Is mode of delivery associated with postpartum depression at 6 weeks: a prospective cohort study

Authors


Dr W Sword, School of Nursing 3H48B, McMaster University, 1200 Main Street West, Hamilton, ON L8N 3Z5, Canada. Email sword@mcmaster.ca

Abstract

Please cite this paper as: Sword W, Kurtz Landy C, Thabane L, Watt S, Krueger P, Farine D, Foster G. Is mode of delivery associated with postpartum depression at 6 weeks: a prospective cohort study. BJOG 2011;118:966–977.

Objective  To examine the relationship between delivery mode and postpartum depression at 6 weeks following hospital discharge.

Design  A prospective cohort study.

Setting  Eleven hospitals in Ontario, Canada.

Sample  A total of 2560 women ≥16 years of age who delivered singleton, live infants at term.

Methods  Women completed a questionnaire in hospital and 74% (n = 1897) participated in a structured telephone interview 6 weeks after discharge. Additional data were extracted from labour and delivery records. Generalised estimating equations (GEEs) were used to investigate factors associated with postpartum depression.

Main outcome measure  Women were screened for depression at 6 weeks following hospital discharge using the Edinburgh Postnatal Depression Scale (EPDS). A score of ≥12 on the EPDS was used as a measure of the primary outcome, depression.

Results  Mode of delivery was not independently associated with postpartum depression, and did not factor into the main-effects model. The multivariable analysis identified 11 predictor variables for depression: young maternal age (OR 5.27; 95% CI 2.73–10.15); maternal hospital readmission (OR 3.02; 95% CI 1.46–6.24); non-initiation of breastfeeding (OR 2.02; 95% CI 0.99–4.11); good, fair, or poor self-reported postpartum health (OR 1.82; 95% CI 1.19–2.80); urinary incontinence (OR 1.79; 95% CI 1.06–3.03); multiparity (OR 1.59; 95% CI 1.22–2.08); low mental health functioning (OR 1.20; 95% CI 1.15–1.25); low subjective social status (OR 1.16; 95% CI 1.02–1.33); high number of unmet learning needs in hospital (OR 1.12; 95% CI 1.03–1.22); low social support (OR 1.06; 95% CI 1.03–1.09); and low physical health functioning (OR 1.03; 95% CI 1.003–1.055). An exploratory interaction model revealed that caesarean section was associated with higher odds of becoming depressed in Canadian-born women, but that in women born outside of Canada it was associated with a lower risk of becoming depressed.

Conclusions  Delivery mode had no significant impact on the development of postpartum depression in the main-effects model. However, it may interact with place of birth and other unmeasured factors to create a risk for depression.

Introduction

Depression in women during their childbearing years is a major public health concern.1,2 Because childbirth is a complex life event associated with numerous biopsychosocial changes, it may trigger psychiatric disorders in women with predisposing genetic or psychosocial vulnerabilities.3 Although the Diagnostic and Statistical Manual for Mental Disorders (DSM IV) specifies postpartum depression as occurring within 4 weeks of childbirth, researchers and clinicians have expanded the definition to include depression without psychotic features present within the first year following childbirth.4 A systematic review of the prevalence and incidence of perinatal depression indicated that as many as 19.2% of new mothers experience depression in the first 3 months after delivery, with 14.5% being new episodes of depression.5 Whereas the onset of postpartum depression is usually within the first 3–4 weeks following childbirth, some women do not experience the onset of depressive symptoms until later in the initial postpartum year.6

Postpartum depression can have profound effects not only on the woman, but also on her infant, partner, and other children. It adversely affects maternal–infant interaction,7 and women with depressive symptoms are more likely to discontinue breastfeeding earlier than non-depressed mothers.8 Common consequences of postpartum depression for the child include emotional and behavioural problems, cognitive delay, and low social competence.9–12 Partners of women with postpartum depression experience major disruptions in their lives and marital relationships, fear, confusion, uncertainty about the future, and risk mental health problems.13,14 Older children are affected in that they may be ‘pushed away’ by their mothers because women with depression have difficulty coping with more than one child.15 In addition, they often take on adult roles prematurely by assuming a care-taking role for siblings and the depressed parent.16

Although there is evidence that reproductive hormonal changes can trigger depression in a vulnerable subgroup of women,17 risk factors for postpartum depression are predominately psychosocial in nature.18 A meta-analysis of postpartum depression predictors conducted by Beck,19 and subsequently updated,20 revealed thirteen significant predictors of depression: social support, prenatal depression, life stress, marital relationship, depression history, infant temperament, socio-economic status (SES), childcare stress, self-esteem, unplanned or unwanted pregnancy, maternity blues, prenatal anxiety, and marital status. A more recent systematic review similarly found psychosocial factors and mental health history to be predictors of postpartum depression: the strongest predictors were depression in pregnancy, anxiety in pregnancy, stressful life events during pregnancy or the early postpartum period, low levels of social support, and previous history of depression.21

Less attention has been given to obstetric risk factors for postpartum depression, including mode of delivery. Carter, Frampton, and Mulder conducted a systematic review and meta-analysis that examined whether caesarean section delivery is associated with an increased risk of depression.22 They concluded that because most studies and the meta-analysis found no significant association, a link between caesarean section and postpartum depression has not been established. Another literature review of the relationship between mode of delivery and various indicators of postpartum health also found no association between delivery mode and postpartum depression.23 However, there is considerable diversity in findings across studies, which may be explained by the presence of unmeasured factors that may moderate the effect of delivery mode on postpartum mental health, and interactive effects of delivery mode and potential mediating variables.24 Although it has been suggested that a caesarean section operates as a risk factor for depression only if a woman is vulnerable for some other reason,22 studies of interaction effects are lacking.

The purpose of this article is to determine whether postpartum depression at 6 weeks following childbirth is associated with mode of delivery. The study design overcomes the methodological limitations of many previous studies, identified by Lobel and DeLuca,24 in that it has a large sample size, includes a comparison group of women who delivered vaginally, uses a cross-sectional prospective approach, and uses a measure of postpartum depression with established psychometric properties. The paper also presents the findings of an exploratory examination of interactions between mode of delivery and potential moderator variables.

Methods

The data for the analyses were collected as part of The Ontario Mother and Infant Study (TOMIS) III, which used a quantitatively driven sequential mixed-methods design. The study examined prospectively the relationship between delivery mode and postpartum depression. A detailed study protocol is available.25

Survey sample

Women were initially recruited from two hospital sites in each of the five geographic regions of Ontario. The sites included both community and teaching hospitals. Because one of the sites withdrew from the study early in the recruitment phase, another hospital in the same geographic area was added as a study site. The inclusion criteria for women to participate in the study were: ≥16 years of age; delivery of a live singleton infant; ≥37 weeks of gestation; assuming care of infant when discharged; competent to give consent; and can be contacted by telephone. Women were ineligible to participate if their infant required admission to a neonatal intensive care or special care nursery for more than 24 hours, or were unable to communicate in one of the four study languages (English, French, Chinese and Spanish).25

Sample size

The primary objective of the study was to test the null hypothesis that the proportion of postpartum depression is identical in two groups defined by mode of delivery: vaginal and caesarean section. Postpartum depression was treated as a binary outcome. The criterion for significance was set at α = 0.05. The Student’s t-test was two-tailed and the power was set at 80% to detect a minimal clinically important difference of 5% between the two groups (e.g. 10% versus 15%). A difference of <5% would not be of clinical or substantive significance. It was also a reasonable difference that could be expected in this type of research. These assumptions led to a sample size of 690 women per group (i.e. 1380 in both groups). These calculations were performed using power and precision™ software.26 The overall sample was inflated to account for the expected intraclass correlation (ICC) structure within a hospital,27 and for a 30% rate of attrition. The calculations were based on an ICC of 0.018 from previous studies. We selected a sample of 3774 to use, which corresponded to an average cluster size of n = 50.25

Data collection

Study participants completed a self-report mothers’ questionnaire in hospital that included questions about: birthweight and gender; infant feeding; infant health problems; obstetrical history; chronic health problems; pregnancy complications; mode of delivery; medical problems post-delivery; perceptions of adequacy of help and support at home; readiness for discharge; concerns related to self and infant; sociodemographics; and subjective social status. The latter was measured using both the SES ladder and community ladder of the MacArthur Scale of Subjective Social Status.28 Site research assistants extracted data from labour and delivery records, specifically gestation, use of forceps/vacuum extraction, integrity of the perineum, anaesthesia, excessive bleeding, and Apgar scores.

At 6 weeks following hospital discharge study participants were contacted to complete a structured telephone interview. The interview schedule incorporated the Edinburgh Postnatal Depression Scale (EPDS).4 The ten-item EPDS is a widely used measure of postpartum depression that has been translated into various languages, and has been tested in diverse cultures.29 Its psychometric properties are as follows: sensitivity 86%, specificity 78%, positive predictive value 73%, split-half reliability 0.88, and α coefficient 0.87.30 In addition, women were asked if they had ever been diagnosed with depression or had been depressed in the past, either prior to or during the current pregnancy.

The interview schedule also included the 12-item short-form health survey (SF-12).31 As reported in the protocol paper,25 the SF-12 captures eight dimensions of health-related functional status (physical functioning, physical role, emotional role, mental health, bodily pain, general health, vitality, social functioning), and yields scores for two summary measures: physical component summary and mental component summary. A single-item question adapted from the SF-12, ‘In general, would you say that since you delivered this baby your health is excellent, very good, good, fair, poor’, was used as a global assessment of maternal health. This type of question is commonly asked to measure self-perceived health status. A large body of international research has found it to be associated with specific health problems, health service use, and functional status, indicating psychometric robustness.32

Other instruments included in the interview were: the modified Duke–UNC Functional Social Support Questionnaire,33 used by the consortium for Longitudinal Studies of Child Abuse and Neglect,34 the Severity Index for Urinary Incontinence,35 and the short-form Incontinence Impact Questionnaire.36 The severity of faecal incontinence was assessed using questions adapted from Sandvik’s index,35 and with language taken from the Manchester Health Questionnaire.37 The impact of faecal incontinence was assessed using the Lifestyle Scale of the Fecal Incontinence Quality of Life Scale.38 The presence of sexual problems was determined using the pain/discomfort and satisfaction domains of the Female Sexual Function Index.39

Information on specific physical health problems was obtained by asking women to respond yes/no to a list of potential problems (e.g. exhaustion/extreme fatigue, frequent headaches or migraines, incision problems) developed from the postpartum health literature. Similarly, women were asked about things they would have liked to have learned more about in hospital (e.g. breastfeeding, signs of illness in infant, physical changes, and care of self) to determine unmet learning needs. Additional questions captured type of infant feeding, length of hospital stay, satisfaction with length of stay, hospital readmission, history of depression, employment status, and satisfaction with services in hospital and in the community. Finally, during the interview women gave details of the information recorded on a modified Ambulatory Health Care Record.40 This record included data on healthcare visits, tests and procedures, medication costs, purchases of supplies and equipment, and direct out-of-pocket costs.

Data analysis

The descriptive results are reported for each group: vaginal delivery and caesarean section delivery. Patient demographics and baseline characteristics are summarised using descriptive summary measures, expressed as mean and standard deviation (SD) or median (minimum–maximum) for continuous variables, and number (percentage) for categorical variables. Generalised estimating equations (GEEs)41 were used to investigate factors associated with postpartum depression, and for modelling depression as a binary outcome to assess the effect of the mode of delivery (vaginal versus caesarean section), adjusting for potential confounding factors. The GEE is a special generalised linear models technique for clustered or correlated data, and, as such, allows for the specification of a correlation structure between patients within a hospital. We specified an exchangeable correlation structure that assumes the same correlation between any two participants within a hospital or site.

To determine which variables to include in the multivariable model, we performed a series of bivariate analyses and considered any variable for inclusion if P ≤ 0.10. All variables included in the bivariate analyses were selected a priori based on the literature and clinical knowledge. Several of the factors that were significant at the bivariate level were eliminated from the multivariable model because they had a high level of colinearity with related variables. We also forced five variables into the model along with mode of delivery for an exploratory examination of interactions between delivery mode and potential moderator variables. The analysis was limited to five variables for model stability. The five variables were: any previous depression (yes versus no), social support (Duke–UNC Functional Social Support Questionnaire score), country of birth (mother born in Canada versus born outside Canada), exhaustion/extreme tiredness (yes versus no), and total household income (<$20 000 versus ≥$20 000). These five variables were selected a priori based on literature that supported these as risk factors for postpartum depression. It was hypothesised that each of these risk factors would interact with delivery mode when predicting postpartum depression. In addition to the main-effects model, we therefore considered an exploratory interaction model that examined the interactions between mode of delivery and the five variables.

All statistical tests were two-sided at the 0.05 level. For all models the results are expressed as odds ratio (OR), corresponding two-sided 95% confidence intervals and associated P values. P values are reported to four decimal places, with values less than 0.0001 reported as <0.0001. Further details on analysis methods, including an assessment of goodness of fit and model assumptions, can be found in the study protocol.25 All analyses were performed using sas v9.2.

Results

A total of 2560 women were recruited into the study, which represents 68% of our target sample of 3774. Five of the hospital sites were able to reach their quota sample, whereas the other hospitals experienced recruitment difficulties and did not reach their targets. One-third of participants (32.3%) had a caesarean section delivery, with 51.7% of caesarean sections being planned before labour (representing 16.7% of all births). Approximately one in five women (18.5%) reported a chronic health problem, and 14.2% reported a pregnancy complication. Nearly half of the women (49.5%) had experienced their first live birth. Mean gestation was 39.3 (± 1.2) weeks and the mean infant birthweight was 3478 (± 4731) grams.

The sociodemographic characteristics of the study participants are shown in Table 1. In comparing women who had a vaginal delivery with those who had a caesarean section delivery, there were statistically significant differences in age, marital status, total household income, ethnicity, and highest level of education, but not in whether they were born in or outside of Canada, or in the language spoken at home. The 6-week data were gathered from 74% of women enrolled into the study from all sites (n = 1897). There were statistically significant differences between women who participated in the 6-week interview and those who did not on all seven characteristics reported in Table 1. Women who completed this interview were: older, and were more likely to have been married or living in common law with a partner; have had higher family incomes; have been born in Canada; have reported their ethnicity as English, French, or Aboriginal Canadian; have spoken English or French at home; and have attained further education than that of high school.

Table 1.   Characteristics of the TOMIS III study participants (n = 2560)
CharacteristicVaginal delivery (n = 1733)Caesarean section delivery (n = 827)Total sample (n = 2560)
  1. *n is <2560 because of missing data for these variables.

Age in years (mean ± SD) (n = 2521)*30.6 ± 5.432.6 ± 4.931.3 ± 5.3
 n (%)n (%)n (%)
Marital status (n = 2542)*
Married1292 (75.1)678 (82.6)1970 (77.5)
Common-law/living with partner308 (17.9)103 (12.6)411 (16.2)
Separated/widowed/divorced18 (1.1)10 (1.2)28 (1.1)
Never married103 (6.0)30 (3.7)133 (5.2)
Household income (n = 2473)*
<$20 000187 (11.2)67 (8.3)254 (10.3)
$20 000–39 999193 (11.6)77 (9.6)270 (10.9)
$40 000–59,999273 (16.4)100 (12.4)373 (15.1)
$60 000–79 000265 (15.9)148 (18.4)413 (16.7)
≥$80 000751 (45.0)412 (51.2)1163 (47.0)
Born in Canada (n = 2545)*1221 (70.9)581 (70.6)1802 (70.8)
Self-reported ethnicity (n = 2544)*
English Canadian955 (55.5)430 (52.3)1385 (54.4)
French Canadian90 (5.2)36 (4.4)126 (4.9)
Aboriginal Canadian49 (2.9)9 (1.1)58 (2.3)
Chinese161 (9.4)68 (8.3)229 (9.0)
Other467 (27.1)279 (33.9)746 (29.3)
Language spoken at home (n = 2552)*
English1370 (79.3)664 (80.6)2034 (79.7)
French47 (2.7)10 (1.2)57 (2.2)
Chinese131 (7.6)56 (6.8)187 (7.3)
Other180 (10.4)94 (11.4)274 (10.7)
Highest level of education (n = 2547)*
Less than high school124 (7.2)25 (3.0)149 (5.9)
High school168 (9.7)63 (7.7)231 (9.1)
Some community college/technical school130 (7.5)59 (7.2)189 (7.4)
Completed community college/technical school400 (23.2)179 (21.8)579 (22.7)
Some university108 (6.3)52 (6.3)160 (6.3)
Completed university794 (46.1)445 (54.1)1239 (48.7)

At 6 weeks the rate of postpartum depression as determined by a score ≥12 on the EPDS was 7.6%. Bivariate analyses revealed that 34 of the 56 factors tested had statistically significant associations with postpartum depression (≤ 0.05) (Table 2). Mode of delivery (vaginal versus caesarean section) was not significantly associated with depression (OR 1.10, P = 0.6586).

Table 2.   Bivariate analysis of factors associated with postpartum depression at 6 weeks
FactorsOdds ratio*95% Confidence intervalP
  1. *Obtained from bivariate logistic regression adjusted for correlation among patients within site.

  2. **Decrease in one point on MacArthur Scale of Subjective Social Status (SES ladder).

  3. ***Decrease in one point on Duke–UNC Functional Social Support Questionnaire.

Data collected in hospital
Age <25 years2.01(1.25–3.21)0.0037
Not first pregnancy1.22(1.01–1.47)0.0360
Inadequate support at home1.68(1.35–2.11)<0.0001
Not ready for discharge1.69(1.12–2.56)0.0130
Health-related concerns1.22(1.10–1.36)0.0002
Language other than English or French spoken at home1.97(1.42–2.75)<0.0001
Born outside Canada1.60(1.20–2.12)0.0013
Not partnered2.14(1.02–4.53)0.0454
Annual household income <$20 0002.77(1.73–4.43)<0.0001
Highest education level achieved: high school1.62(1.22–2.15)0.0008
Low subjective social status**1.15(1.04–1.28)0.0057
Data collected at 6 weeks
Unmet learning needs in hospital1.23(1.15–1.31)<0.0001
Baby’s health good/fair/poor3.04(2.10–4.40)<0.0001
Cannot tell when baby is sick1.79(1.23–2.61)0.0025
Breastfeeding not initiated1.74(1.09–2.79)0.0202
Maternal hospital readmission3.75(2.18–6.45)<0.0001
Health before pregnancy good/fair/poor2.17(1.52–3.12)<0.0001
Health since delivery good/fair/poor4.35(3.09–6.13)<0.0001
Low SF-12 physical component score1.02(1.01–1.04)0.0067
Low SF-12 mental component score1.20(1.16–1.24)<0.0001
Physical health problems1.29(1.19–1.41)<0.0001
Low affective social support***1.35(1.29–1.41)<0.0001
Low confidant social support***1.27(1.23–1.32)<0.0001
Low instrumental social support***1.26(1.20–1.32)<0.0001
Low total social support***1.12(1.10–1.15)<0.0001
Urinary incontinence2.33(1.59–3.43)<0.0001
Unable to get help for a maternal physical health problem2.38(1.73–3.28)<0.0001
Unable to get care for a maternal mental health problem10.25(4.11–25.57)<0.0001
Rating of community health services fair/poor/didn’t use any1.63(1.18–2.25)0.0028
Rating of services in hospital during labour/delivery fair/poor2.40(1.00–5.77)0.0497
Rating of services in the community after discharge fair/poor1.94(1.24–3.03)0.0035
History of depression2.37(1.64–3.43)<0.0001
Depression in pregnancy5.66(2.59–12.36)<0.0001
Any previous depression2.89(1.88–4.43)<0.0001

The main-effects multivariable analysis showed 11 predictor variables to be significantly associated with postpartum depression (Table 3): maternal age <25 years; maternal hospital readmission; non-initiation of breastfeeding; good, fair, or poor self-reported postpartum health; urinary incontinence; multiparity (previous pregnancy); low mental health functioning; low subjective social status; high number of unmet learning needs in hospital; low social support; and low physical health functioning. Odds ratios were highest for maternal ages of <25 years and maternal hospital readmission.

Table 3.   Logistic regression (GEE) main effects model for postpartum depression at 6 weeks*
VariableAdjusted odds ratio**95% confidence intervalP
  1. *Of the 1897 women who completed the 6-week interview, 1758 (92.7%) were included in the multivariable analysis; 139 women were excluded because of missing data, with nine of these women having missing outcome data, and all but one of these women having missing data on at least one predictor variable.

  2. **Obtained from multiple-variable logistic regression adjusted for correlation among patients within site.

  3. ***Odds ratio associated with a decrease in one point on score.

  4. ****Decrease in one point of total score on the Duke–UNC Functional Social Support Questionnaire.

Mother’s age
≥25 yearsReference  
<25 years5.27(2.73–10.15)<0.0001
First pregnancy
YesReference  
No1.59(1.22–2.08)0.0006
MacArthur SES score1.16(1.02–1.33)0.0385
Number of unmet learning needs1.12(1.03–1.22)0.0115
Breastfeeding initiation
YesReference  
No2.02(0.99–4.11)0.0513
Maternal hospital readmission
NoReference  
Yes3.02(1.46–6.24)0.0030
Health since delivery
Excellent/very goodReference  
Good/fair/poor1.82(1.19–2.80)0.0060
SF-12 mental component score***1.20(1.15–1.25)<0.0001
SF-12 physical component score***1.03(1.003–1.055)0.0289
Urinary incontinence
NoReference  
Yes1.79(1.06–3.03)0.0297
Mode of delivery
VaginalReference  
Caesarean section1.06(0.61–1.85)0.8402
Low social support****1.06(1.03–1.09)0.0006
Exhaustion/extreme fatigue
NoReference  
Yes0.91(0.62–1.32)0.6108
Any previous depression
NoReference  
Yes1.66(0.97–2.84)0.0635
Total household income ≥$20 000Reference  
<$20 0000.76(0.31–1.89)0.5583
Country of birth
Outside CanadaReference  
Canada0.92(0.59–1.44)0.7220

All 11 statistically significant variables in the main-effects model remained in the exploratory interaction model. Only one of the interaction terms was statistically significant: delivery mode by country of birth (Canada versus outside Canada) (Table 4). For women born in Canada, those who had a caesarean section delivery had higher odds of being depressed than those who delivered vaginally (OR 1.389, < 0.0001). For women born outside of Canada, those who had a caesarean section delivery had lower odds of being depressed than those who had a vaginal delivery (OR 0.333, < 0.0001).

Table 4.   Logistic regression (GEE) interaction model for postpartum depression at 6 weeks
VariableAdjusted odds ratio*95% Confidence intervalP
  1. *Obtained from multiple-variable logistic regression adjusted for correlation among patients within site.

  2. **Odds ratio associated with a decrease in one point on score.

  3. ***Decrease in one point of total score on the Duke–UNC Functional Social Support Questionnaire.

Mother’s age
≥25 yearsReference  
<25 years5.73(3.06–10.76)<0.0001
First pregnancy
YesReference  
No1.61(1.22–2.11)0.0006
MacArthur SES score1.16(1.01–1.34)0.0437
Number of unmet learning needs1.12(1.03–1.22)0.0090
Breastfeeding initiation
YesReference  
No2.07(1.02–4.21)0.0449
Maternal hospital readmission
NoReference  
Yes3.04(1.45–6.35)0.0031
Health since delivery
Excellent/very goodReference  
Good/fair/poor1.88(1.23–2.88)0.0037
SF-12 mental component score**1.20(1.15–1.26)<0.0001
SF-12 physical component score**1.03(1.002–1.06)0.0344
Urinary incontinence
NoReference  
Yes1.85(1.08–3.16)0.0245
Mode of delivery
VaginalReference  
Caesarean section1.39(0.27–7.04)0.6920
Low social support***1.06(1.02–1.10)0.0010
Interaction with delivery mode1.01(0.97–1.05)0.6399
Exhaustion/extreme fatigue
NoReference  
Yes0.95(0.67–1.37)0.7976
Interaction with delivery mode0.77(0.28–2.12)0.6060
Any previous depression
NoReference  
Yes1.75(0.95–3.23)0.0714
Interaction with delivery mode0.88(0.50–1.53)0.6438
Total household income
≥$20 000Reference  
<$20 0000.60(0.15–2.34)0.4577
Interaction with delivery mode2.41(0.66–8.73)0.1820
Country of birth
Outside CanadaReference  
Canada1.48(0.91–2.44)0.1180
Interaction with delivery mode0.24(0.12–0.47)<0.0001

Discussion

We did not find an association between postpartum depression and mode of delivery at the bivariate level of analysis. We did, however, find that 34 other variables were associated with depression. Eleven of these variables were determined to be predictor variables in the main-effects model, and all 11 statistically significant variables remained in the exploratory interaction model. This interaction model suggested an interaction between mode of delivery and being born in Canada versus outside of Canada.

The rate of postpartum depression at 6 weeks in our study is similar to that reported in the Canadian Maternal Experiences Survey,42 which also used the EPDS, but with a cut-off score of 13 rather than 12. It is important to note that the EPDS is not a diagnostic instrument, but rather a tool that was developed to detect depression in the postpartum period; secondly, a positive score could reflect a transient mood change rather than a depressive illness.4 A cut-off score of 12 or 13 is typically used for ‘probable depression’, and as a marker for major depression, whereas a cut-off score of 9 or 10 is used for ‘possible’ or minor depression.43 A systematic review revealed ‘convincing diagnostic evidence’, based on summary positive predictive values, for the ability of the EPDS to detect both major depression and combined major and minor depression using a cut-off score of 12/13.43

Our finding of no association between postpartum depression and mode of delivery is consistent with other recent studies and a meta-analysis.22,44–46 Several studies that have reported an association between caesarean section and postpartum depression, or that have had mixed results, have been criticised for being methodologically weak, with small samples, retrospective designs, and poor-quality psychometric instruments.22,24 Many of these studies were conducted in the 1980s and early 1990s, a time when contextual factors associated with caesarean section birth may have resulted in women being psychologically more vulnerable to postpartum depression than today. Over the last decade changes in the medical care of women experiencing caesarean birth, such as the increased use of regional instead of general anaesthesia, which allows women to be more involved in their childbirth experience, and better postoperative pain management,47 have made caesarean birth less painful and less stressful.

In addition, recent shifts in societal and women’s attitudes away from the longstanding belief that a ‘normal birth’ is a ‘vaginal birth’ have made caesarean section delivery more acceptable.24,46,48 The shift is evident not only in the number of women requesting elective caesarean section,46 but also in the rising number of pregnant women who no longer expect to have a vaginal delivery.49 Given the increasing acceptability of caesarean section delivery, women are more likely to be satisfied with an operative delivery than was reported in a large meta-analysis of psychosocial outcomes that included studies from the 1970s to the early 1990s.50 Satisfaction with childbirth is an important outcome that is likely to decrease women’s psychological vulnerability to postpartum mood disorders.51

Whereas many variables in our study were found to be bivariately associated with postpartum depression, just 11 factors were found to be statistically significant in the multivariable models. Only one factor that has previously been identified as a strong predictor in three meta-analyses was identified as being predictive of postpartum depression in our multivariable model, that is, low social support.20,21 A number of other predictor variables in our model have been reported in studies of risk factors for postpartum depression, including young maternal age,52–54 having more than one child,52,54 self reporting of poorer health,55,56 and not breastfeeding.44,57 Interestingly, although a history of depression and perinatal depression and anxiety have commonly been found to be strong independent predictors of postpartum depression,20,21 they were not predictors in our multivariable model. However, a history of depression and perinatal depression were strong bivariate predictors of postpartum depression in our study.

Urinary incontinence as a predictor of postpartum depression has received little attention in the literature. Depression and urinary incontinence have only recently been linked in studies of postpartum women.58,59 In a small study (n = 149) using the EPDS, Hullfish and colleagues found that urge urinary incontinence was associated with depression at 6 weeks postpartum.58 They postulated that the dysregulation of serotonergic mechanisms alone or together with other neuropathways may explain the potential association of urge incontinence and depression; serotonergic activity enhances bladder storage through the inhibition of central pathways regulating maturation.58 It may also be that the stress associated with urinary incontinence contributes to depression.

No literature related to three of our multivariable model predictors of postpartum depression could be found: maternal readmission to hospital, high number of unmet learning needs in hospital, and low mental health functioning. Readmission to hospital was one of the strongest predictors of postpartum depression, with an OR 3.02 (95% CI 1.46–6.24), yet only two of the 40 maternal readmissions to hospital within the first 6 weeks of postpartum discharge were for postpartum depression. Readmission is a stressful life event during the postpartum period, and research has found that stressful life events pre- and postnatally are strong predictors of postpartum depression.20,21 As for unmet learning needs, a lack of information about postpartum transitions, self and infant care, and community resources are likely to heighten maternal anxiety and stress among women who are vulnerable to depression, thereby increasing the risk of psychiatric morbidity. In their randomised controlled trial, Ho and colleagues found that discharge education on postpartum depression was effective in lowering rates of postpartum depression at 3 months.60 It was not surprising that low mental health function was a predictor of postpartum depression. In a study to assess the screening utility of the SF-12 mental health component scale, Gill and colleagues reported this scale to be a valid measure of depression.61

We did not find low total household income to be a significant predictor of postpartum depression, contrary to the findings of several other studies.20,45,62 This may be because we used the limits <$20 000 versus ≥$20 000 as our measure of income, instead of the more sensitive low-income cut-offs set by Statistics Canada, which take into account the size of the family and the community,63 which are data that we did not collect. Segre and colleagues found that although poverty is an important predictor of postpartum depression, not all women in the lowest income category became depressed.54 They hypothesised that poverty alone is inadequate to explain the development of depressive symptoms, and that intermediary buffering factors, such as social support, can facilitate resilience to depression. A study of low-income families in Early Head Start programmes conducted by Malik and colleagues supports this tenet, as they found that maternal socio-economic variables were only indirectly associated with maternal depression.64

Few studies have examined interactions between delivery mode and potential effect modifier variables in predicting postpartum depression.24 By doing so our model provides a better understanding of the predictors of postpartum depression, although we do recommend exercising caution in interpreting the interaction findings, as these analyses are exploratory and may be the result of chance, given multiple testing. We found that for women born in Canada, caesarean birth was associated with an increased risk of depression compared with vaginal delivery, whereas for women born elsewhere, caesarean birth was associated with a decreased risk of postpartum depression compared with vaginal birth. It may be that immigrant women have different expectations of childbirth, or experience different degrees of social support, dependant on delivery mode, than women born in Canada. Clement pointed out that beliefs about caesarean section delivery differ by sociocultural group.51 She gave the example that in Brazilian culture, caesareans are esteemed and viewed positively, as they signify modernity. In a study in Nigeria, women most often reported choosing a caesarean section because it is perceived as the surest mode to have a live birth.65 For Canadian born women, although attitudes are shifting regarding caesarean section, a vaginal birth may still be the desired mode of delivery for many. Having a caesarean section may be associated with feelings of failure, lack of control, and reduced self-esteem, resulting in an increased risk for depression.57

Strengths of our study include our large sample of sociodemographically diverse women recruited from both community and teaching hospitals across the province of Ontario, Canada. Data were collected in four languages purposefully selected to promote the participation of the largest language groups represented at the hospital sites. The sociodemographic data indicate that the study sample is generally comparable with the Ontario, Canada, population, which enhances the generalisability of the findings. The participants are similar in age and education to women of childbearing age/giving birth in Ontario.42,66–70 The proportion of aboriginal women in our study (2.3%) is comparable with the 2.6% reported for Ontario.69 Thirty percent of women were born outside of Ontario; Statistics Canada reports a slightly lower rate of 24.1% for new mothers in Canada who were born outside of the country.71 The larger percentage of foreign-born women in our sample is not surprising given that Ontario is the ‘province of choice’ for over half of all newcomers to Canada.72

The caesarean section rate in our study sample (33.2%) is higher than the 28.4% reported in The Ontario Perinatal Surveillance System Report 2008.69 However, given that our data collection period continued for 1 year later than that of the Ontario report, the difference probably reflects a continuation of the rising trend in caesarean section rates in Canda.67 The rate of planned caesarean section (16.7%) in our study was only slightly higher than the Ontario and national rates reported in The Canadian Maternity Experiences Survey (14.1 and 13.5% respectively).42

However, the study is not without its limitations. We used a convenience sample of women who gave birth to generally healthy term infants who were discharged with their mothers. There is a potential selection bias in that we might have had an under-representation of women who had social risk factors, mood disorders and other morbidities, and who had experienced stressful life events. Our analysis included 68% (2580/3774) of the target sample size, with 75% (1897/2580) of the participants completing the 6-week interview with EPDS data. This may have limited our ability to fully test the null hypothesis of interest. Also, we know that there were statistically significant differences between women who completed the 6-week interview and those who did not on three of the variables included in the multivariable model (age, household income, born in/outside Canada). Overall, women who participated in this interview were likely to be older, to have lower household incomes, and to have been born outside of Canada than those who did not. As such, the findings of the multivariable analyses should be interpreted with caution. Finally, although we collected data on numerous independent variables, we lacked some potentially relevant measures that have been implicated in postpartum depression, such as quality of the marital relationship and stress.20,21

Conclusion

We found no association between postpartum depression and mode of delivery. Many of the factors that were found to be predictive of postpartum depression are consistent with other research findings. Low social support, young maternal age, previous pregnancy, poorer self-reported postpartum health, and not breastfeeding have commonly emerged as predictors of depression in other studies. However, this study revealed additional factors that have received little or no attention in the postpartum depression literature: high number of unmet learning needs, maternal readmission to hospital, and urinary incontinence. In addition to the commonly recognised risk factors for depression, these potentially important factors should also be acknowledged and considered as risk factors for depression.

Identified risk factors provide valuable ‘red flags’ that can alert healthcare providers to pay particular attention to the assessment of maternal mood. Screening or clinical assessment for possible postpartum depression is an important first step in early identification and intervention, which are important given that the length of time until obtaining adequate treatment is the most significant factor in the duration of depression.73 Untreated depression has negative implications for women, children, and families, yet many cases are undetected.74 The assessment of risk factors can easily be integrated into postpartum home visits by nurses and the routine 6-week postpartum maternal medical check-up. Screening for these factors should also be integrated into medical well-baby visits in the early months, as it has been suggested that the optimal screening time is between 2 weeks and 6 months.74

Future analyses of study data will determine whether risk factors for postpartum depression differ for women who have had a planned versus unplanned caesarean section. Although more adverse psychological outcomes have been reported after emergency caesarean section than after elective or planned caesarean section,75 we did not find an association between postpartum depression and whether the caesarean section was planned before labour. The findings of our exploratory interaction model are perplexing: women born in Canada were more likely to experience postpartum depression if they delivered by caesarean section, whereas women who were born elsewhere were more likely to experience postpartum depression if they delivered vaginally. More research is needed to tease out the reasons for these findings; qualitative approaches might be the most appropriate for enhancing our understanding of these differences. Also, there is a need for rigorous studies that are adequately powered to examine the interactive effects of caesarean section and a variety of potential moderator variables.

Disclosure of interests

None declared.

Contribution to authorship

WS, CLK, LT, SW, PK, and DF contributed to the study design. WS had overall project responsibility and took a lead role in writing the article. CKL and LT wrote sections of the manuscript. GF performed the statistical analysis under the supervision of LT. All authors contributed to the interpretation of the results, and have seen and approved the final draft of the article.

Details of ethics approval

This study was approved by the Hamilton Health Sciences/McMaster University Faculty of Health Sciences Research Ethics Board (17 February 2006, project number 05-455), and secondly by the research ethics board at each participating hospital.

Funding

This study was funded by the Canadian Institutes of Health Research MOP – 77594.

Acknowledgements

We thank all of the women who participated in the study and the many individuals who assisted with participant recruitment, data collection, and data management. We also thank the TOMIS III Advisory Committee (Donna Fedorkow, Amiram Gafni, Helen McDonald, Debbie Sheehan, and Russell Springate) for their guidance regarding the study design.

Ancillary