Maternal depression from early pregnancy to 4 years postpartum in a prospective pregnancy cohort study: implications for primary health care

Authors

  • H Woolhouse,

    Corresponding author
    1. Healthy Mothers Healthy Families, Murdoch Childrens Research Institute and Royal Children's Hospital, Parkville, Vic., Australia
    • Correspondence: H Woolhouse, Healthy Mothers Healthy Families, Murdoch Childrens Research Institute, Royal Children's Hospital, Flemington Road, Parkville, Vic. 3052, Australia. Email hannah.woolhouse@mcri.edu.au

    Search for more papers by this author
  • D Gartland,

    1. Healthy Mothers Healthy Families, Murdoch Childrens Research Institute and Royal Children's Hospital, Parkville, Vic., Australia
    Search for more papers by this author
  • F Mensah,

    1. Clinical Epidemiology and Biostatistics Unit, Murdoch Childrens Research Institute and Royal Children's Hospital, Parkville, Vic., Australia
    2. Department of Paediatrics, University of Melbourne, Melbourne, Vic., Australia
    Search for more papers by this author
  • SJ Brown

    1. Healthy Mothers Healthy Families, Murdoch Childrens Research Institute and Royal Children's Hospital, Parkville, Vic., Australia
    2. General Practice and Primary Health Care Academic Centre, University of Melbourne, Melbourne, Vic., Australia
    Search for more papers by this author

Abstract

Objective

To describe the prevalence of maternal depression from pregnancy to 4 years postpartum, and the risk factors for depressive symptoms at 4 years postpartum.

Design

Prospective pregnancy cohort study of nulliparous women.

Setting

Melbourne, Australia.

Sample

In all, 1507 women completed baseline data in pregnancy (mean gestation 15 weeks).

Methods

Women were recruited from six public hospitals. Questionnaires were completed at recruitment and 3, 6, 12 and 18 months postpartum, and 4 years postpartum.

Main outcome measures

Scores ≥13 on the Edinburgh Postnatal Depression Scale were used to indicate depressive symptoms.

Results

Almost one in three women reported depressive symptoms at least once in the first 4 years after birth. The prevalence of depressive symptoms at 4 years postpartum was 14.5%, and was higher than at any time-point in the first 12 months postpartum. Women with one child at 4 years postpartum were more likely to report depressive symptoms at this time compared with women with subsequent children (22.9 versus 11.3%), and this association remained significant in adjusted models (Adjusted odds ratio 1.71, 95% confidence interval 1.12–2.63).

Conclusions

Maternal depression is more common at 4 years postpartum than at any time in the first 12 months postpartum, and women with one child at 4 years postpartum report significantly higher levels of depressive symptoms than women with subsequent children. There is a need for scaling up of current services to extend surveillance of maternal mental health to cover the early years of parenting.

Introduction

In light of mounting evidence regarding the importance of maternal mental health to child health outcomes,[1-5] it is surprising that virtually all studies of maternal mental health have restricted their focus to the perinatal period (i.e. pregnancy and the first 12 months postpartum).[6] It has long been considered that women are at an increased risk of depression during this time, in large part because of the significant changes associated with the birth of a first child.[7-9] Comparatively, we know very little about the prevalence and consequences of maternal depression after the first 12 months postpartum. The fact that few studies examine maternal mental health beyond the perinatal timeframe contributes to the view that depression is most common in pregnancy and soon after giving birth. Counter to this, two large longitudinal mother and baby cohort studies—one from Australia and one from Norway—report a higher prevalence of maternal depression at 18 months postpartum[10] and 5 years postpartum[11] than during the immediate postpartum period. Another US-based study found the prevalence of depressive symptoms was highest in the month after birth (26%), and then remained stable at around 15% in assessments up to 36 months postpartum.[12] Interpreting the results of these studies is complicated by the fact that they do not take into account the timing and impact of subsequent pregnancies and births, which may trigger episodes of ‘perinatal depression’. There is also the probability of selective attrition over time in longitudinal studies, where women most at risk of depression are the most likely to be lost to follow-up, leading to under-estimation of the true prevalence of depression in the population, particularly at later follow-up points. Risk factors for maternal mental health problems (such as previous depression, intimate partner abuse, stressful life events and socio-economic disadvantage) have also been well examined in the perinatal period, but not in the context of longer-term maternal mental health.[6, 13-16]

To our knowledge, this is the first longitudinal study of maternal mental health that takes into account the impact of subsequent births, and the influence of selective attrition over time. The aims of the paper are: (1) to describe the prevalence of depressive symptoms from early pregnancy to 4 years postpartum in a nulliparous pregnancy cohort; and (2) to assess risk factors for maternal depression at 4 years postpartum, including subsequent pregnancies and births, relationship transitions, intimate partner violence and social adversity.

Methods

Study design

Women who were registered to give birth at six metropolitan public hospitals in Melbourne Australia (with a mix of high-risk and low-risk perinatal services) were recruited to the Maternal Health Study between 1 April 2003 and 31 December 2005. The eligibility criteria were as follows: nulliparity; ≤24 weeks of gestation at the time of completing the baseline questionnaire; ≥18 years; and sufficient proficiency in English to complete the study questionnaires. Follow-up questionnaires were completed at 3, 6, 12 and 18 months postpartum, and at 4 years postpartum. Eligible women were identified by hospital staff and mailed an invitation package after their first phone contact with the hospital to book in to give birth, including an invitation letter, an information sheet and consent form, the baseline questionnaire, and a reply-paid envelope. This was followed up by a mailed reminder postcard. Australian privacy legislation prevented the research team from having direct access to women's contact details, so this process was managed by hospitals on our behalf. The Maternal Health Study was approved by the relevant human research ethics committees in the following Melbourne institutions: La Trobe University, Royal Women's Hospital, Southern Health, Angliss Hospital and The Royal Children's Hospital. Further details regarding the study are available in a published study protocol.[17]

Outcome measures

The Edinburgh Postnatal Depression Scale (EPDS) was included in the baseline questionnaire and all follow-up questionnaires (3, 6, 12 and 18 months and 4 years postpartum). The EPDS is a ten-item self-report scale designed to identify women experiencing depressive symptoms in the postnatal period.[18] It has been validated for use in pregnancy,[19] and in nonpostnatal women.[20] A standard cut-off score of ≥13 is recommended when screening for probable major depression.[18, 21, 22]

Intimate partner abuse was assessed at 12 months postpartum and at 4 years postpartum using the Composite Abuse Scale, a validated scale to assess intimate partner abuse.[23, 24] The short version of the scale contains 18 items of behaviours by a partner or ex-partner that constitute emotional or physical abuse. Examples of items include: ‘Slapped me’, ‘Told me I was ugly’, ‘Blamed me for causing their violent behaviour’. The scale provides data on the prevalence and type of intimate partner abuse over a 12-month period. For each item, women are required to note the frequency of a particular behaviour over the last 12 months, with response options of ‘never’, ‘only once’, ‘several times’, ‘once per month’, ‘once per week’, and ‘daily’, scored 0–5, respectively. Emotional abuse is indicated by a score of ≥3 for ‘emotional’ items, and physical abuse by a score of ≥1 for ‘physical’ items. Women were categorised as experiencing intimate partner abuse if their scores indicated either physical or emotional abuse.

Stressful life events and social health issues were assessed at 4 years postpartum using a 20-item measure drawing on items from the Pregnancy Risk Assessment and Monitoring Study.[25] Women were asked ‘Have any of the following things happened to you in the last 12 months?’ and asked to tick ‘yes’ or ‘no’ to the list of items, which included major life events such as moving house, separation or divorce, getting married, death or serious illness of a close friend or family member, as well as social health issues such as not having enough money to buy food, legal troubles, or serious conflict between family members.

Sociodemographic characteristics including maternal age, country of birth, education, relationship status and income were assessed in the baseline questionnaire. Relationship status and family income were re-assessed at subsequent follow ups. Relationship transitions over the study period were determined by examining women's report of relationship status at each of the follow-up time-points. A relationship transition was considered to have occurred if a woman went from married to single, living with a partner to single, single to married or single to living with a partner. Change in relationship status from living with a partner to married was not coded as a relationship transition.

Statistical analysis

Data were analysed using STATA version 12.[26] Very few variables in the data set had missing data levels of >2%. The exceptions were income and smoking in pregnancy. For example, at baseline, 8.6% of data was missing for income, and 3.3% was missing for smoking status. The missing data levels for depression, employment status and pension status at baseline were 1.3, 1.7 and 1.5%, respectively. At 4 years postpartum, 1102 women completed the follow-up questionnaire and missing data for key variables was as follows: depression, 0.5%; intimate partner abuse, 0.9%; income, 6.3%; and child internalising and externalising behaviours, 0.7%. To assess representativeness of the sample, we compared participant data with routinely collected Victorian Perinatal Data for nulliparous women giving birth as public patients during the recruitment period. To account for selective attrition, multiple imputation of missing data was conducted.[27] The model included variables associated with attrition, variables in the analysis model and variables significantly associated with the primary outcome variable (depression at 4 years). The imputation model included (1) baseline report of maternal factors including highest educational qualification, country of birth (Australian/overseas), income, marital status, smoking in pregnancy; (2) index child and family factors including number of children in the family, the index child's sex, gestation and birthweight, the index child's physical health and emotional/behavioural difficulties at age 4; (3) repeated measures including intimate partner violence, maternal depressive symptoms and maternal physical health. Forty data sets were imputed using chained equation modelling.[27] The study analyses were conducted for the participants with complete data and repeated including the full cohort with multiple imputation of missing data to assess the robustness of the observed findings for the complete cohort. The two approaches produced similar results, therefore the results are presented for the full cohort with multiple imputation of missing data.

Univariable logistic regression was used to examine the association between depressive symptoms and a variety of maternal and family factors at 4 years postpartum. We hypothesised that women with a previous history of depression, women with two or more children at 4-year follow up, women who had experienced intimate partner abuse and women experiencing greater social adversity would be more likely to report depressive symptoms at the time of 4-year follow up. Data are presented as unadjusted odds ratios (OR) with 95% confidence intervals (95% CI).

Five multivariable logistic regression models were conducted to explore the association between number of children (focal variable of main interest) and depressive symptoms at 4 years postpartum (outcome variable). Factors to control for were identified a priori and included in the models as follows: model one adjusted for maternal age at baseline; model two additionally adjusted for intimate partner abuse over the study period; model three additionally adjusted for relationship transitions over the study period (early pregnancy to 4 years postpartum); model four additionally adjusted for two measures of social adversity (family income at 4 years postpartum, and stressful life events over the past 12 months); and model five, additionally adjusted for previous report of depressive symptoms (in either pregnancy or the first 12 months postpartum).

Results

Participants

Of 1537 enrolled women, 30 were excluded due to ineligibility. The final sample comprised 1507 eligible women. Exact response figures are not possible to calculate as hospitals identified participants and mailed invitation packages on behalf of the research team. This meant we were unable to assess wrongly addressed mail, duplicate mailings, returns to sender and mail-outs to non-eligible women. Over 6000 invitation packages were distributed. If we assume that 80–90% of invitations were correctly addressed and mailed to eligible participants, we estimate the final response fraction to be in the range of 28–31%. However, this is a conservative estimate, and it may have been higher. The mean gestation at enrolment was 15 weeks (SD 3.1, range 6–24 weeks).

The representativeness of the Maternal Health Study sample was assessed by comparing participant characteristics with routinely collected perinatal data for all nulliparous women over 18 years of age who gave birth in public hospitals in Victoria during the period of recruitment. The sample was representative in terms of method of birth and infant birthweight. However, the cohort included fewer young women (18–24 years, 14.1 versus 29.8%), and fewer women born overseas of non-English speaking background (16.2 versus 21.5%). Maternal age at birth ranged from 18 to 50 years, with a mean age of 30.9 years (SD 4.8 years). Most women were living with their partner (60.7% married and 34.6% cohabiting), and most were Australian-born (74.4%) and tertiary educated (72.1%). The potential for cluster effects associated with hospital of recruitment was examined in regression analyses investigating the outcomes of depressive symptoms at multiple time-points, and 12-month-period prevalence of intimate partner violence. No significant changes to odds ratios were found with the addition of birth hospital to multivariable regression models. Further details on the characteristics of the study participants can be found in previous publications.[28-34]

Participation in follow up

Sample retention in phase 1 of the study (up to 18 months postpartum) was 1431 (95%), 1400 (93%), 1357 (90%) and 1327 (88.1%) at 3, 6, 12 and 18 months postpartum, respectively. In all, 1345 women consented to taking part in Phase 2, which involved follow up at 4 years postpartum, with 1102 of these women (83.4%) returning the 4-year follow-up questionnaire. At the time of the last follow up, the index children ranged in age from 4.4 to 5 years, with a mean age of 4.5 years (SD 0.7). There was selective attrition between the baseline questionnaire in early pregnancy and the follow-up questionnaire at 4 years postpartum. Women completing the 4-year follow-up questionnaire were more likely to be older, Australian-born, tertiary-educated, not on a government benefit as their main source of income, and less likely to have reported intimate partner abuse or depressive symptoms in the first year following birth (data not shown). Differential attrition meant that women with depressive symptoms were more likely to miss future stages of follow up. Data imputation was conducted to give a more representative picture of the prevalence of depressive symptoms over time (estimating the rates that would have been observed if the whole sample had completed each stage of follow up). Study analyses were initially conducted using complete case analysis, and then re-conducted with multiply imputed data to check for concurrence of the results. Imputed data analyses supported the complete case analyses, with very similar findings. Comparisons over time, and associations between variables were alike, but slightly different prevalence estimates were observed (i.e. point prevalence of depressive symptoms) at 4 years was 11.5% using complete case analysis, and 14.5% using multiply imputed data. Given that the imputed data are likely to give a more representative picture of the prevalence over time, the results in this paper are presented using multiple imputation data (n = 1507).

Prevalence of depressive symptoms

Period prevalence was calculated by combining data from specific time-points. Almost one in three women (31.4%) reported depressive symptoms (EPDS ≥13) at least once over the period from early pregnancy to 4 years postpartum. One in five women (22.5%) reported depressive symptoms in early pregnancy and/or the first 12 months postpartum (i.e. during the perinatal period). Among the women reporting depressive symptoms over the period from early pregnancy to 4 years postpartum, 53.5% scored ≥13 on the EPDS on a single occasion, 20.1% on two occasions, 11.5% on three occasions and 14.9% on four or more occasions. Less than 1% of women reported depressive symptoms at every follow-up point.

The point prevalence of depressive symptoms at each follow-up point is reported in Figure 1. The highest prevalence of depressive symptoms (14.5%; 95% CI 14.2–14.8) was at 4 years postpartum and the lowest at 3 months postpartum (8.1%; 95% CI 7.9–8.3). As the confidence intervals do not overlap, this can be considered a statistically significant difference. Among women reporting depressive symptoms at 4 years postpartum, 59.5% had previously reported depressive symptoms: 27.2% had reported depressive symptoms in early pregnancy and 48.1% on at least one occasion in the first 12 months postpartum.

Figure 1.

Point prevalence of depressive symptoms (EPDS ≥13) at each follow up (n = 1507; all values based on multiple imputation of 40 data sets to account for missing data).

Depressive symptoms and subsequent children at 4 years postpartum

To assess the impact of subsequent births on the prevalence of depressive symptoms, we conducted analyses stratified by number of children. Figure 2 shows the prevalence of depressive symptoms at each follow up distinguishing between women with one child (28.2%) and those with two or more children (71.8%) at the time of the 4-year follow up. Women with one child at the time of the 4-year follow up reported approximately double the prevalence of depressive symptoms at every time-point compared with women with two or more children. At 4 years postpartum, 22.9% of women with one child reported depressive symptoms compared with 11.2% of women with two or more children (unadjusted OR 2.34; 95% CI 1.63–3.37).

Figure 2.

Point prevalence of depressive symptoms (EPDS ≥13) over the study period by number of children at 4 years postpartum (n = 1507; all values based on multiple imputation of 40 data sets to account for missing data). Values provided are point prevalence and 95% confidence intervals.

Risk factors for depressive symptoms at 4 years postpartum

Table 1 reports the relationship between social characteristics, previous depressive symptoms, relationship transitions, intimate partner abuse, stressful life events/social health issues, and depressive symptoms at 4 years postpartum. The strongest predictor of depressive symptoms at 4 years postpartum was having previously reported depressive symptoms either in early pregnancy, or in the first 12 months postpartum. Other factors associated with depressive symptoms were: young maternal age at baseline (18–24 years), stressful life events/social adversity in the year before the 4-year follow up, intimate partner violence and low income. Exposure to intimate partner abuse in the first 12 months postpartum or in the year before follow-up at 4 years postpartum was associated with a four-fold increase in the odds of reporting depressive symptoms at 4 years postpartum (unadjusted OR 4.09; 95% CI 2.82–5.92). Women who reported intimate partner abuse at 12 months postpartum, but not 4 years postpartum, did not have significantly raised odds of reporting depressive symptoms at 4 years postpartum.

Table 1. Variables associated with depressive symptoms at 4 years postpartum (n = 1507a)
 Prevalence of depressive symptoms (EPDS ≥13) at 4 years postpartum (14.5% of overall sample)
%95% CIUnadjusted OR (95% CI)
  1. a

    All values based on multiple imputation of 40 data sets to account for missing data.

  2. Significant differences are presented in bold.

Maternal age at baseline
18–24 years21.915.0–28.8 2.00 (1.2–3.3)
25–29 years13.19.4–16.71.08 (0.7–1.6)
30–34 years12.29.2–15.3 1.0 ref
≥35 years15.810.4–21.21.35 (0.8–2.2)
Country of birth
Australia14.111.7–16.61.0 ref
Overseas15.711.0–20.51.13 (0.8–1.7)
Maternal education level at baseline
Tertiary13.811.1–16.61.0 ref
Year 12 or less16.411.7–21.01.22 (0.8–1.8)
Depressive symptoms in pregnancy
No (EPDS <13)11.69.3–13.9 1.0 ref
Yes (EPDS ≥13)44.834.4–55.2 6.19 (3.9–9.8)
Depressive symptoms in first 12 months postpartum
No (EPDS <13)9.27.2–11.2 1.0 ref
Yes (EPDS ≥13)38.431.1–45.8 6.15 (4.3–8.8)
Number of children at 4 years postpartum
Two or more children11.28.8–13.7 1.0 ref
One child22.917.8–28.0 2.34 (1.6–3.4)
Relationship transitions (early pregnancy to 4 years postpartum)
No transitions11.69.3–13.9 1.0 ref
Transitions28.621.4–35.8 3.04 (2.0–4.6)
Relationship status at 4 years postpartum
Living with a partner11.29.0–13.4 1.0 ref
Not living with a partner (incl. single/separated)33.024.7–41.3 3.89 (2.5–6.0)
Intimate partner abuse
Never8.96.8–10.9 1.0 ref
First year postpartum only12.95.4–20.41.51 (0.8–3.0)
4 years only31.722.1–41.3 4.77 (2.9–7.9)
First year and 4 years35.726.3–45.1 5.71 (3.5–9.2)
Any report of intimate partner abuse28.422.6–34.3 4.09 (2.8–5.9)
Family income at 4 years postpartum ($AUD)
>$100,0008.75.2–12.1 1.0 ref
$60,001–$100,00012.79.4–16.01.54 (0.9–2.6)
$40,001–$60,00014.49.2–19.6 1.78 (1.0–3.2)
$20,001–$40,00025.217.4–33.0 3.57 (2.0–6.3)
<$20,00027.715.6–39.8 4.03 (1.9–8.8)
Stressful life events/social adversity at 4 years
No stressful life events5.83.0–8.6 1.0 ref
One or two stressful life events10.77.7–13.7 1.96 (1.1–3.6)
Three or more stressful life events26.521.3–31.8 5.93 (3.3–10.6)

Contrary to our hypothesis, women with one child at the 4-year follow up had a more than two-fold increase in odds of reporting depressive symptoms at this time. In an additional series of univariable regression analyses exploring sociodemographic characteristics, we found that compared to women with subsequent children, women with one child at the 4-year follow up were more likely to have experienced relationship transitions (unadjusted OR 4.28; 95% CI 2.97–6.19), intimate partner abuse (unadjusted OR 2.16; 95% CI 1.61–2.90), and to have experienced a greater number of stressful life events and social health issues in the preceding 12 months (unadjusted OR 2.68; 95% CI 1.85–3.88). They were also more likely to have a low income (unadjusted OR 3.54; 95% CI 1.92–6.52), and to have experienced depression in pregnancy (unadjusted OR 1.96; 95% CI 1.29–3.00) and the first 12 months postpartum (unadjusted OR 1.99; 95% CI 1.43–2.78). Women with one child at 4 years postpartum were no more likely than women with two or more children to have experienced a miscarriage, pregnancy termination or stillbirth since the index birth (data not shown).

To explore these relationships further, we developed a series of multivariable logistic regression models with number of children as the variable of main interest and depressive symptoms at 4 years postpartum as the outcome variable (see Table 2). In each successive multivariable model, the adjustment for additional variables resulted in an attenuation of the association between depressive symptoms and number of children at 4 years. In the final model, the relationship between depressive symptoms and number of children at 4 years postpartum was attenuated but remained significant (adjusted OR 1.71; 95% CI 1.12–2.63). Other variables that remained significantly associated with depressive symptoms in the final model were: intimate partner abuse at 4 years; ongoing intimate partner abuse; three or more stressful life events in the past 12 months; and depressive symptoms in pregnancy or the first year postpartum.

Table 2. Multivariable regressions to examine the relationship between number of children (variable of interest) and depression (outcome of interest) at 4 years postpartum (n = 1507a)
 Depressive symptoms (EPDS ≥13) at 4 years postpartum
Unadjusted OR (95% CI)Model 1 Adjusted OR (95% CI)Model 2 Adjusted OR (95% CI)Model 3 Adjusted OR (95% CI)Model 4 Adjusted OR (95% CI)Model 5 Adjusted OR (95% CI)
  1. a

    All values based on multiple imputation of 40 data sets to account for missing data.

  2. Significant differences are presented in bold.

Number of children at 4 years
Two or more children 1.0 ref 1.0 ref 1.0 ref 1.0 ref 1.0 ref 1.0 ref
One child 2.34 (1.6–3.4) 2.36 (1.6–3.5) 2.13 (1.4–3.2) 1.94 (1.3–2.9) 1.84 (1.2–2.8) 1.71 (1.1–2.7)
Maternal age at baseline
18–24 years 2.00 (1.2–3.3) 2.03 (1.2–3.3) 1.57 (0.9–2.7)1.42 (0.8–2.5)1.27 (0.7–2.2)1.14 (0.6–2.1)
25–29 years1.08 (0.7–1.6)1.13 (0.8–1.7)1.12 (0.7–1.7)1.10 (0.7–1.7)1.00 (0.7–1.6)0.93 (0.6–1.5)
30–34 years 1.0 ref 1.0 ref1.0 ref1.0 ref1.0 ref1.0 ref
≥35 years1.35 (0.8–2.2)1.10 (0.7–1.8)1.03 (0.6–1.7)1.03 (0.6–1.7)0.96 (0.6–1.6)0.99 (0.6–1.7)
Intimate partner abuse trajectory
Never 1.0 ref   1.0 ref 1.0 ref 1.0 ref 1.0 ref
First year postpartum only1.51 (0.8–3.0) 1.18 (0.6–2.4)1.07 (0.5–2.2)0.97 (0.5–2.0)0.75 (0.4–1.6)
Four years only 4.77 (2.9–7.9)   4.49 (2.7–7.5) 4.19 (2.5–7.0) 3.32 (1.9–5.7) 3.48 (2.0–6.1)
First year and 4 years 5.71 (3.5–9.2)   4.87 (3.0–8.0) 4.32 (2.6–7.2) 3.13 (1.9–5.2) 2.18 (1.2–3.8)
Relationship transitions
No 1.0 ref   1.0 ref1.0 ref1.0 ref
Yes 3.04 (2.0–4.6)   1.53 (0.9–2.5)1.18 (0.7–2.0)1.10 (0.6–1.9)
Family income at 4 years ($AUD)
>$100,000 1.0 ref    1.0 ref1.0 ref
$60,001–$100,0001.54 (0.9–2.6)   1.24 (0.7–2.1)1.25 (0.7–2.2)
$40,001–$60,000 1.78 (1.0–3.2)    1.04 (0.6–2.0)1.07 (0.6–2.1)
$20,001–$40,000 3.57 (2.0–6.3)    1.53 (0.8–3.0)1.61 (0.8–3.3)
<$20,000 4.03 (1.9–8.8)    1.36 (0.6–3.2)1.44 (0.6–3.5)
Stressful life events/social adversity in past 12 months
No stressful life events 1.0 ref     1.0 ref 1.0 ref
One or two stressful life events 1.96 (1.1–3.6)    1.64 (0.9–3.0)1.55 (0.9–2.8)
Three or more stressful life events 5.93 (3.3–10.7)     3.16 (1.7–5.9) 2.53 (1.3–4.8)
Depressive symptoms in pregnancy or first postnatal year
No (EPDS <13) 1.0 ref      1.0 ref
Yes (EPDS ≥13) 6.04 (4.2–8.7)      4.30 (2.9–6.5)

Discussion

Main findings

In our study, almost one in three women reported depressive symptoms at least once between pregnancy and 4 years postpartum. Counter to the prevailing view that the perinatal period is a peak time of vulnerability to depression, the prevalence of depressive symptoms was higher at 4 years postpartum than any point in the first 12 months after birth. The high prevalence of depressive symptoms at 4 years postpartum was not explained by subsequent births (i.e. subsequent episodes of ‘perinatal depression’). In fact, women with one child had a more than two-fold increase in odds of reporting depressive symptoms at 4 years postpartum, compared to women with two or more children. This association was partially explained by the higher levels of social adversity experienced by this group (relationship transitions, intimate partner abuse, lower family income and higher number of stressful life events) compared to women with subsequent children.

Women who had experienced depressive symptoms in early pregnancy and/or the first 12 months postpartum were more likely to report depressive symptoms at 4 years postpartum. Other factors associated with depressive symptoms at this time were: intimate partner abuse, relationship transitions, low income, and reporting three or more stressful life events and social health issues in the 12 months before the 4-year follow up. Women who had experienced intimate partner abuse in the 12 months before the 4-year follow up were four times more likely to report depressive symptoms than women who had never experienced abuse.

Strengths and limitations

Strengths of the study include: recruitment of first-time mothers in early pregnancy, intensive follow up from early pregnancy to 18 months postpartum, and extension of the study to include follow up 4 years after the index birth, and very high levels of participant retention. Other features of the Maternal Health Study which make it unique, are the collection of data on exposure to intimate partner abuse in the first 12 months postpartum, and 4 years after a first birth.

As with all studies, there are some important limitations that should be taken into consideration when interpreting the results. While representative in terms of key obstetric variables of method of birth and infant birthweight, women recruited to the study were not wholly representative in terms of sociodemographic characteristics. Younger women, single women and women born overseas from a non-English-speaking background were under-represented. As a result, prevalence estimates for depression and intimate partner violence are likely to underestimate the true prevalence of these conditions, and the results presented here may not be wholly generalisable. Despite achieving excellent retention of study participants, especially in the first 12 months postpartum, our analyses identified selective attrition likely to result in under-ascertainment of depressive symptoms and intimate partner abuse. Multiple imputation of missing data was used to provide estimates for key covariates that were more representative of the original cohort. Although data were collected prospectively, it remains difficult to comment on causal pathways. Finally, the measures of relationship transitions used in the study may not have captured all relationship transitions, particularly ones occurring between the follow ups at 18 months and 4 years postpartum.

Interpretation

This is the first study to report the prevalence over time of maternal depressive symptoms from pregnancy to 4 years postpartum, in a large, prospectively recruited cohort of first time mothers, taking into account the impact of subsequent births. The findings attest to the extent of psychological morbidity affecting first-time mothers in the 4 years after a first birth. The fact that almost one in three first-time mothers reported depressive symptoms on at least one occasion from early pregnancy to 4 years postpartum, coupled with the finding that the prevalence of depressive symptoms was highest at 4 years postpartum, provide a compelling case for re-thinking current policy frameworks for maternal mental health surveillance. As reported in a previous publication,[14] more than half of women experiencing depressive symptoms in the first 12 months after birth did not report symptoms until 6 months postpartum or later. At 4 years postpartum, 40% of women reporting depressive symptoms had not previously reported depressive symptoms. Based on these findings, it is likely that current systems of maternal mental health surveillance in Australia and the UK (where guidelines focus on pregnancy and the early months after birth)[35-37] will miss more than half the women experiencing depression in the early years of parenting. In particular, women who have not had subsequent children may be especially vulnerable to ‘falling through the gaps’ as they will not be reconnected back into primary-care services such as Maternal and Child Health programmes. Recent calls to improve collaboration between mental health researchers, policy makers, and primary-care systems to ensure action that advances both maternal and child health, included recommendations for the integration of core mental health services within routine primary health care (i.e. antenatal and postnatal visits).[38] Our findings provide a strong argument that such integrated systems of maternal mental health surveillance should be extended beyond the initial perinatal period, to include the first 5 years of parenting, when there is a high risk of mental health problems.

The strong univariable association between depression at 4 years and relationship transitions over the study period was no longer significant in multivariable models. It is likely that the experience of intimate partner abuse (sometimes but not always associated with relationship transitions) exerts a more powerful influence on maternal mental health than relationship transitions per se. Indeed for some women, relationship transitions may have positive effects. The finding that a ceasing report of intimate partner abuse (at 12 months postpartum but not 4 years postpartum) was no longer significantly associated with depressive symptoms at 4 years, lends further support to the value of early intervention for families where partner abuse is present in the early postnatal period. The emergence of social health issues as a strong predictor of maternal mental health problems presents particular challenges for health professionals working with mothers, and being attuned to the wider context of women's lives must be an important component of all services aimed at improving maternal mental health.

Conclusion

Our findings indicate that maternal depression is more common 4 years after a first birth than at any time in the first 12 months postpartum. Women with one child at 4 years postpartum show higher levels of depressive symptoms than women with two or more children, a difference which is in part explained by greater levels of social adversity experienced by women with one child at this time. There is a need for the surveillance of maternal mental health to extend beyond the perinatal period, to encompass at least the first 4 years of parenting, and to incorporate a focus on social health. At a time when so much attention is given to the surveillance of child health, an increased focus on maternal health is warranted, particularly given the strong connections between maternal and child health outcomes.

Disclosure of interests

The authors have no potential conflicts of interest.

Contribution to authorship

All authors have significantly contributed to this article and approved the final version of the manuscript. HW was involved in data collection, conducted literature searches, completed data analyses and interpretation, and wrote the paper. SB was responsible for the study concept and design, data analysis and interpretation, and co-wrote the paper. DG was involved in data collection, analysis and interpretation of the data, and critical revisions of the manuscript. FM assisted with data analysis and interpretation, and revision of the manuscript. All authors accept responsibility for the paper as published.

Details of ethics approval

The Maternal Health Study was approved by the relevant human research ethics committees in the following institutions: La Trobe University (2002/38), Royal Women's Hospital, Melbourne (2002/23), Southern Health, Melbourne (2002-099B), Angliss Hospital, Melbourne, and The Royal Children's Hospital, Melbourne (27056A).

Funding

This work was supported by grants #199222, #433006 and #491205 from The National Health and Medical Research Council (NHMRC), an NHMRC Early Career Fellowship #1037449 (FM), a VicHealth Research Fellowship (SB), an ARC Future Fellowship (SB), an NHMRC Career Development Fellowship (SB), a grant from the Medical Research and Technology in Victoria Fund (ANZ Trustees) and Murdoch Childrens Research Institute research is supported by the Victorian Government's Operational Infrastructure Program. The funding organisations had no involvement in the conduct of the study, and the authors are independent of the funding sources. All authors had access to the study data and were responsible for the decision to submit the paper for publication.

Acknowledgements

We are extremely grateful to all of the women taking part in the study; to members of the Maternal Health Study Collaborative Group (Christine MacArthur, Jane Gunn, Kelsey Hegarty, Shaun Brennecke, Peter Wein and Jane Yelland) who contributed to the design of study instruments; to Susan Donath who has made contributions to data analysis decisions, and to members of the Maternal Health Study research team who have contributed to data collection and coding (Liesje Brice, Maggie Flood, Ann Krastev, Ellie McDonald, Kay Paton, Renee Paxton, Sue Perlen, Martine Spaull, and Marion Tait).

Ancillary