Perinatal mental health is a key issue in contemporary maternity care, with suicide continuing to be a leading cause of maternal death in the UK (Knight et al, 2014). The most recent Confidential Enquiries into Maternal Deaths and Morbidity report found that maternal death rates by indirect causes have not reduced in recent years, with a woman being more likely to die as a result of suicide (0.67/100 000) than obs tetric complications such as genital tract sepsis (0.50/100 000) (Knight et al, 2014). Perinatal mental illness is defined as a mental health condition that is prevalent in pregnancy, around childbirth or up to 1 year following birth, and may include depression, anxiety, psychosis or bipolar disorder (O'Hara and Wisner, 2014).
Although the National Institute for Health and Care Excellence (NICE, 2014) estimates that 10% of all new mothers are affected by a perinatal mental health condition, it has been argued that actual figures may be far higher as many women do not seek help (Flynn et al, 2006). Depression is the most commonly diagnosed perinatal mental illness (NICE, 2014). Perinatal depression is defined in this paper as depression, new or pre-existing, experienced during preg nancy (antenatal depression) or up to 1 year postpartum (postnatal depression). Contemporary research acknowledges that the entire period is an inter linking transitionary phase of life for women, with external factors influencing maternal mental health even prior to conception (Rallis et al, 2014). In addition, having an untreated antenatal or pre-existing mental health issue significantly increases the risk of developing a postnatal mental health condition (O'Hara and Wisner, 2014).
Untreated depression can increase the risk of premature birth, intrauterine growth restriction and low birth weight (Grote et al, 2009; Misra et al, 2010). Prolonged exposure to increased levels of circulating cortisol can affect fetal brain development, potentially correlating with emotional and behavioural problems in childhood such as attention deficit disorder, anxiety and hyperactivity disorder (Letourneau et al, 2006; Talge et al, 2007). In addition, lifelong health inequalities increase when children are not exposed to effective consistent social and emo tional relationships from an early age (Marmot, 2010). Depressed mothers are more likely to limit episodes of play and the associated cognitive and emotional development (Bowlby, 1978; Murray et al, 2011). Consequently, it is essential that perinatal depression is detected and treated early to prevent the potential long-term impact on the child and the associated economic impact on the NHS.
Research suggests that the current UK health system does not always offer timely assessment or treatment of women identified as being at risk from perinatal mental health conditions (Bauer et al, 2014). Almost half of all women have no access to a specialist perinatal mental health service and only 3% of clinical commissioning groups have a strategy for increasing future provision (Bauer et al, 2014).
Women cite pregnancy-related triggers to depression including changing body shape, transition to motherhood, relationship changes in social circles and perceived loss of an established career (Rallis et al, 2014). The incidence is reduced when the transition to parenthood is supported by family, friends and the wider community (Solmeyer and Feinberg, 2011). Contemporary society and increased globalisation are not conducive to this, with families frequently living geographically separate lives.
Talking therapies and medication are considered best practice for the treatment of perinatal depression (Dennis and Hodnett, 2007; NICE, 2014). Many women may decline medication owing to fears of adverse neonatal outcomes or medicine transmission through breast milk (Dennis and Chung-Lee, 2006; Hanley and Oberlander, 2014). Talking therapies are successful but the restricted availability of appointments means that many women cannot access the help that is recommended. Logistical barriers, such as travel or childcare restrictions, may prevent women from engaging in long-term counselling (Goodman, 2009; Kim et al, 2010).
When considering the barriers to women accessing recommended treatment, alternative support should be considered. Although NICE (2014) does not make any recommendations supporting the use of psychosocial intervention, this may fill the gap in meeting the needs of women with perinatal depression.
Psychosocial interventions are any intervention of a non-pharmacological nature that addresses both the mental health condition and secondary issues such as social and relational problems (Walker, 2014). Arguably, talking therapies delivered by health professionals are a psychosocial intervention, but there are a number of published studies examining the effectiveness of the use of other perinatal psychosocial interventions delivered by non-clinicians. Further review of these papers may offer alternative or additional care pathways for women accessing care in an overstretched NHS system. Some published papers also evaluate data based on the clinician's perception of the woman's depression using structured clinical interviews or a population deemed at risk of becoming depressed in the future. It is essential to consider the woman's lived experience and perception of her own mental health (Johnstone and Dallos, 2013) and no evidence could be found that this perspective has been examined systematically. Consequently, this systematic review addresses the question: Do psychosocial support interventions have an impact on maternal perception of perinatal depression?
A further review objective considers: Is any single psychosocial intervention more effective than another?
Methods
The Centre for Reviews and Dissemination's (CRD, 2009) structure for systematic review protocol was used. The Database of Abstracts of Reviews of Effects (DARE), Cochrane Central Register of Con trolled Trials, Cochrane Data base of Systematic Reviews, NHS Economic Evaluation Database, National Institute for Health Research (NIHR) Health Technology Assessment and the Campbell Collaboration databases were searched, and three systematic reviews in this topic area were iden tified. Dennis and Hodnett (2007) considered psycho social interventions for treatment of post partum depression, but only one study in their paper met the date parameters set for this review. Lavender et al (2013) assessed the effect of telephone support for women during pregnancy, with many studies not meeting the inclusion criteria for the current review because the inter ventions were clinician-supported or studied women deemed ‘at risk’ of depression rather than screened for depressive symptoms. One study from the review was excluded after reading the full text. A narrative review of peer support by Leger and Letourneau (2015) identified two studies that were later excluded from this review.
Search terms and associated synonyms (Table 1) were collated from the scoping exercise and uti lised to perform a methodical search of CINAHL, Medline, PubMed and PsycINFO electronic databases. Grey literature and internet searching was conducted using Google and Google Scholar. Reference lists of key authors, similar reviews and all included papers were searched. Owing to feasibility constraints, the following limiters were applied: primary research published in the English language between 2005 and 2015. Alerts did not capture any newly published papers.
| Population | Intervention/comparator | Outcomes |
|---|---|---|
| Perinatal |
Psychosocial support |
Depression |
PICOS–population, intervention, comparator, outcomes, study design
The PICOS (population, intervention, comparator, outcomes, study) tool (Higgins and Green, 2008) was utilised to define the individual elements of the research question and the inclusion and exclusion criteria (Table 2). A total of 3463 records were identified through database searching and an additional 13 studies were identified through hand searching. Duplicate citations were removed, leaving 3156 records. A further 3049 were excluded by title and 85 by abstract. The remaining 22 full-text articles were assessed against the inclusion and exclusion criteria, leaving seven studies suitable for inclusion in the review (Figure 1). Reasons for exclusion are documented in Table 3.
| PICOS | Inclusion criteria | Exclusion criteria |
|---|---|---|
| Population | Pregnant women |
Non-pregnant women or women beyond 1 year postpartum at the start of the intervention |
| Intervention | Any psychosocial support intervention that does not utilise elements of psychotherapy or pharmotherapy as part of the intervention itself | Any interventions which comprise combined elements of structured psychotherapy or pharmotherapy |
| Comparator | Standard treatment |
None |
| Outcomes | Maternal reports on perinatal depression symptoms/status |
Outcomes observed by a researcher as subjective change in the woman's depression status |
| Study design | Primary research including randomised controlled trials and non-randomised studies, including pilot studies and quasi-experimental studies | Secondary research including literature reviews, discussion papers, audit of services |
PICOS–population, intervention, comparator, outcomes, study design
| Study | Reason for exclusion |
|---|---|
| Barnes et al (2009) | Population deemed ‘at risk’ of depression rather than diagnosed with a perinatal mental health issue |
| Caramlau et al (2011) | Protocol and pilot study with no usable data |
| Dennis et al (2009) | Population deemed ‘at risk’ of depression rather than diagnosed with a perinatal mental health issue |
| Dubus (2014) | Population deemed ‘at risk’ of depression rather than diagnosed with a perinatal mental health issue |
| Guszkowska et al (2013) | General population rather than with an identified perinatal mental health issue |
| Hall and Grundy (2014) | Intervention included elements of cognitive behavioural therapy and therapeutic group work |
| Logsdon et al (2010) | Unable to extract relevant data from the study |
| Milgrom et al (2011) | Mixed population of depressed and non-depressed women. Unable to extract relevant data |
| Morton and Forsey (2013) | Evaluation and case study with a very small sample |
| Norman et al (2010) | General population of postnatal women |
| O'Mahen et al (2014) | Online psychotherapy-based intervention |
| Perry et al (2008) | Population up to 2 years postnatal, scant facts or data for analysis |
| Rackett and Macdonald (2014) | Dataset received from authors did not correlate with primary dataset in the paper and therefore was deemed unreliable |
| Ross et al (2009) | Study focuses more on HIV than perinatal mental health. Results not generalisable to general pregnant population |
The Critical Appraisal Skills Programme (CASP, 2013) checklist for randomised controlled trials (RCTs) was selected for critical appraisal of the papers. There appears to be no consensus on the most appropriate tool for non-randomised studies; Quigley et al (2014) state that the CASP RCT checklist is often used to appraise both types of study, even though it is not validated to do so. Use of a single checklist, in this instance, was deemed to be the most efficient way of appraising all of the papers consistently, but is acknowledged as a potential source of bias in the review. A summary of results of the CASP checklist (and relating them to the PICOS framework) is displayed in Table 4. Level of bias in the studies was assessed using the Cochrane Collaboration risk of bias tool and displayed using the tabular form suggested by Higgins et al (2011) (Figure 2). Data were extracted using a piloted data extraction tool based on the Cochrane Collaboration data collection form (Higgins and Green, 2008). One paper (Rackett and MacDonald, 2014) was excluded after contacting the authors for more information, which led to discrepancies in the data.
| S | Study | Battle et al (2015) | Field et al (2013a) | Field et al (2013b) | Gjerdingen et al (2013) | Kamalifard et al (2013) | Letourneau et al (2011) | Heh et al (2008) |
| P | Did the trial/study address a clearly focused issue? | Yes | Yes | Yes | Yes | Yes | Yes | Yes |
| Were the study participants recruited in an acceptable way/randomised? | No | Yes | Yes | Yes | Yes | Unsure | Unsure | |
| I | Was assignment of women blinded? | No | Yes | Yes | Yes | Yes | Unsure | Unsure |
| Were groups similar at the start of the trial/study? | No | No | Yes | No | Yes | Yes | Unsure | |
| C | Aside from the intervention, were groups treated equally? | No | No | Yes | Yes | Yes | Yes | Unsure |
| O | Were all participants properly accounted for at its conclusion? | No | Yes | Unsure | Yes | Yes | Yes | Yes |
| Can the results be applied in your context? | Unsure | Yes | Yes | Yes | Yes | Yes | Yes |
CASP–Critical Appraisal Skills Programme; PICOS–population/intervention/comparator/outcomes/study design
A meta-analysis of the data was not possible owing to the heterogeneity of the review papers. A narrative synthesis approach, using a systematic method, considered each paper, aiming to reduce bias by not inappropriately stressing the results of one paper over another.
Findings
Table 5 summarises the papers included in the review. No qualitative papers fulfilled the inclusion criteria, suggesting a lack of contemporary qualitative data on this topic. The remaining studies had usable data involving a total of 449 women. Three of the studies were RCTs, with the remaining four papers consisting of a pilot study, two randomised experimental studies and a controlled trial. Although non-randomised studies are lower in the hierarchy of evidence, it is acknowledged that ‘real world data’ from smaller studies is having a significant impact on decision-making in health care (Quigley et al, 2014).
| Another snd year | Study design | Study population | intervention | Outcomes | Country |
|---|---|---|---|---|---|
| Battle et al (2015) | Experimental pilot study (34 participants) | Women at 12–26 weeks' gestation with a singleton pregnancy who were not regular yoga practitioners. Diverse population of white/Latina/African-American women from predominantly low-income households | A prenatal yoga class taught in a community setting. A 10-week programme of 75-minute classes offered twice per week. Women encouraged to attend one class per week | Significant decrease in Edinburgh Postnatal Depression Scale scores with mean baseline screening score of 13 reduced to mean final screening score of 7.5 (P = 0.05). |
USA |
| Field et al (2013a) | Randomised experimental study* (44 participants) | Women around 22 weeks' gestation with a singleton, uncomplicated pregnancy, younger than 40 years old. Primarily Hispanic or African-American women from low-income households and high school level education | A peer support group participated in 20-minute leaderless groups once per week for 12 weeks. The control group (an interpersonal therapy group) met for 1 hour per week for 12 weeks | Center for Epidemiologic Studies Depression Scale scores decreased in both groups, with a slightly higher decrease in the peer support group (although they had higher initial scores in comparison to the interpersonal therapy group). Salivary cortisol levels were also measured pre- and post-intervention. The decrease in cortisol was greater for the peer support group. The authors report medium effect sizes. Peer support was as effective as psychotherapy | USA |
| Field et al (2013b) | Randomised experimental study* (92 participants) | Women aged 20–40 years with a singleton pregnancy, assigned to the intervention or control group at 22 weeks' gestation. Primarily Hispanic or African-American women from low-income households and high school level education | A prenatal yoga group met for 20-minute sessions once per week for 12 weeks. The control group (a peer support group) also met for 20 minutes per week for 12 weeks | Center for Epidemiologic Studies Depression Scale scores decreased almost equally in both groups, showing that group peer support was as effective as yoga for decreasing symptoms of depression | USA |
| Gjerdingen et al (2013) | Pilot RCT (39 participants) | English-speaking women recruited at about 2–4 weeks postpartum, with a 0–6 month infant. Population 95% white women, 84% of whom were married and 74% with degree level education and moderate to high household income | Women were assigned to either one of two intervention groups or a control group of standard postnatal care. The intervention groups were postpartum doula support or peer telephone support. Doulas provided up to 24 hours of support over a 6-week period. Peer telephone support was offered over a period of 3 months | Center for Epidemiologic Studies Depression Scale scores increased in women who had received doula support (believed to be a result of allocation bias and doula's experience in prompting women or recognising depressive behaviours). Multiparous women did not value peer telephone support. None of the results were statistically significant | USA |
| Kamalifard et al (2013) | RCT–randomised controlled trial | Primiparous women aged 18–35 years old with an uncomplicated pregnancy and full-term birth. Two thirds of the sample had private health insurance, almost all women were housewives, and almost all husbands were business owners rather than employees | Women in the intervention group received peer support via telephone at least once per week for 2 months following birth. The control group was standard care | Edinburgh Postnatal Depression Scale scores decreased significantly for women received peer telephone support. The authors suggest the intervention may decrease social stigma for a population of women who do not readily talk about depression, and may remove access barriers such as geographical distance and transportation; however, there is no qualitative data presented to support this suggestion | Iran |
| Letourneau et al (2011) | RCT–randomised controlled trial | English-speaking women who had a singleton birth, with an infant less than 9 months old. The infant had no significant health issues. Participant also had to live within driving distance of the research cities. Most women were married with a good standard of education and had a household income >$ 40000 (US) per year | Women received 12 weeks of home-based peer support, which included an element of maternal–infant interaction teaching. The control group was standard care | Home-based peer support was less effective at reducing Edinburgh Postnatal Depression Scale scores than standard care. In this study, peer support had a negative impact on women—the authors report that this may have been influenced by the teaching component of the intervention | Canada |
| Heh et al (2008) | Controlled trial (80 participants) | Married, primiparous Taiwanese women aged 20–35 years old. All births were full-term normal deliveries. Socioeconomic status not documented, although the authors state that these data were collected | Women completed a 1-hour group exercise programme which they were asked to attend weekly. They also received a CD of the exercise programme and were asked to repeat the exercises at home twice per week | Edinburgh Postnatal Depression Scale scores measured at 5 months postpartum decreased significantly in both the exercise and the control group, with a larger decrease in the exercise group. The sample size was small, with a high attrition rate. Bias was present with the author having direct telephone contact with participants to remind them to exercise as per the schedule | Taiwan |
RCT–randomised controlled trial
No UK studies were identified; four were conducted in the USA, with the others originating from Canada, Iran and Taiwan. The control group was generally ‘standard care’, varying from country to country, and not comparable with free-at-point-of-contact UK NHS maternity care. Alternative control groups included an interpersonal psychotherapy group (Field et al, 2013a) and a purely social support group (Field et al, 2013b). In the latter of these studies, the control group itself met the remit of a psychosocial intervention and was considered accordingly. Battle et al (2015) did not have a control group, detrimentally affecting the validity and reliability of the results. Gjerdingen et al's (2013) study compared both doula support and telephone-based peer support (i.e. two psychosocial interventions) with standard care. Kamalifard et al (2013) and Field et al (2013b) were judged to have the lowest risk of bias, reducing allocation bias by randomly assigning participants to groups through the use of internet blocking or random number tables. Two papers had a high risk of bias owing to high attrition rates, the researcher telephoning participants as a reminder to complete the exercise programme (Heh et al, 2008) and the absence of a control group (Battle et al, 2015).
Physical activity-based interventions
A postnatal group exercise intervention (Heh et al, 2008) initially appeared to be an effective intervention, but Edinburgh Postnatal Depression Scale (EPDS) scores also reduced significantly in the control group. The results were statistically significant but the data show that there was a minimal difference in terms of mean EPDS screening scores at 5 months postpartum (10.2 in the exercise group vs 12.7 in the standard postnatal care group). The results also have poor external validity, with inadequate allocation concealment and evidence of selection and performance bias.
Prenatal yoga was the focus of two papers. However, Battle et al's (2015) paper was methodologically poor with no control group and, therefore, no comparison data. In contrast, Field et al (2013b) presented one of the strongest papers methodologically. Both social support and yoga resulted in statistically significant reduction in depression screening scores, with a higher mean reduction in Center for Epidemiologic Studies Depression Scale (CES-D) scores for the social support group (9.9) vs the prenatal yoga group (9.2). There were large standard deviations in data sets, a relatively small sample (n = 92) and multiple ways of assessing efficacy of the intervention (of which the CES-D scale was only one). Consequently, from the results of this review, prenatal yoga cannot be recommended as an effective psychosocial intervention for maternal perinatal depression.
Peer support-based interventions
Telephone-based postnatal peer support appeared to be the most effective psychosocial intervention and was found to be significantly more effective than standard private postnatal care for primiparous women in Iran (Kamalifard et al, 2013). This paper was methodologically strong and ranked low for bias. Of 50 women receiving peer telephone support, 28 had an endpoint EPDS score less than 11, in comparison to 13/50 women in the control group. In contrast, a small amount of not statistically significant data from Gjerdingen et al (2013) found that multiparous white American mothers did not appreciate telephone postnatal peer support owing to a lack of time and a preference for practical, hands-on support rather than the opportunity to talk about their feelings.
One-to-one home-based peer support (Letourneau et al, 2011) was the only intervention in the review that appeared less effective than standard postnatal care. Women who had experienced doula peer support still had a mean EPDS score of 11.8 in comparison to the control group, whose score dropped to 8.68. There were similar findings in Gjerdingen et al's (2013) paper, which implemented home-based doula support. The doula intervention group had more women who were depressed at 6 months postpartum than in the telephone support intervention or control groups. Data from CES-D screening scores were not statistically significant and the authors suggest that the doula's own training and experience may have influenced women in acknowledging their depressive symptoms more openly, therefore affecting the validity of the data in this study.
This review found group-based peer support to be the second-most-effective intervention (Field et al, 2013a) with statistically significant data (P < 0.005) and medium effect sizes. CES-D screening scores fell from a mean of 26.8 to 21 in the social support group, in comparison to a fall from 20 to 17.5 in the psychotherapy group. However, the small sample (n = 44) may mean that the effect sizes are overstated, as under powered studies may erroneously report effect sizes larger than the data actually suggest (Panagiotakos, 2008). Consequently, these results should be considered with caution; further research into group-based peer support is required.
Discussion
The overall sample of women were from varied backgrounds including low-income African-American and Latina women in the USA, Iranian women from affluent, well-educated families, and predominantly married white Canadian women from middle-class households. Sociodemographic data were not revealed in the Taiwanese paper (Heh et al, 2008), although the authors state that this information was collected.
In the selected papers, the authors screened for depression using either the EPDS (Cox et al, 1987) or the CES-D (Radloff, 1977). An EPDS score of ≥ 10 suggests depression, although it has been suggested that a score of > 15 in the antenatal period and > 13 in the postpartum period is a more accurate indicator of depression in English-speaking women (Matthey et al, 2006). In this review, only Heh et al (2008) adhered to Cox et al's (1987) recommendations including women with a score of ≥ 10. They justified that this threshold was culturally appropriate based on previous research within a Taiwanese population (Heh, 2001). Other authors used different cut-off scores: 11 in the Iranian study (Kamalifard et al, 2013) and 12 in the Canadian study (Letourneau et al, 2011).
The CES-D screening tool (Radloff, 1977) is viewed more favourably with a moderate criterion validity when used in vulnerable populations including low-income and minority ethnic groups (Thomas et al, 2001), reflecting the population in the majority of the studies in this review. A score of ≥ 16 is suggestive of depression. Interestingly, none of the three studies that used the CES-D (Field et al, 2013a; Field et al, 2013b; Gjerdingen et al, 2013) relied on this criterion for inclusion in the study, instead using baseline scores of 20–33.
A Cochrane review did not support the use of screening tools (Austin et al, 2008) and, in this context, it became clear that their use limits the breadth and depth of data collected. However, such tools do provide some evidence of self-perceived change in depressive symptoms in women receiving psychosocial interventions.
Psychosocial interventions and depression scores
Review papers assessed depression screening scores across the antenatal and postpartum continuum. All papers displayed an overall reduction in depression screening score for both intervention and control groups, suggesting that depressive episodes without interventions may be self-limiting for some women. Supporting research by O'Mahen and Flynn (2008) suggests women may not seek medical help when they believe that depressive symptoms may be transient; this concept is particularly relevant to women in the antenatal and postpartum period. Women may attribute sleep disturbance, hormonal changes and adaptation to parenthood as reasons for periods of low mood.
Matthey (2010) suggests that 50% of women scoring highly on perinatal depression screening tools (particularly the EPDS) are actually not depressed, and that there is a risk of over-pathologising motherhood. Many stressors are transient in early parenthood, and the use of a screening tool at the wrong time may give a false impression of perinatal distress. This is an important consideration and could arguably detract from the validity of the review findings.
Implications of findings
This review has found that some psychosocial interventions—particularly postnatal peer telephone support for primiparous women, and antenatal peer support groups—may be effective at reducing maternal perception of perinatal depression. There was limited evidence found in the reviewed papers to support the use of physical activity or one-to-one peer support interventions. The psychosocial interventions were too heterogeneous to effectively compare, which led to the review being insufficiently powered to make strong recommendations for practice. It is important that midwives interpret the findings only tentatively, owing to the limitations of this review and the multifaceted causes of perinatal depression. A larger review with wider date parameters for literature searching may change the review findings.
Perinatal depression is influenced by a number of variables and risk factors, including ethnicity (Gavin et al, 2011), finances (Lancaster et al, 2010) and level of social support (Solmeyer and Feinberg, 2011). The papers in this review acknowledge this complexity but are limited by the ability to only assess set points in time with limited demographics of the samples. None of the papers considered the entire antenatal and postpartum period, and it is possible that those interventions which demonstrated an initial effect on depression may have only had a short-term benefit. In addition, the results suggest that perinatal depression may be time-limited for some women, whose symptoms may improve without the use of psychosocial interventions.
The quantitative design of the papers restricts the depth of information collated about the participants' lives, including their cultural construction of mental ill health, whether they felt that their depressive symptoms were part of a normal transition to parenthood, or whether they desired help to actively address their own mental health concerns. The majority of papers did not assess whether the women either enjoyed or felt that they had benefited from the interventions, the effectiveness of which could affect future engagement and compliance.
None of the studies reviewed were undertaken in the UK. Nevertheless, they provide an insight into the response of women from different ethnic backgrounds to psychosocial interventions.
This review recommends future research that considers the entire antenatal and postpartum period, including qualitative studies that explore in greater depth women's experience of mental health and response to psychosocial interventions. Additionally, owing to the complex nature of perinatal depression, it is recommended that service users are involved in the design of socioculturally appropriate psychosocial interventions, which can be tested in adequately powered future studies.
Conclusions
This systematic review provides tentative evidence that psychosocial interventions in the form of antenatal peer group support and postnatal telephone support for primigravid women could be beneficial for improving women's perceptions of their symptoms of perinatal depression. Such interventions are relatively inexpensive and could help improve pregnancy outcomes as well as the health and wellbeing of women and infants throughout their lives. They could fill a gap in service provision owing to limited access to specialist perinatal mental health services for some women. There may already be local services provided by the voluntary sector to which the midwife can signpost women after assessing their perinatal mental health. However, further large-scale, well-designed studies are essential.