDS, which is now used routinely as a screening instrument for PND, has been validated in Australia [25] and the UK [26] and has also been shown to detect depression in non-childbearing women [27]. In this study, women who scored above 12 on the EPDS were defined as depressed. This cut-off score has previously been shown to have high sensitivity, specificity and positive predictive power for PND [24–26,28]. The EPDS was mailed to participants at 8 weeks postpartum with a reply-paid envelope. Those women identified as depressed were referred for appropriate treatment. The high response rate is due to careful attention to follow-up during this section of the experimental methodology.
Personality was assessed using the Vulnerability Personality Style Questionnaire Scale (copies of this scale are available from the corresponding author) which measures two personality dimensions: vulnerability and organized/responsive personality style. The instrument has good internal consistency (coefficient 〈?= 0.63 for the whole scale, 0.77 for the vulnerability subscale and 0.17 for the organized/responsive subscale) and acceptable test–retest reliability.
Data analysis
Determination of a statistically significant relationship between an independent and the dependent variable (case/non-case), expressed in terms of the odds ratio (OR), was based on the χ2 statistic (Mantel-Haenzel weighted), or the Fishers Exact test where a cell value was less than five. Ninety five per cent confidence intervals (CIs) were based on Cornfield's approximation or exact limits where appropriate (i.e. where cells contained small numbers).
Results
A total of 504 women consented to participate in the study, across the two regions (with 268 and 236 participants from the Wentworth and Central West regions respectively). Data from 14 of these women (seven from each region) could not be used due to a partial or complete failure to complete the EPDS, reducing the total sample size to 490.
Sociodemographic characteristics of the sample
The age range of the sample was 16.0–42.8 years, with a mean age of 28.0 years (SD = 5.2); 58.7% of participants were aged between 25 and 34 years. Three hundred and ninety (79.6%) of the women were married or in de facto relationships, while 61 (12.4%) women had never been married, and nine (2%) women were divorced or separated. Two hundred and four (41.6%) of the women were primiparous. More than half of the participants (n = 272, 55.7%) lived in their own house or unit, while 188 (38.5%) lived in a rented house or unit, and 28 (5.7%) lived in other accommodation (e.g. 18/28 in parent's home). The highest level of education completed was less than year 12 (high school certificate) for 46.3% (n = 227) of participants, with 12.9% (n = 63) completing year 12, 25.9% (n = 127) attaining technical college level, and 14.3% (n = 70) attaining university level. Employment status was assessed as at 3 months antenatally, with the largest proportion (n = 190, 38.9%) of women involved in home duties, 24.9% (122) being employed full-time, 17.2% (84) employed part-time, 4.1% (20) self employed, 12.3% (60) unemployed, and 2.7% (13) being full- or part-time students.
Sample representation: sociodemographics
There were no significant differences between the two samples, nor between the samples and normative New South Wales state-wide data for 1995 on sociodemographic variables.
Point prevalence of postnatal depression
The mean EPDS score was 6.2 (SD = 4.7), with the median score being 5.0. Using a cut-off score of < 12 of the EPDS to identify cases of PND, 64 women (13.1%) were defined as having PND at 8 weeks postpartum.
Univariate analysis
Sociodemographic factors associated with postnatal depression
We first tested for associations between PND and sociodemographic factors. Women with postnatal depression were significantly more likely to have had a technical college education compared with those who had not completed year 12 (OR = 2.24, CI = 1.18–4.23), live in rented accommodation (17%) as opposed to owning their own home (9.6%) (OR = 1.94, CI = 1.08–3.51) and be receiving a pension/benefit (23.2%) (OR = 2.53, CI = 1.36–4.69). No other significant differences were found.
Personality factors associated with postnatal depression
Increased odds of PND were associated with women who rated themselves as being either nervy (OR = 3.25, CI = 1.38–7.56), shy/self-conscious (OR = 4.13, CI = 2.09–8.20), obsessional (OR = 3.44, CI = 1.68–7.03), angry (OR = 2.18, CI = 1.01–4.63) or a worrier (OR = 4.40, CI = 2.07–9.57). Being timid (i.e. unable to assert yourself) was also associated with higher odds of developing PND, but this increase approached statistical significance only (OR = 1.94, CI = 0.80–4.61). Women showing mid-high scores on the vulnerability scale were at increased risk of PND (OR = 3.86, CI = 1.44–11.23).
Psychopathological factors associated with postnatal depression
We next examined whether a history of psychopathology (depression, anxiety, PND) or having a family member with a psychiatric illness increased the risk of PND. Higher odds of developing PND were associated with a history of psychopathology or familial mental illness (OR = 2.53, CI = 1.43–4.5). Specifically, women with a past history of depression (OR = 3.21, CI = 1.55–6.58), anxiety (OR = 4.2, CI = 1.77–9.85) or PND (OR = 4.0, CI = 1.77–9.10) had increased risk of developing PND. Increased odds of PND were also associated with the occurrence of depression (OR = 2.83, CI = 1.29–6.15) and PND (OR = 4.41, CI = 1.21–15.52) in the participant's mother. The only other significant familial psychiatric risk factor for PND was a history of alcoholism in the brother of the participant (OR = 5.91, CI = 1.51–22.75).
Life event and parental relationship factors associated with postnatal depression
The only significant associations between recent life events (i.e. those within 12 months preceding completion of the postnatal questionnaire) and PND were: major health problem (OR = 2.86, CI = 1.30– 6.23) and arguments with partner (OR = 4.57, CI = 2.13–5.29).
Obstetric factors associated with postnatal depression
Finally we examined the relationship between PND and a range of obstetric factors including obstetric history; complications of pregnancy, labour and delivery; and infant details. These were also compared to the normative 1995 New South Wales data. Our sample was representative in terms of the occurrence of obstetric variables. None of the obstetric factors were associated significantly with developing PND. There was an increased, but statistically non-significant, risk of developing PND for multiparous women (OR: 1.42), those having three previous pregnancies of greater than 20 weeks in duration (OR = 2.46), antepartum haemorrhage (OR = 3.97), gestational diabetes (OR = 2.05), pre-eclampsia (OR = 1.39), artificial rupture of membrane (OR = 1.72), epidural anaesthetic (OR = 1.36), forceps delivery (OR = 2.51), elective caesarean section (OR = 2.03), emergency caesarean section (OR = 1.40), and third-degree tear (OR = 1.61). There were no infant variables which were associated with an increased risk of PND. However, women with multiple births had a 4.5-fold increased risk of developing PND, but this was not significant since there were only five women in this study who had multiple births. Smoking during pregnancy was not identified as a risk factor for developing PND (OR = 1.13).
Multivariate analyses
Hierarchical logistic regression was used to determine the effect of the addition of variable categories on the predictive power of the model (see Table 1). With the exception of the sociodemographic variables, only variables which showed a significant univariate association with increased risk for PND were included in the multivariate analysis. When the sociodemographic variables were entered alone, the model was non-significant. The model was, however, highly significant with the addition of the personality variables (i.e. nervy, shy/self-conscious, worrier, obsessional, angry), with significant improvements in predictive power obtained with addition of the psychiatric history variables (personal or family history of mental illness, personal history of depression, history of depression in mother, personal history of anxiety, personal history of PND, history of alcoholism in brother), and further improvement with addition of the life-event and parental relationship variables (health problems, arguments with partner, arguments with others, financial difficulties, problems with the law, problems in relationship with parents). No improvement in the model was obtained with addition of the obstetric variables (number of previous pregnancies < 20 weeks, antepartum haemorrhage, gestational diabetes, normal vaginal delivery, forceps delivery, caesarean section, multiple births).
Results from the hierarchical logistic regression analysis
Discussion
The age, marital status, country of birth and insurance status characteristics of this unselected sample of 490 women were representative of the regions from which the sample was drawn. Furthermore, the current sample was also representative in terms of the occurrence of obstetric factors, with only six of the 43 variables showing greater than 5% deviation from the regional/state-wide rate (the largest deviation was 10.3% less epidural injections in the current sample). The point prevalence of PND found in the current sample (13.1%) was very similar to the overall rate of 13% derived from a recent meta-analysis of 59 studies (n = 12 810), and was slightly higher than the rate for self-report assessment (12%) [1].
Overall, the results of this study reiterate the importance of psychosocial risk factors for PND [1–3], via both univariate and multivariate analyses. A regression model which contained only the sociodemographic variables did not have significant predictive power, but showed significant improvement with the incremental inclusion of the personality, psychiatric history and lifeevent variables. When obstetric variables were added to the model containing sociodemographic, personality, psychiatric and life-event variables, the predictive power of the model did not improve. Furthermore, on a univariate basis, significant associations with PND outcome were found for sociodemographic (education, residence type and pension/benefit status), personality (nervy, shy/ self-conscious, worrier, obsessional, angry), previous psychopathology and familial psychiatric history (history of personal/family mental illness, history of depression in self and mother, history of anxiety in self, history of PND in self and mother, and history of alcoholism in brother) and recent life-event factors (health problem, arguments with partner, arguments with others), while none of the obstetric variables were associated with significantly increased odds of PND.
It is interesting to note that, while non-significant, women from the rural community sample were 1.6 times more likely to have developed PND by 8 weeks postpartum than those from the urban community sample. A high rate (58%) of PND has been reported previously in women from a rural community in Australia [29], while another study reported decreased rates of PND in those that lived outside the metropolitan area of Victoria [18].
The multivariate analysis also revealed information about the independence or association of factors in predicting PND. Being divorced/separated and receiving a pension/benefit were significantly associated with PND when the sociodemographic variables alone were in the model, and also when the personality factors were added, indicating the independence of these factors. Of the personality factors, being shy/self-conscious was the only variable significantly associated with PND, suggesting that it had a significant independent influence of PND. Upon addition of the psychiatric history variables, being divorced/separated and shy/self-conscious were no longer significant, suggesting that the influence of these two factors on PND was through their association with previous psychopathology and familial psychiatric history. At this step, receiving a pension/benefit or having a brother with a history of alcoholism were the only variables associated with PND. Both of these variables were still significant with the addition of the lifeevent factors, with arguments with partner also reaching significance, suggesting its significant and independent influence on PND.
Receiving a pension/benefit emerged as an important factor in the development of PND in this study. The majority of previous research has focused on the overall income or socioeconomic status of the participant's family to assess the financial situation of participants, with results showing a small but significant association with PND [1]. No association between socioeconomic status and the development of PND was found here. It should be noted that the pension/benefit status variable related directly to the participant, as opposed to an overall family/household measure which was assessed by the socioeconomic status variable. Approximately one in four women (23.2%) who were receiving a pension/benefit were postnatally depressed at 8 weeks, compared with one in 10 (10.6%) of those who did not receive this financial assistance. Pension/benefit status was a significant risk factor for PND at each stage of the hierarchical logistic regression analysis (i.e. after controlling for other sociodemographic factors, as well as personality, psychiatric history and life-events factors) indicating that this variable exerted an independent effect on development of PND at 8 weeks postpartum, and was not the result of other risk factors.
Exactly how pension/benefit status influences development of PND is unclear. Certainly, this factor is very likely to be related to physical ill-health and psychosocial disadvantage, however, it may also be possible that part of this effect may be related to the various benefits of employment. Seventy-nine per cent of women receiving a pension/ benefit (at 1 week postpartum) were not employed (i.e. were either unemployed, a student or undertaking home duties) at 3 months prior to the birth. It could be hypothesized that the psychological (e.g. self-esteem, confidence, independence) and practical (e.g. responsibility, timemanagement and awareness, social/professional contact) benefits of employment, which the majority of pension/ benefit receivers do not receive, may, in combination with other factors specific to this group (e.g. negative feelings associated with lack of contribution to society), lead to higher rates of PND in this group. The nature of this effect requires further investigation.
Although none of the obstetric factors showed statistically significant associations with PND, the univariate results provide valuable information on a wide range of potential obstetric risk factors and their association with PND. In terms of the subcategories of obstetric factors, weak univariate associations with PND were found for the majority of factors in the ‘Complications of labour and delivery’ (mainly due to low numbers) and ‘Baby details’ subcategories, with generally stronger associations evident in the other three subcategories (i.e. obstetric history, complications during pregnancy and factors during labour and delivery). Findings which approached statistical significance suggested increased rates of PND in multiparous women, those with three previous pregnancies of at least 20 weeks, those who had an antepartum haemorrhage, those whose delivery was other than normal vaginal, and those who had forceps or elective caesarean delivery. Other univariate findings are worthy of note due to their potential practical or policy implications. These include (i) an increased rate of PND in mothers who attended their first antenatal visit in the third trimester of pregnancy, had gestational diabetes, S.J. JOHNSTONE, P.M. BOYCE, A.R. HICKEY, A.D. MORRIS-YATES, M.G. HARRIS 73 whose onset of labour was augmented via artificial rupture of membranes, whose infant had an APGAR (a scoring system of physical signs, with 10 being perfect) score of eight or less 5 minutes after birth; and (ii) a decreased rate of PND in mothers of children born either pre-or post-term compared with those at term.
It is acknowledged that using self-report methods to assess some risk factors (e.g. life events, personality, and psychopathology) is not as reliable as other methods such as clinical interview, however, given the large sample size this was the best way to gather this data. Additionally, it is acknowledged that the sample size is small for a true population-based study, but is large for this type of study.
In summary, this prospective study of an unselected sample reports on the potential association of a wide range of obstetric risk factors for PND, with concurrent consideration of sociodemographic, personality, psychiatric history and recent life-event variables. The results (i) confirmed the significant risk associated with psychosocial factors; (ii) indicated that, although not reaching statistical significance, several obstetric factors related to the pregnancy, complications of pregnancy, and the labour and delivery are associated with an increased risk of PND (and warrant further study); and (iii) suggested the potential importance of pension/benefit status in the development of PND.
Footnotes
Acknowledgements
This study was supported by a grant from RADGAC.