• Open Access

A geographic comparison of the prevalence and risk factors for postnatal depression in an Australian population

Authors


Correspondence to:
Dr Justin Bilszta, Heidelberg Repatriation Hospital (bldg 129A), PO Box 5444, West Heidelberg, VIC 3081. Fax: (03) 9496 4223; e-mail: justin.bilszta@austin.org.au

Abstract

Objective: This study sought to compare the contribution of demographic and psychosocial variables on the prevalence of, and risk for, PND in urban and rural women.

Methods: Demographic, psychosocial risk factor and mental health data was collected from urban (n=908) and rural (n=1,058) women attending perinatal health services in Victoria, Australia. Initial analyses determined similarities and significant differences between demographic and psychosocial variables. The association between these variables and PND case/non-case was evaluated using logistic regression analysis.

Results: There were a number of significant differences between the two cohorts in terms of socio-economic status (SES), age, marital status and past history of psychopathology Antenatal depression was more common in the urban group compared to the rural group (8.5% vs 3.4%, p=0.006); there was no significant difference in the prevalence of PND (6.6% vs 8.5%, p=0.165). For urban mothers, antenatal EPDS score was the best predictor of PND. For rural mothers antenatal EPDS score, SES and psychiatric history had a significant influence on postnatal mood.

Conclusions: Findings confirm the contribution of established risk factors such as past psychopathology, antenatal EPDS score and SES on the development of PND and reiterate the need for procedures to identify and assess psychosocial risk factors for depression in the perinatal period. Other predictors such as efficacy of social support and perceived financial burden may strengthen statistical models used to predict PND for women living in a rural setting.

The rural experience of postnatal depression (PND) in Australia is not well-understood.1 Knowledge is based on urban cohorts with findings generalised to rural populations. Research evaluating PND in rural populations is limited and significant methodological differences – screening instrument selection and cut-off score; timing of screening and selection of study participants – have confounded the issue. Of the studies that have investigated this phenomenon, point prevalence rates have been reported at 11.3%2 and 57.8%.3 There is conflicting data as to whether women from a rural community are at an increased risk (OR=1.6, p=NS) of depression4 or have significantly lower odds of depression (OR=0.56, p<0.05) than those living within metropolitan areas.2,5

When comparing the psychosocial risk factors associated with the development of a PND, there is a danger of assuming these factors are the same for both rural and urban women. However, as both groups of women live in different physical, economic and social environments, it is likely the impact of place of residence, and the association between residence and risk factors, on PND will differ considerably.1,6,7

It can be argued place of residence might interact with known psychosocial risk factors that contribute to development of PND, however, this is yet to be investigated. This study compared the contribution of demographic and psychosocial variables on the prevalence, and risk for, development of PND in women living in urban and rural settings.

Methods

Participants and setting

Urban data was collected from the Northern and Angliss hospitals, approximately 20km and 35km respectively, from the central business district of Melbourne, as part of the beyondblue National Postnatal Depression Program.8 These hospitals serve predominantly low to middle class areas of Melbourne. Women were recruited as part of their routine antenatal booking visit. At this time, they completed a demographics/psychosocial risk factor questionnaire and an Edinburgh Postnatal Depression Scale (EPDS). Women were followed-up at six-eight weeks postnatally via their maternal-child health nurse who asked them to complete a second EPDS.

Rural antenatal data was collected from Wodonga Regional Health Service (WRHS) medical records. Wodonga is situated on the Victoria/New South Wales state border, approximately 300km from Melbourne. The Rural, Remote and Metropolitan Area (RRMA) Index,9 uses population size and an index of remoteness related to distance from an urban centre to classify population centres as ‘metropolitan’, ‘rural’ or ‘remote’. According to RRMA, Albury-Wodonga is considered a ‘large rural centre’ with a joint population of approximately 100,000 people. WRHS has been conducting an ‘Antenatal Risk Assessment Project’ funded by the Victorian Department of Human Services since 2002. Antenatal data, collected as part of the routine antenatal booking-in appointment at the WRHS maternity unit, was matched with postnatal EPDS scores collected routinely by the Maternal Child Health Centres in Albury and Wodonga.

The final sample (n=1,966) consisted of 908 urban (59% Northern Hospital; 41% Angliss Hospital) and 1,058 rural (41.9% Wodonga; 32.3% Albury; 18.3% Lavington; 7.9% other localities) women who gave birth between 2002 and 2004.

For the rural cohort, only antenatal data with corresponding postnatal data was included in the study. Analysis of rural localities using the Selection of States (SOS) classification codes suggest that 88.3% of the sample comes from a large rural population (Code =12) with the remainder (11.7%) coming from an ‘absolute’ rural population (Codes = 15, 22 & 31).

The design of the beyondblue study was naturalistic and data collection was integrated into routine perinatal care.8 For the period that urban data was collected, the follow-up response rate (i.e. the number of women screened antenatally who were also screened postnatally) was 61% at the Angliss Hospital and 47% at the Northern Hospital.

Where possible, the following demographic and psychosocial data was collected from both cohorts of women: maternal age at the time of delivery, socio-economics status (SES), marital status, parity, language spoken at home, indigenous status, past psychiatric history, past history of childhood abuse, major life events in the past 12 months, presence and relationship with partner as well as presence and source of social support. These variables were selected following a comprehensive literature review and clinical experience of the investigators.

The Farish Index A10 was used to estimate SES based on equal amounts of occupational status, education and income according to the individual's postal area. This estimate was necessary since the education level and annual household income were unknown for rural women. The SES estimate assumes that participants live in areas that correspond with their social and financial situation. It also assumes each postal area is socially homogenous.

Current mental health status was evaluated using the EPDS.11 A threshold score of 15 or more was used to indicate likely antenatal major depression; a threshold score of 13 or more was used to indicate likely postnatal major depression.12 The higher antenatal threshold score reflects the often transient, heightened anxiety that women may experience at this time, without actually meeting diagnostic criteria for a mood disorder.13 These respective cut-off scores have been validated for English-speaking populations, at the timepoints indicated.

Minor differences in the wording of the psychosocial questionnaire were noted between the urban and rural cohorts. For the urban cohort, participants were asked for ‘diagnosed’ past ‘minor depression’, ‘major depression’, ‘anxiety’ and ‘other’ conditions, whereas the rural cohort were asked for past ‘emotional/coping problems’ with depression and panic attacks given as examples. For this study, minor and major depression was collapsed as ‘depression’ and panic attacks and anxiety were collapsed as ‘anxiety’. Similarly, the urban cohort was asked about the ‘…emotional support…’ received from a partner while the rural cohort was asked how the couple ‘…(got) along…?’. Since the two statements do not necessarily mean the same thing, responses weren't compared directly but were only used in the logistic regression model for the separate groups.

Statistical analysis

Cases with missing values were excluded in the analyses comparing demographic and psychosocial variables between urban and rural women but were included in the logistic regression calculations.

Initial data analyses determined the similarities and significant differences between the urban and rural groups in terms of demographic and psychosocial variables. The frequencies of categorical variables and the proportions of women with antenatal and postnatal depression were compared using cross-tabulations and chi-square tests. For continuous variables like age and SES, which were normally distributed, independent sample t-tests were used to compare the means between the two groups. The means of the antenatal and postnatal EPDS scores from the two groups were compared using the non-parametric Mann-Whitney test since the distributions were not normal.

The association between demographic, psychosocial and PND case/non-case was determined using binary logistic regression. The relationship between each risk factor was considered separately but thereafter a model considering all the available risk factors simultaneously was considered to obtain adjusted odds ratios (OR) and their 95% confidence intervals. The significant risk factors in this final model were confirmed using forward and backward selection of predictors and this was investigated further in order to develop a model for predicting PND in mothers.

The statistical package used was SPSS version 13.0.

Results

Demographics and psychosocial comparison

The demographic and psychosocial composition of the study cohort is summarised in Table 1 and Table 2, respectively. Associations with a Cramer's V statistic of above 0.20 should be particularly noted as this indicates a relatively strong relationship between the variables.

Table 1.  Demographic comparison of urban and rural women.
 UrbanRuralSignificance Cramer's V
 n%bn%bp-valuea 
  1. Notes:

  2. (a) from chi-square tests.

  3. (b) Counts and % for each variable may not add up to the sample size and 100% due to missing values.

  4. (c) Single/Other consists of never married, separated, divorced and widowed women.

Marital status    <0.0010.101
  Married63269.866663.0  
  De facto20322.424523.1  
  Single/Otherc707.714713.9  
Parity    <0.0010.187
  Primiparous19426.445843.3  
  1 child35448.134532.6  
  2 children13918.917516.5  
  ≥ 3 children496.7807.5  
Language spoken    <0.0010.077
  English85595.198092.6  
  Other444.980.8  
Indigenous status    0.0020.071
  Non-Ab/TSI89099.6103297.5  
  Ab/TSI40.4222.1  
Table 2.  Psychosocial history comparison of urban and rural women.
 UrbanRuralSignificance Cramer's V
 n%cn%cp-valuea 
  1. Notes:

  2. (a) from chi-square tests.

  3. (b) chi-square performed for a 2x2 table, presence/absence of a partner vs. urban/rural residence.

  4. (c) Counts and % for each variable may not add up to the sample size and 100% due to missing values.

Psychiatric history    0.0320.138
  No52477.580876.4  
  Yes15222.525023.6  
Past depression    NS0.006
  No56383.388783.8  
  Yes11316.717116.2  
Past anxiety    <0.0010.098
  No62592.5102696.8  
  Yes517.5343.2  
Other psychiatric condition    <0.0010.090
  No64871.498893.4  
  Yes171.9706.6  
Past childhood abuse    <0.0010.222
  No66982.3101395.7  
  Yes14417.7454.3  
Recent life events    <0.0010.233
  No41546.573069.0  
  Yes47753.532131.0  
Relationship with partner    0.024b0.226
  No partner283.1505.2  
Poor171.9141.5  
Moderate707.7606.2  
Good78987.383987.1  

Rural women were more often living in lower socio-economic areas (94.9 ± 4.4 vs 97.6 ± 8.6, p<0.001), tended to be slightly younger (30.8 ± 5.5 years of age vs 31.5 ± 4.7 years of age, p=0.006), were more likely to be single (13.9% vs 7.7%, p<0.001) and primiparous (43.3% vs 26.4%, p<0.001) compared to their urban counterparts. The urban cohort was more culturally diverse with more women who did not speak English at home (4.9% vs 0.8%, p<0.001), while the rural group had more Indigenous women (2.1% vs 0.4%, p=0.002).

There were significant differences between urban and rural women across all the psychological variables tested, except past history of depression. Urban women were more likely to report a past history of anxiety (7.5% vs 3.2%, p<0.001), childhood abuse (17.7% vs 4.3%, p<0.001) or recent major life events (53.5% vs 31.0%, p<0.001). Rural women were more likely to report they had experienced ‘other psychiatric conditions’ (6.6% vs 1.9%, p<0.001).

Both rural and urban women with a partner reported a ‘good relationship’−87.1% vs 87.3%, respectively. Chi-square testing was not used to compare how women rated their relationships. However, despite more rural women being single, significantly more reported antenatally that they believed they would receive adequate support upon returning home after delivery −97.9% compared to 94% of urban women (p<0.001).

Prevalence of antenatal and postnatal depression

Antenatal depression, classified by an EPDS score of 15 or more, was more common in the urban group – 8.5% vs 3.4% (p=0.006). Rural women also had lower mean EPDS scores than urban women (5.2 ± 4.2 vs 6.7 ± 5.1, p<0.001). Postnatally, 6.6% (62/908) of urban compared to 8.5% (90/1058) of rural mothers scored 13 or more on the EPDS (p=0.165). No significant differences in antenatal or postnatal EPDS scores within each geographic region were noted.

Predictors of postnatal depression

Separate binary logistic regression analyses (see Table 3) for the available risk factors suggests a significant relationship between PND and the following variables – antenatal EPDS score (OR=1.27; 95% CI 1.22 – 1.52), past psychiatric history (OR=3.69; 95% CI 2.60 – 5.21), being without a partner (OR=2.29; 95% CI 1.46 – 5.59), recent life events (OR=1.61; 95% CI 1.16 – 2.26) and past childhood abuse (OR=1.97; 95% CI 1.24 – 3.11). In comparison, having a good partner relationship (OR=0.44; 95% CI 0.21 – 0.92) and having adequate social support (OR=0.42; 95% CI 0.22 – 0.79) protects against PND.

Table 3.  Binary Logistic Regression Analysis for PND for the overall sample, all risk factors.
 Single PredictorMultiple Predictors
 p-valueOR95% CIp-valueOR95% CI
Residence
  1 urban(n=1966)
  2 rural0.166a1.270.91 -1.780.0641.610.97-2.67
Past psychiatric history<0.001  0.008  
  1 none(n=1966)
  2 yes< 0.0013.692.60-5.210.0041.941.24-3.02
  3 missing0.4190.790.40-1.550.6050.770.29-2.04
Antenatal EPDS score0.001     
 (n=1879)1.271.22-1.32<0.0011.271.22-1.32
SES0.096 (n=1955)0.980.96-1.000.2490.980.95-1.02
Age0.412 (n=1963)0.990.96-1.020.9651.000.96-1.05
Marital Status0.001  0.387  
  1 Married(n=1963) 
  2 De facto0.0631.450.98-2.160.3161.290.78-2.13
  3 Single/Other<0.0012.291.46-3.590.5760.810.40-1.67
Parity0.246  0.573  
  1 primiparious(n=1794)
  2 1 child0.3240.810.54-1.230.7060.910.54-1.52
  3 2 children0.8711.040.64-1.700.2961.390.75-2.66
  4 ≥ 3 children0.1741.520.83-2.790.6861.180.53-2.64
Recent life events      
  1 No(n=1943)
  2 Yes0.0051.611.16-2.260.4140.830.54-1.29
Past childhood abuse      
  1 No(n=1871)
  2 Yes0.0041.971.24-3.110.0991.700.91 -3.21
Relationship partner<0.001  0.545  
  1 No partner(n=1867)
  2 Poor0.4341.520.54-4.300.9981.000.25-3.99
  3 Moderate0.3101.530.67-3.510.5801.370.45-4.15
  4 Good0.0290.440.21 -0.920.7600.850.30-2.41
Social Support      
  1 No(n=1830)
  2 Yes0.0080.420.22-0.790.7820.880.34-2.25

However, in order to predict PND, it is the results of the multiple predictor analysis that are significant. It appears the only variables needed to be included in a model to predict PND are: past psychiatric history, antenatal EPDS score and, perhaps, residence. This suggests when controlling for antenatal EPDS and past psychiatric history, the influence of marital status, recent life events, past childhood abuse, relationship with partner and social support are no longer significant.

Binary logistic regressions with forward and backward selection of variables confirmed this as the case. Residence (urban/rural), past psychiatric history and antenatal EPDS score had a significant relationship with PND (Table 4). However, when these variables were considered individually, residence was not significant (p=0.166) suggesting it was only when past psychiatric history and antenatal EPDS were statistically controlled that residence was a significant predictor of PND. An interaction effect between residence and past psychiatric history is suggested (z=2.29, p=0.022) by the finding that PND is more common in rural rather than urban settings but only for people with a past psychiatric history – 20.8% of rural mothers with a psychiatric history experienced PND compared to 11.8% of urban mothers. For mothers without a psychiatric history, the chance of PND is slightly higher for urban (6.4%) rather than rural (4.7%) settings.

Table 4.  Significant variables for predicting PND.
 Single PredictorMultiple Predictors
 p-valueOR95% CIp-valueOR95% CI
  1. Note:

  2. (a) 1,968 cases included; (b) 1,879 cases included; (c) 1,878 cases included

Residence
  1 urban
  2 rural0.166a1.270.91-1.780.015a1.681.10-2.55
Past psychiatric history
1 none
2 yes<0.001a3.692.60-5.21<0.001c2.041.36-3.05
3 missing0.419a0.790.40-1.550.8970.940.45-2.00
Antenatal EPDS score0.001b1.271.22-1.32<0.001c1.271.22-1.32

As a result of this interaction, different binary logistic regression models are required in order to predict PND in the urban and rural cohorts. Table 5 shows the significant predictor variables when separate forward binary logistic regressions are fitted for urban and rural mothers. It suggests for urban mothers, the antenatal depression score is the best predictor of PND (OR=1.27; 95% CI 1.20–1.34). For rural mothers, SES and psychiatric history also have a significant influence.

Table 5.  Factors associated with depression for urban and rural women.
 Single PredictorMultiple Predictors
 p-valueOR95% CIp-valueOR95% CI
  1. Notes:

  2. (a) 906 cases included.

  3. (e) 967 cases included.

  4. (b) 973 cases included.

  5. (f) 1,060 cases included.

  6. (c) 1,058 cases included.

  7. (g) 967 cases included.

  8. (d)  1,052 cases included.

Urban
  Antenatal EPDS score<0.001a1.271.20-1.34<0.001a1.271.20-1.34
Rural
  Antenatal EPDS score<0.001b1.301.23-1.37<0.001e1.261.20-1.34
Past psychiatric history
  1 none
  2 yes<0.001c5.323.41-8.32<0.001e3.131.87-5.23
Past depression<0.001f4.112.60-6.510.008g2.121.22-3.70
Past Anxiety<0.001f4.191.89-9.280.021g2.901.18-7.13
Past Childhood Abuse< 0.001f5.572.84-10.910.013g2.931.25-6.88
SES0.040d0.950.90-1.000.025e0.930.87-0.99

It is interesting to note the impact of the antenatal EPDS score is similar for both cohorts, especially when the effects of past psychiatric history and SES are statistically controlled for rural mothers.

The model for the rural cohort in Table 5 shows that independent of antenatal EPDS and past psychiatric history, the odds of PND decreased for each one-unit increase in SES (OR=0.95; 95% CI 0.90 – 1.00). When antenatal EPDS and past psychiatric history were controlled statistically, this decline increased (OR=0.93; 95% CI 0.87 – 0.99) for each additional SES unit. When SES and antenatal EPDS were ignored, past psychiatric history significantly increased the likelihood of developing PND (OR=5.32; 95% CI 3.41 −8.32) and even when the effects of SES and antenatal SPSS were statistically controlled, past psychiatric history still had a pronounced influence (OR=3.13; 95% CI 1.87 – 5.23).

In confirmation of the above model for rural mothers, Table 5 also shows the results when ‘past depression’, ‘past anxiety’ and ‘past childhood abuse’ are included in the model instead of ‘past psychiatric history’. As before, when the other predictor variables were controlled, on average one unit increase in antenatal EPDS score increased the odds of PND (OR=1.27; 95% CI 1.20-1.34) while each one unit increase in SES decreased the odds of PND (OR=0.92; 0.86 – 0.98). When the effect of the other predictor variables was ignored, past depression (OR=4.11; 95% CI 2.60 −6.51), past anxiety (OR=4.19; 95% CI 1.89 – 9.28) and childhood abuse (OR=5.57; 95 CI% 2.84 – 10.91) were all significant risks but it is important to note the wide 95% confidence internal. When the other predictors were controlled these odds ratios dropped to 2.12, 2.90 and 2.93 respectively, but again still retaining the wide 95% confidence intervals.

Discussion

Research comparing PND in urban and rural populations has been limited within both the Australian and international context. This present study compares, in an Australian population, the contribution of demographic and psychosocial variables on the prevalence, and risk for, development of PND in urban and rural women.

The results presented here demonstrate the prevalence of antenatal depression was more common in the urban than rural cohort, while there was no significant difference in rates of PND. Logistic regression analysis indicates rural residence, past psychiatric history and antenatal EPDS score were significant variables for predicting PND for the overall sample. However, when different regression models were assessed for the individual cohorts, antenatal EPDS score only held as the best predictor for urban mothers, while for rural mothers antenatal EPDS, SES and past psychiatric history remained significant.

It is somewhat difficult to understand why antenatal depression was more common in the urban women compared to their rural counterparts, but this was not reflected in the rates of PND. This may be due to the fact that, antenatally, rural women believed they would receive adequate support upon returning home after delivery, resulting in lower levels of anxiety about how they would cope postnatally Alternatively, differences in the impact of variables related to satisfaction with obstetric/immediate postnatal care might explain this finding.

The finding that past psychopathology,4,14–17 antenatal EPDS scores18–20 and low SES21,22 are strongly predictive of PND is in-keeping with the findings of others. However, it is interesting to note in this model, social support appears to have no predictive effect. This is in contrast from observations that poor/unsatisfactory or absent social supports is a risk factor for PND.21–23 This result should be viewed cautiously in this context as the survey item used may have inadequately assessed the concept of ‘social support’. We propose further research specifically examining the efficacy, as well as the form of social support, and their interaction with place of residence and mental health status.

As noted earlier, it has been reported that women living outside the greater metropolitan area of Melbourne have significantly lower odds of depression than those living within the metropolitan area2,5. The results of this study, however, suggest the opposite; for the overall sample, rural residence was significantly associated with the presence of PND. It is important to highlight that reports describing the methodology of these earlier studies24 did not provide any demographic details on the rural women aside from stating women living in rural areas constituted approximately 25% of respondents, and the residential area was either ‘metropolitan’ or ‘other’. Later reports25 also refer to women as living in either ‘metropolitan’ or ‘non-metropolitan’ areas. As there is no explanation as to how ‘other’, ‘non-metropolitan’ or ‘metropolitan’ were defined or to describe from which localities women were recruited, in comparison to the detailed methodology employed in this study, it is impossible to confidently make a direct comparison between the different findings.

Some caution is advised because due to the nature of participant recruitment, the women sampled may not be truly representative of their respective populations. The methods employed to collect data from the rural cohort, and the depression-screening protocol involving the urban cohort,8 may have not all resulted in all potential pregnant women participating and/or being included in the postnatal follow-up. Despite this, the present study had a large rural sample size of over 1,000 women compared to the smaller numbers involved in previous studies. This improved the power to determine relationships between independent variables and PND. Furthermore, by not restricting the sample to only married, primiparous or English speakers gave diversity to the sample that better reflects the population from which it was drawn.

Women's relationship with their partner is of interest. Due to the difference in wording, the variable concerning the relationship with the partner was only tested in the separate urban and rural regression models. It was not a significant factor. This could be because the majority of women (87%) reported ‘good’ relationships, resulting in an under-representation of women with marital conflict, which has been demonstrated to increase the risk of PND.14,15,26,27 Another possibility was the questionnaire was completed with the partner present – unfortunately, we are unable to confirm whether this was the case. However, others have suggested asking women about sensitive health issues, for example domestic violence, in the presence of a partner might compromise the responses provided.28 Thus, a questionnaire to be completed in private for both urban and rural women should be used to explore this notion further.

The SES estimate was necessitated by the lack of information on education level and income of rural women. This estimate assumes that residents are homogeneous in education, income and occupation within a given suburb, and does not distinguish between those who own or rent property. For a rural area consisting of few suburbs, women would be allocated to a limited range of SES values. A rural community may consist of a diversity of rural residents, thus postcode may not be the best indicator of SES. Therefore, perceived financial stress or perceptions of financial adequacy, rather than income level, may be a better measure of the impact of socio-economic factors on maternal mental health outcomes.

Along similar lines, a measure of accessibility to health-care services and opportunities for social interaction, such as that assessed by the Accessibility/Remoteness Index of Australia (ARIA) may be a better indicator to define living circumstances29 and may better reflect differences in access to services such as health and education between those living in, and those living outside, major metropolitan areas.

This study did not evaluate the availability of appropriate services for women within the different geographic regions, nor the incentives or barriers to uptake-and-use of these services. Its aim was merely to highlight any differences in prevalence of PND and/or the association with demographic/psychosocial risk factors. We would hope that this paper provides the impetus and evidence to service providers within these communities to consider the importance of mental health issues in the perinatal period. We contend that it is difficult to assume how many women would require intervention – a score on a screening item provides no indication of morbidity, nor the behaviours that influence selection of an appropriate intervention. We also contend that investigation of this issue would require a detailed evaluation, outside the scope of this study, of the women with elevated EPDS scores and their presentation (or not) to particular services and supports and the decision making process that led to this presentation.

PND is an important issue regardless of one's place of residence. This study found no significant differences in the prevalence of PND in urban and rural women, although significantly more urban women experienced antenatal depression. This study also reiterates the contribution of established risk factors, such as past psychopathology and antenatal EPDS score, to the development of PND. Identification of psychosocial and socio-economic risk factors may help facilitate early intervention preventing development of postnatal mood disorders or limiting their long-term impact. The introduction of procedures, which identify and assess these factors, should be promoted. This finding provides impetus for further studies investigating the contribution of variables such the efficacy and form of social support, and the accessibility to perinatal health services, on the development of PND in rural women.

Acknowledgments

We gratefully acknowledge the assistance of Ms Jennie Ericksen and Prof Jeannette Milgrom, Parent Infant Research Institute (PIRI), Austin Health, Heidelberg Victoria, Australia.

We also thank the Northern and Angliss Hospitals for allowing collection of the urban data and the Wodonga Regional Health Service for the sharing of their antenatal and postnatal data with us.

Funding for this study was provided by beyondblue: the national depression initiative, as part of their commitment to the beyondblue National Postnatal Depression Program 2001-2005.

Ancillary