Article Text
PDF
Abstract
Background Low social support is consistently associated with postpartum depression. Previous studies do not always control for previous mental health and do not consider what type of support (tangible, emotional, informational or positive social interaction) is most important. The objectives are: to examine if low social support contributes to subsequent risk of depressive or anxiety symptoms and to determine which type of support is most important.
Methods Data from the All Our Families longitudinal pregnancy cohort were used (n=3057). Outcomes were depressive or anxiety symptoms at 4 months and 1 year postpartum. Exposures were social support during pregnancy and at 4 months postpartum. Log binomial models were used to calculate risk ratios (RRs) and absolute risk differences, controlling for past mental health.
Results Low total social support during pregnancy was associated with an increased risk of depressive symptoms (RR 1.50, 95% CI 1.24 to 1.82) and anxiety symptoms (RR 1.63, 95% CI 1.38 to 1.93) at 4 months postpartum. Low total social support at 4 months was associated with an increased risk of anxiety symptoms (RR 1.65, 95% CI 1.31 to 2.09) at 1 year. Absolute risk differences were largest among women with previous mental health challenges resulting in a number needed to treat of 5 for some outcomes. Emotional/informational support was the most important type of support for postpartum anxiety.
Conclusion Group prenatal care, prenatal education and peer support programmes have the potential to improve social support. Prenatal interventions studies are needed to confirm these findings in higher risk groups.
- pregnancy
- mental health
- social and life-course epidemiology
- cohort studies
- social capital
Statistics from Altmetric.com
Introduction
As many as 20% of women experience depressive symptoms or elevated anxiety in the first year postpartum, and these symptoms often co-occur.1–4 Maternal mental health problems can be debilitating for mothers, partners and for children, who are at increased risk of emotional and behavioural problems.5 In recent reviews, factors consistently associated with postpartum depression and anxiety include, but are not limited to: depression or anxiety during pregnancy, history of depression, early life adversity or stressful life events, parity, lower socioeconomic status, marital relationship, low social capital and low levels of social support.6–11
For the purposes of this paper, we define social support as the resources obtained from social relationships, primarily friends and family.12 This support can be in the form of emotional or informational support (ie, someone to ask advice of or talk to about problems), tangible support (someone to help with practical needs) or interaction support (someone to do fun things with).13
Previous studies have typically measured mental health symptoms and social support at the same time period.4 7 9 10 However, women who are already experiencing depressive symptoms may be less likely to socialise and may perceive that less social support is available. To understand the potential benefit of increased social support, it is important to assess this relationship prospectively.14
Considering that past mental health influences current mental health and that mental health symptoms such as anxiety and depression often co-occur, longitudinal studies are critical to understanding the effect of social support on current mental health independent of previous mental health.2 15 16 While some longitudinal studies fail to control for previous mental health altogether,17 other studies only control for either previous depression or anxiety,14 18 which does not account for the cumulative and clustering nature of mental health challenges.2 15 19 20 For example, a study of 877 women in Japan found that women with more people in their support networks during pregnancy had lower levels of postpartum depression, controlling for earlier depression symptoms, but this study did not control for earlier anxiety or history of depression.14
Qualitative evidence suggests that different ethnic or socioeconomic groups may prioritise different types of support.21 Some quantitative studies have examined types of support individually.22–24 For example, a study of first-time mothers in Ireland found moderate inverse correlations between several types of support (tangible, emotional and so on) and depressive symptoms; however, this analysis was cross-sectional and only examined each type of support in isolation.22 Because different support types overlap,13 comparing them at the same time can reveal whether one type is more important than another. For example, does a woman with low tangible support have a higher risk of postpartum depression regardless of availability of other types of support?
This study has two aims. First to understand if low levels of total social support contribute to subsequent risk of maternal mental health problems (depressive symptoms and elevated anxiety) at 4 months and 1 year postpartum. Second to determine which type of support (if any) is most important for the risk of later mental health problems.
Methods
Participants
This study used data from the All Our Families (AOF—previously All Our Babies) pregnancy cohort from Calgary, Canada. Details of the study design, recruitment and retention are described in detail elsewhere.25 Briefly, women were recruited between 2008 and 2010 during the second trimester of pregnancy. Participants had to be 18 years or older, understand English well enough to complete the questionnaires and live in Calgary. Women were given questionnaires twice in pregnancy and at 4 months and 1 year postpartum. At each follow-up, eligible women were provided with a questionnaire and prepaid return envelope. Participants were contacted by telephone to resolve missing or unclear data. Non-responders were contacted by telephone and email to provide additional opportunities to respond. At 4 months postpartum, 3057 women completed questionnaires, a response rate of 90%. At 1 year, 1573 of 1942 eligible women (81%) completed the questionnaire (online supplementary file 1). Compared with the target population, the women in these samples have higher incomes, are more educated and more likely to identify as white.25 Women were asked sociodemographic questions, repeated psychosocial measures of current anxiety, depression, social support as well as past mental health and early life experiences. All questionnaires were developed with input from healthcare providers, epidemiologists and community programme staff. All participants provided informed consent to participate. The study received ethical approval from the University of Calgary and complies with the STROBE statement for reporting on cohort studies.
Supplemental material
Variables
Primary outcomes
Depression and anxiety symptoms were assessed at 4 months and 1 year postpartum, resulting in four outcomes. Depressive symptoms were measured using the Edinburgh Postnatal Depression Scale.26 A cut-off of 10 or more is recommended for community settings and indicates a risk of minor or major depression.27 Anxiety was measured using the Spielberger State Anxiety Index, with a score of 40 or more representing clinically significant levels of anxiety.28 29
Exposure
The Medical Outcome Survey Social Support Scale was used to measure total perceived support and three types of support: tangible, positive social interaction and emotional/informational support.13 For example, items ask respondents how often they had someone ‘to help you if you were confined to bed’ (tangible); ‘to confide in or talk to about yourself or your problems’ and ‘to give you good advice about a crisis’ (emotional/informational) and ‘to do something enjoyable with’ (positive interaction). There are 19 items and responses were measured on a five-point Likert scale. There is no validated cut-point for this scale and responses were skewed. Therefore, the lowest 20% of responses of the total scale, and of each subscale, were considered to be ‘low support’. Perceived support captures what resources women think they have available to them, regardless of the need to use the support (ie, received support) or how many people provide the support.13 Also, because evidence suggests that women perceive support differently after giving birth,30 we used the most recent measure of support as our exposure. Specifically, for 4-month outcomes, the exposure was low social support at any time during pregnancy. For 1-year outcomes, the exposure was low social support at 4 months postpartum.
Covariates
Based on the previous literature, we developed a list of potential covariates for inclusion: income, self-reported ethnic background, parity and previous mental health. We did not consider marital status or relationship quality as these variables are embedded in how supported a woman feels.12 In addition, almost all of the women in our sample (94%) were married, leading to limited variability in this measure.
Household income was reported at baseline and dichotomised at $60 000. This cut-off represents approximately 65% of the median income in Calgary at the time and is the level at which residents become eligible for low-income housing.31 Household income is a proxy measure for socioeconomic status and may directly impact postpartum mental health or may moderate the type of support that is most important.
Self-reported ethnic background information was collected at baseline and categorised into white and minority. Although validated cross-culturally, screening tools may operate differently in minority ethnic groups. Ethnic background may also moderate which type of support is considered most important.
Women with a previous live birth were multiparous and those for whom the All Our Families child was their first live birth were primiparous.
We used a life course approach to operationalise previous mental health.2 15 16 A cumulative risk score can account for the aggregate and overlapping nature of mental health challenges,16 32 including possible collinearity between variables as well as the cumulative effect of past mental health on current mental health.2 6 7 16 We developed a separate cumulative mental health risk variable for each of the two outcome timepoints. We summed four variables: history of treatment of depression, having experienced four or more Adverse Childhood Experiences (ACEs) and recent symptoms of anxiety and depression. ACEs include maternal report of abuse, neglect or family dysfunction in her own childhood and a threshold of 4 or more is recommended as a cut-off for the at risk category.33 Recent symptoms for 4-month outcomes were elevated anxiety or depressive symptoms at any time during pregnancy. For 1-year outcomes, we included anxiety or depressive symptoms at 4 months postpartum. Cumulative mental health risk scores ranged from 0 to 4. A model outlining the association between social support and mental health is illustrated using a directed acyclic graph in figure 1.
Directed acyclic graph to demonstrate cumulative risk. For outcomes 1 and 2: T1 is pregnancy and T2 is 4 months postpartum. For outcomes 3 and 4: T1 is 4 months postpartum and T2 is 1 year postpartum. The cumulative risk variable is the sum of history of depression, early adversity and T1 depressive or anxiety symptoms. Analysis which controls for cumulative risk allows for the isolation of the effect of low social support at T1 on depression or anxiety symptoms at T2, independent of previous mental health. Other hypothesised possible confounding effects (age, ethnicity, income and parity) are not depicted for reasons of simplification.
Analysis
We calculated descriptive statistics for all variables described above. For aim one, our exposure was low total support. For aim two (the relative importance of support type), our exposure was three support subscales. All exposures and covariates were measured at time points prior to the outcome. Log-binomial regression models were built to estimate risk ratios (RRs) for each outcome.34 We first assessed possible effect modification by stratifying by income, ethnic background, parity and cumulative mental health risk. We then eliminated covariates that were not statistically associated with the outcome and did not change the point estimate of our exposure by more than 10%, which would indicated confounding.35
Results are reported as crude and adjusted RRs. We calculated risk differences based on the estimated prevalence in each exposure group. Based on risk differences, we calculated a number needed to treat (NNT) to inform possible interventions. Cumulative effects and risk differences were depicted graphically. All analyses were carried out in STATA IC V.13.
Results
The mean age of women at delivery was 31 years (SD 4.4), and the majority of the women had higher income and self-identified as white. Approximately half the samples (50.3%) were primiparous (table 1). At 4 months postpartum, 12.1% reported depressive symptoms and 14.9% reported elevated anxiety. At 1 year postpartum, 12.4% of women reported depressive symptoms and 16.9% reported elevated anxiety. Both prenatal and postpartum cumulative mental health scores followed expected patterns, with the majority reporting 0, and decreasing numbers reporting higher levels.
Sample characteristics
For all four outcomes, there were substantial changes in the point estimate for low social support from the crude to the adjusted models, which indicates confounding by previous mental health risk factors (as depicted in figure 1). In the adjusted model, low total support during pregnancy was associated with an adjusted 1.5-fold increased risk (95% CI 1.24 to 1.82) of depressive symptoms at 4 months (table 2). Increasing numbers of prenatal cumulative mental health risk factors were associated with increased adjusted risk of depressive symptoms, with one, two and three or four risk factors being associated with a RR of 2.52 (95% CI 1.86 to 3.41), 4.60 (95% CI 3.45 to 6.13) and 7.64 (95% CI 5.71 to 10.20), respectively. The results were similar for elevated anxiety at 4 months postpartum.
Total support log-binomial regression models for 4 month and 1 year outcomes
Adjusted RR estimates for the association between low social support at 4 months and depressive symptoms at 1 year were similar, but not statistically significant. There was a statistically significant association between low total support at 4 months and elevated anxiety at 1 year (RR 1.65, 95% CI 1.31 to 2.09). Postpartum cumulative mental health risk factors showed similar patterns for both 1 year outcomes with increasing risks associated with an increasing number of mental health risk factors.
Age, income, ethnic background and parity were not associated with any of the outcomes and did not substantially affect any of the point estimates and are therefore not included in the models. There was no evidence of effect modification by any of the covariates, including number of cumulative mental health factors.
To illustrate the multiplicative nature of the log-binomial models, figure 2 shows the adjusted RRs (primary axis) and associated CIs for each outcome. Because there is no effect modification, at each level of mental health risk, low social support in pregnancy adds an additional 63% risk of elevated anxiety at 4 months postpartum (panel B). This translates into a low absolute increase for women with no mental health risk factors (RR of 1.63 for low social support compared with baseline of 1); for women with three or four mental health risks, the increase is from 7.74 (95% CI 5.96 to 10.05) for high support to 12.62 (95% CI: 9.79 to 16.26) for low support. Absolute differences can also be seen by comparing the estimated prevalence of elevated anxiety in each group (secondary axis in figure 2—details in supplementary file 2). Among women with a cumulative mental health risk score of 0, only 4.5% of women with high social support and 7.3% of those with low social support are expected to have elevated anxiety at 4 months postpartum (RD=2.8%). Among women with a cumulative mental health risk score of 3 or 4, estimated prevalences are 34.7% for high support and 56.5% for low support (RD=21.9%).
Supplemental material
These same absolute differences can also be understood using a ‘NNT’ analysis (NNT=1/RD) (online supplementary file 2). The RD between high and low support in women with 0 mental health risk factors is 2.8% compared with 21.9% for women with 3–4 mental health risk factors, translating into an NNT of 36 or 5, respectively. Other outcomes show similar trends (online supplementary file 2).
![[SP1.jpg]](https://jech.bmj.com/content/jech/72/10/933/DC1/embed/inline-supplementary-material-1.jpg?download=true){kind=link}
Risk ratios and estimated prevalence: Showing the risk ratios and estimated prevalence by level of mental health risk and level of social support: of depressive symptoms at 4 months (panel A); elevated anxiety at 4 months (panel B); depressive symptoms at 1 year (panel C); elevated anxiety at 1 year (panel D). EPDS, Edinburgh Postnatal Depression Scale; MH, mental health; RR, risk ratio SS, social support; SSAI, Spielberger State Anxiety Index.
To understand the relative importance of support types (aim 2), all four models were re-run with the three support types as exposures (table 3). Crude estimates show that each type of support is associated with increased risk. In the adjusted model, no one type of support during pregnancy appeared more important for depressive symptoms at 4 months. For anxiety at 4 months, only low emotional/informational support during pregnancy was associated with an increased risk (RR 1.64, 95% CI 1.32 to 2.05).
Types of support log-binomial regression models for 4-month and 1-year outcomes
Crude estimates of the associations between low support type and depressive symptoms at 1 year become non-significant after adjustment. For anxiety at 1 year, both low interaction support and low emotional/informational support at 4 months were associated with an increased risk (RR 1.48, 95% CI 1.09 to 2.00 and RR 1.42, 95% CI 1.06 to 1.90, respectively).
Discussion
Our analyses show that low total support in pregnancy was associated with an increased risk of both depressive symptoms and anxiety symptoms at 4 months. This is consistent with previous research which suggested a greater number of support persons in pregnancy reduced postpartum depressive symptoms.14 Our study complements this finding as we considered social support to comprise the availability of different resources provided by support people, not only the number of people.
In terms of 1-year outcomes, low total support at 4 months was associated with an increased risk of anxiety symptoms (RR 1.65, 95% CI 1.31 to 2.09). The results for depressive symptoms at 1 year were similar, but not statistically significant. Martini et al found that higher levels of support throughout pregnancy and postpartum were correlated with reduced levels of both anxiety and depression up to 16 months postpartum.20 Currently, most perinatal mental health programmes in Canada occur during the postpartum period,36 which misses the opportunity for women to build support prenatally. Prenatal education classes, group prenatal care and pregnancy outreach workers have all been shown to improve support prenatally.37 38 Support interventions specifically targeted to women with multiple mental health risks should be evaluated.
Our analysis also clearly shows the cumulative nature of mental health risk, as women with increasing cumulative mental health risk scores had increasing risks of postpartum anxiety and depression at both 4 months and 1 year. For 3 of our 4 outcomes, low social support was associated with increased risks of poor subsequent mental health across all levels of our cumulative mental health index. We posit that social support is beneficial for everyone but provides the most absolute benefit to those with the highest mental health risks (figure 2). Moreover, increased social support is protective for multiple outcomes. If seven women with low support and high mental health risk (score of 3 or 4) were provided with additional support in pregnancy, we could prevent one case of postpartum depression, and one case of postpartum anxiety (NNT for anxiety being 5). This could result in very cost-effective targeted strategies to decrease symptoms for subsequent mental health problems.
The crude analysis (table 3) also shows that when considered independently, each type of support is important. However, when mutually adjusted for multiple support types, our results are less consistent. For both time points, low emotional/informational support was associated with increased risk of postpartum anxiety, whereas no single type of support was clearly superior for postpartum depression. We suggest that while interventions that provide only emotional/information support may be beneficial for some outcomes, a more holistic approach to providing support may be warranted. Previous studies show that group prenatal care and group prenatal education can provide multiple types of support.37 38 A randomised control trial of pregnancy outreach workers in the UK designed to provide tangible, emotional and informational support to vulnerable women showed a reduction in depressive symptoms.39 For areas where in-person support is not feasible, randomised clinical trials have shown that peer support interventions delivered over the phone or internet during the postnatal period can provide emotional and informational support and reduce depressive symptoms.36 Further studies focusing on interventions in the prenatal period and measuring anxiety outcomes are needed.
A study by Bassuk et al40 in the USA found that lack of tangible support during pregnancy was the most strongly associated with symptoms of postpartum depression. However, that study was conducted in a high-risk population of women, who may have more immediate tangible support needs that our study population. We found no evidence of modification of type of support by any of our demographic variables or previous mental health risk. However, we did have a comparatively homogeneous sample with relatively educated, higher income women, which may limit the generalisability of our results to more vulnerable populations.
Strengths and limitations
One of our study’s strengths is its large sample size, with over 3000 women for our 4-month outcomes, and over 1500 for our 1-year outcomes. While sample size decreases at 1 year, the samples at both time points remain generally representative of the pregnant and parenting population in urban Canada and levels of mental health outcomes were similar in our study to previous studies.1–4 However, our sample was slightly more educated, more likely to self-identify as white and have a higher income. Our sample may not be fully representative of the target population and may not be generalisable to more vulnerable groups. As with all observational studies, our study could not measure all potential confounders and may be subject to residual confounding or omitted variable bias. In addition, our mental health measures were self-reported, which may be less accurate than those measured by trained professionals. However, our study uses validated screening tools which are widely used in perinatal populations in community settings.25 27 29 Our study takes full advantage of its longitudinal design, with all exposures preceding outcomes, which mitigates challenges associated with measuring social support and mental health concurrently. Finally, our life-course approach allowed us to account for the cumulative nature of mental health risks, considering both past and more proximal measures of mental health, while avoiding the challenges associated with over-controlling for mental health which occur when multiple mental health measures are entered independently in a model. Our results show consistent patterns over four outcomes, which lends support to our approach and findings.
Conclusion
This study provides evidence that low levels of social support are associated with increased risk of both depression and anxiety at 4 months and with increased risk of anxiety at 1-year postpartum, taking into account previous mental health risk. Clinicians and prenatal education resources should promote developing strong support networks before birth, not just postnatally. Furthermore, optimising social support for all women has benefit, but may have a larger absolute risk reduction for those with high mental health risk. Low emotional/informational support is consistently associated with increased risk for elevated anxiety at both 4 months and 1 year postpartum, whereas only low total support impacts postpartum depression. Group-based care and other group activities that promote positive social networks and holistic support may provide the most benefit compared with interventions that focus on one type of support. Further studies that focus on providing support in the prenatal period and that include measures of perinatal anxiety are needed. Intervention studies that focus on women with high levels of previous mental health challenges, not just social or demographic risk, are needed to confirm findings.
What is already known on this subject
Up to one in five women experiences symptoms of depression or anxiety in the first year postpartum.
Low social support is consistently associated with mental health challenges in cross-sectional studies, but the longitudinal association is less clear.
There is little information on what type of support is most important to improve postpartum mental health outcomes.
What this study adds
Low social support contributes to a subsequent risk of postpartum anxiety and depressive symptoms, across multiple levels of previous mental health challenges.
Emotional/informational support has the largest impact on subsequent anxiety symptoms, but all types of support are important.
Interventions should focus on encouraging women to develop strong support networks prenatally to prevent mental health problems postpartum.
Acknowledgments
We are extremely grateful to all the families who took part in this study and the All Our Families research team. We are extremely grateful to the investigators, coordinators, research assistants, graduate and undergraduate students, volunteers, clerical staff and managers.
References
Footnotes
Contributors EH developed the idea, analysed the data and drafted the manuscript. SCT and SM conceived, implemented and managed the AOF study, assisted in interpretation and provided critical input to the manuscript. TW provided statistical expertise. TW and SBP assisted in interpretation and provided critical input to the manuscript.
Funding This study was funded by Alberta Children’s Hospital Foundation, Alberta Innovates Health Solutions, Interdisciplinary Team Grant #200700595, Canadian Institutes of Health Research.
Competing interests EH receives scholarship funding from the University of Calgary, Alberta Innovates Health Solutions and the Canadian Institutes for Health Research Vanier scholarship.
Patient consent Not required.
Ethics approval Conjoint Health Research Ethics Board—University of Calgary.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The All Our Families questionnaires is stored at Secondary Analysis for Generating Evidence (SAGE), a secure data repository managed by PolicyWise for Children and Families. Requests for data and collaborations are welcome. For further information, please visit: https://policywise.com/initiatives/sage/