A systematic review, evidence synthesis and meta-analysis of quantitative and qualitative studies evaluating the clinical effectiveness, the cost-effectiveness, safety and acceptability of interventions to prevent postnatal depression

Prevalence

Postnatal depression is a public health problem4,6 which occurs in most cultures. 6–8 The prevalence of both major or minor depression during the first postnatal year is 7–13%. 9 Among a sample of more than 8000 women in England, 13% scored 13 or more (the threshold to identify women with probable major depression)2 on the Edinburgh Postnatal Depression Scale (EPDS)10 on at least one postnatal assessment. 11

Some women recover by the time their infant is 6 months old, but in 50% of women depression can last for more than 6 months. 12 Although PND is defined as depression within the 12 months after the birth of an infant, a significant number of women remain depressed for over 1 year,13 and some women remain depressed for 4 years. 12

Although depression postnatally may not be different from depression occurring in non-pregnant women, some women become depressed for the first time postnatally, some experience postnatal recurrence of previous depression13 and, for others, depression begins antenatally and continues postnatally. 14–16 Antenatal depression is the strongest predictor of PND,14 being as common as PND, with 18.4% of women having depressive symptoms throughout pregnancy. 17 Antenatal anxiety is commonly comorbid with antenatal depression and also increases the likelihood of PND. 14,15,18

Additional factors have consistently been associated with PND. Some PND may be biologically mediated and specifically linked to childbirth. 1 Some women with PND may be genetically more reactive to the environmental trigger for depression. 19 In other women, who have a general vulnerability to depression, PND may occur because childbirth is a stressor. 1 The strongest predictors of PND are antenatal anxiety and antenatal depression,14 lack of social support, a history of depression, neuroticism, low self-esteem, stressful life events during pregnancy, poor marital relationship and domestic violence. 1,20,21 Women themselves have reported that the causes of their PND were lack of support, pressure to do things right, their personality (prone to mental health problems), pressure (work or money), hormonal changes and resurfaced memories. 22 As the aetiology is diverse, it is difficult to predict accurately which women will develop PND.

Impact of health problem

The burden of PND can extend, in its most severe form, to suicide and, less frequently, infanticide. 23 The impact of PND on mothers is compounded by impairments to the mother–infant interaction24 and impairments to the infant’s longer-term emotional, cognitive, behavioural and social development. 25,26 The impact of withdrawn behaviour24 and vocally communicated sadness27 appears to be worsened when women live in poorer socioeconomic circumstances, and is worse if the infant is a boy28,29 or if depression becomes a chronic problem. 30,31 Additional later risks for infants are mediated through the effect of chronic depression on the hypothalamic–pituitary–adrenal axis functioning in offspring, into adolescence. 25,32,33

Depressed pregnant women have a greater risk of delivering a low-birthweight infant. 34 Antenatal depression is a risk factor for infant mood33,35 and for depression in offspring at 18 years of age, with higher risk among offspring whose mothers are less educated. 16,36 There is a potential impact on fathers, around 10% of whom are at risk of depression, particularly during the 3–6 months after the infant is born. 37 This depression is moderately positively correlated with maternal depression, but it is unclear if there is an association or a causal influence, and the direction of the influence, if any, is unknown. 37 Furthermore, postnatal paternal depression is associated with depression in offspring. 16

Variation in service and uncertainty about best practice

Free maternity care in the UK, delivered predominantly by midwives and obstetricians, provides opportunities for women to have contact with health-care services. The National Institute for Health and Care Excellence (NICE) provides evidence-based guidelines for antenatal, intrapartum and postnatal care, and for antenatal and postnatal mental health. 38 Among those at low obstetric and medical risk, nine antenatal consultations are recommended for women expecting their first baby and seven consultations for those expecting a subsequent child. 39 Most women give birth in hospital maternity units, or in free-standing or alongside midwifery units and stay in for less than 2 days; fewer than 3% give birth at home. 40

Traditionally in the UK, hospital midwives have provided care in hospital for antenatal, labouring and postnatal women. Community midwifery teams have provided antenatal care in the community, and postnatal care during visits to the woman’s home, community health centres and children’s centres for up to 28 days after birth. Care is usually transferred on postnatal day 10 to the health visiting service and is provided by health visitors; specially trained public health nurses. Most health visitors now offer antenatal visits.

National Institute for Health and Care Excellence guidance38 recommends that primary health-care professionals should routinely enquire about past and current mental illness, and family history of perinatal mental illness, at a woman’s first appointment in early pregnancy, and postnatally (4–6 weeks and 3 or 4 months) to identify predictive risk factors. NICE guidance38 also recommends that midwives enquire within the first 24 hours after birth about a woman’s experience of her labour. In some locations, midwife-provided services have developed to provide an opportunity for women to discuss their birth experiences, but these do not always include access to formal psychological support.

The community midwife’s role includes an increased focus on improving public health and current pre-registration midwifery education covers the identification of potential mental health issues for childbearing women. The Maternal Mental Health Pathway41 guidance focuses on the health visitor’s role in maternal mental health and wellbeing during pregnancy and postnatally, recognising the contribution of midwives, mental health practitioners and general practitioners (GPs).

Other maternity support roles include maternity support workers and volunteers, such as breastfeeding peer supporters, counsellors and doula support (women who provide support to other women), during pregnancy, labour and birth and the early postnatal period.

Infrequently in the UK, and more commonly in the USA and a small number of other countries, CenteringPregnancy^® (Centering Healthcare Institute, Boston, MA, USA) is available. 42,43 The CenteringPregnancy44 approach provides group care to women at similar stages of pregnancy by means of a health assessment and provision of education and peer support. Health-care professionals help women to participate in their own care and to learn from each other about pregnancy and care of the new infant.

Identification of postnatal and antenatal depression

There has been a lack of consistency in the routine approach to the identification of PND9,45,46 by primary health-care professionals. 47 NICE advocates a case-finding approach for depressive symptoms,38 based on two questions, the Patient Health Questionnaire (PHQ)-2, from the PHQ-9, as follows:48,49 ‘Over the last 2 weeks, how often have you been bothered by any of the following problems?’ (1) ‘Little interest or pleasure in doing things’ and (2) ‘Feeling, down, depressed, or hopeless’. 49 The EPDS,10 the Hospital Anxiety and Depression Scale (HADS)50 and the full PHQ-9 are to be used as follow-up tools as part of a fuller assessment process. The EPDS is frequently used as it performs well for major and minor depression,45 and is acceptable to women and health-care professionals. 51 The EPDS is not used systematically throughout the UK to identify depressive symptoms during pregnancy or postnatally partly because it ‘does not satisfy the National Screening Committee’s criteria for the adoption of a screening strategy as part of national health policy’. 52

Current service costs

Apart from the distress for women and the potential long-term consequences for infants, there are additional public health, social and economic consequences of maternal depression. 4 The cost of PND to the UK government is estimated as £45M53 to £61M per year. 4 For each exposed child, the estimated cumulative economic costs of adverse child development linked to a mother’s depression is £8190. 54 The health-care costs associated with postnatal paternal depression have been estimated for fathers with depression as £11,041, for fathers at high risk of developing depression as £1075 and for fathers without depression as £945 at 2008 prices. 55 In New Zealand, the potential value for money of implementation of a PND screening programme was assessed and the programme was found to be cost-effective. 56 In contrast, following a cost-effectiveness analysis, a system to identify PND in the UK was reported not to represent value for money based on the assumed cost of false positives. 57 Little is known about the economic consequences of PND or the cost-effectiveness of interventions aiming to prevent or alleviate PND symptoms. 58 Substantial economic returns have been estimated for investment in the prevention of mental health problems, with potential long-term pay-offs continuing into adulthood. 59

Despite the ‘case-finding’ approach to identify women at greater risk of PND, mainly based on earlier experience of mental health problems, little attention is paid to the prevention of PND, and no specific instruments are available to reliably predict PND among asymptomatic women. Some health visitors in the UK use the EPDS, but this practice varies nationally. It is likely that even less attention is paid to identifying depression and anxiety antenatally than postnatally.

Preventive interventions for postnatal depression

This section provides an overview of the rationale for the prevention of PND and a description of approaches that have been explored to prevent PND. There is evidence of the effectiveness of pharmacological60 and psychological interventions61–63 to treat PND within four main approaches: general counselling, interpersonal psychotherapy (IPT), cognitive–behavioural therapy (CBT) and psychodynamic therapy. 1 Prevention of a major depressive episode implies reducing the intensity, duration and frequency of depressive symptoms. 64

NHS England has provided a £1.8M budget for public health responsibilities, covering screening, immunisation and health-visiting services. 65 Less than 5% of NHS funding in England is spent on prevention, of all conditions. 65 The Marmot et al. 66 review aims to strengthen the role and impact of ill-health prevention, prioritising prevention and early detection of mental health conditions and early intervention. Traditionally primary, secondary and tertiary prevention activities are designed, respectively, to reduce the risk of developing health problems, to identify and manage pre-symptomatic ill health and to reduce the impact of the disease.

Three levels of preventive intervention are relevant to the prevention of PND:67

1. Universal preventive interventions are available to all women in a defined population not identified on the basis of increased risk for PND.

2. Selective preventive interventions are offered to women or subgroups of the population whose risk of developing PND are significantly higher than average, because they have one or more social risk factors.

3. Indicated preventive interventions are offered to women at high risk of developing PND on the basis of psychological risk factors, above-average scores on psychological measures or other indications of a predisposition to PND but who do not meet diagnostic criteria for PND at that time.

Universal preventive approaches may be less stigmatising than selective preventive interventions, but little attention has been paid to universal prevention in pregnant women, partly because the cost of a universal programme is likely to be high63 compared with a selective approach to identify higher-risk women. For example, 81% of women do not have an EPDS score 13 or more during pregnancy. 14 However, there is a rationale for providing a preventive intervention to women with subthreshold symptoms of depression who may otherwise go on to develop depression. 18,64

The outcomes for a selective intervention depend on how the population and risks are identified and defined. 63 Although indicated preventive interventions for PND could be regarded as addressing prodromal symptoms and therefore are not actually preventive, they could be regarded as early intervention. 68

The rationale for antenatal prevention of PND is based on data from the Avon Longitudinal Study of Parents and Children study14 showing that 43.7% of women with an EPDS score 13 or more at 32 weeks of pregnancy experienced elevated symptoms postnatally. Aiming to prevent, identify and treat antenatal depression presupposes that this will lead to a reduction in antenatal maternal morbidity and severity, deleterious effects on the developing infant, postnatal maternal morbidity and severity and other adverse outcomes in the offspring. 16,69 Hence, investment during pregnancy and postnatally may yield future benefits and financial savings in different areas of health and social care.

Psychological approaches to the prevention and treatment of depression

The psychological literature attests to the large effort expended on research into differing psychological approaches to the prevention70 and treatment of depression. 71–75 Although depression has often been the initial target condition for testing psychological approaches, it has equally often proved to be a more challenging condition when attempting to establish mechanisms of change that are specific to particular models of therapeutic interventions. A review of 101 randomised controlled trials (RCTs) on the treatment of major depression concluded that IPT, CBT and behaviour therapy are effective, while brief dynamic therapy and emotion-focused therapy are possibly effective. 72

A different body of literature suggests relatively small differences between the outcomes of different psychological interventions for depression. An earlier review which controlled for researcher allegiance (belief in the superiority of a treatment) found small effect sizes from comparisons between specific therapies. 73 This finding has been broadly supported in a meta-analysis of 58 outcome studies for depression which made direct comparisons between specific therapies, which yielded similarly small effect sizes. 74 However, arguments suggesting that researcher allegiance bias is related to treatment effects have been both supported76 and challenged. 77

A wide-ranging review of the efficacy and effectiveness of psychological therapies in general concluded that they were broadly effective for depression with little difference between theoretically diverse interventions. 78 Estimates of the proportion of outcome variance attributable to components of therapy comprised the following: extra-therapeutic factors, 40% (e.g. delivered individually or in a group or the number of sessions); relationship, 30%; placebo/expectancy effects, 15%; and specific techniques, 15%. 78,79 A subsequent meta-analysis in which common factor control groups were employed supported these estimates. 80

Extensive efforts have been afforded in relation to the development of measures81 and the measurement of outcomes82 in psychotherapeutic interventions, and the role of non-specific (common) factors, such as congruence, positive regard and empathy, has long been recognised. 83,84 The account of broadly similar outcomes despite diverse therapeutic interventions (termed the equivalence paradox)85 has yielded sophisticated accounts to explain this phenomenon, with the existence of common factors persisting as one major explanatory source. 85 However, others have argued that there is no clear evidence supporting a causal link between common factors and therapeutic outcomes. 86 The debate is not so much focused on the validity of the concept but rather on the absence of experimental manipulation as a route to determining which common factors, if any, impact on therapeutic change. The concepts of hope and expectancy, among others, have been posited as common factors, but the main focus for research has been on the concept of the therapeutic relationship or alliance.

Educational interventions

Attention has been paid to developing preventive strategies or interventions that focus on couple communication or parenting skills to ease the transition to parenthood. 87 Antenatal preparation for parenthood has traditionally focused on aspects of the woman’s pregnancy and on preparation for childbirth, with less attention paid to what to expect when the infant arrives or to couple communication or parenting. 88,89 Dyadic relationship quality is adversely affected90 in 67% of new mothers91 and 45% of new fathers92 during the first year of parenthood. Despite the central role of partner support in maternal mood,93 new parent couples have reported being shocked by and unprepared for adverse changes in their relationship, feeling sad and bemused that no one had talked to them about the changes they would experience in their relationships. 94

Some preventive educational interventions have been delivered universally to all expectant parents, making use of the opportunities to access this population through established antenatal care pathways, thereby reaching couples who may not otherwise seek such support. 95 These, and more targeted, approaches cover a variety of levels of intensity and format and timings.

Social support

Social support is a multidimensional concept that incorporates appraisal, companionship, informational, motivational and instrumental support; that is ‘. . . information leading the subject to believe that they are cared for and loved, esteemed and a member of a network of mutual obligations’. 96 Social support involves both social relationships that are embedded, such as relationships with family members or friends, and those that are created. 97

There are several pathways through which social relationships and social support can affect mental health. Social support can operate to promote health directly by enhancing feelings of well-being or by buffering the negative influences of stressful events. Integration in a social network might also directly produce positive psychological states, including sense of purpose, belonging and recognition of self-worth. 98 These positive states, in turn, might benefit mental health because of an increased motivation for self-care, as well as the modulation of the neuroendocrine response to stress. 98 Being part of a social network enhances the likelihood of accessing various forms of social support, which in turn protects against distress. 99 Members of a social network can exert a salutary influence on mental health by role modelling health-relevant behaviours. 100

Several different psychosocial mechanisms link aspects of social relationships to physical and emotional well-being: social influence/social comparison, social control, role-based purpose and meaning (mattering), self-esteem, sense of control, belonging and companionship and perceived support availability. 101 Given the importance of social support on mental health outcomes, enhancing social support has been used as a strategy for both the prevention and treatment of PND.

Pharmacological interventions or supplements

Some of the earliest interventions for the treatment and prevention of PND were hormonal. Uncontrolled studies used progesterone,102–104 but no controlled studies have been conducted of progesterone or oestradiol, as either a treatment or prevention.

Compared with the results of trials supporting antidepressant treatment for major depression, there is relatively little evidence to guide the clinician in treating or preventing PND. The mainstay of treatment has been antidepressant medication but women are reluctant to take antidepressants,60 as they are concerned about their safety when breastfeeding and the potential for side effects to disturb their interaction with their infant. 105

It has been reported that fish consumption and omega-3 status after childbirth are not associated with PND,106 but there is still interest in exploring the role of omega-3 fatty acids in PND, alone or combined with supportive psychotherapy. 107

Complementary and alternative medicine

This review adopts a generic definition of complementary and alternative medicine (CAM): ‘A group of diverse medical and health-care systems, practices and products that are not presently considered to be part of conventional medicine’. 108 Although this definition meets with problems in many areas of medical practice, in that what were once regarded as CAM are now provided as part of conventional medical service, it works reasonably well in perinatal depression, as CAMs are not generally provided in perinatal services.

Complementary and alternative medicine is widely used by pregnant women in the Western world, particularly those who are highly educated and have high incomes,109 often to reduce stress and improve mood; however, their use remains controversial. 110 Controversy extends beyond the definition of CAM, to the nature of the effects of CAM and to the quality of CAM research. CAM is also widely used by the general public, particularly women,111,112 many of whom do not report its use to their doctors. It is often used to promote wellness in the positive holistic sense as well as in the management of symptoms and disease. CAM has been offered to women with the aim of treating both antenatal depression63,113–115 and PND,63,116 alone or in combination.

The CAM interventions most commonly explored in these studies include aromatherapy, massage, hypnosis and other forms of relaxation therapy, herbal medicine, mindfulness and meditation, acupuncture and general traditional Chinese medicine, Ayurvedic medicine and homeopathy. Acupuncture is a popular form of treatment for depression outside the perinatal period, and there is evidence that its effectiveness is equivalent to that of antidepressants117 and that side effects are rare. Acupuncture in the context of antenatal depression was examined by a Cochrane review118 that reported inconclusive evidence.

Mind–body therapies have also been used to treat depression in general and in the perinatal period specifically,116,119 and for many there is some evidence of effectiveness. 120 Mindfulness has received specific attention in the context of perinatal depression121 and is supported by an evidence base showing that it is effective in depression in general. 122

Yoga and tai chi/qi gong are practised both alone and as a component of Ayurvedic and traditional Chinese medicine and are used by pregnant women to improve their health. 110,119 The health effects of these traditional medical approaches are held to extend beyond physical fitness, suppleness and strength, and they overlap with those of simple physical activity, which has also been investigated as an intervention to reduce depressive symptoms in pregnant women. 123

Service user involvement

The Nottingham Expert Patient (EP) committee is a group of women who have experienced the distressing effects of severe PND. Three of the women in the group were admitted to a mother and baby unit and all received community psychiatric care. The EP committee, established in 2009, has acted as the patients’ ‘voice’, advising the East Midlands Perinatal Mental Health Clinical Network Board on how to develop local services to meet the needs of women who experience mental health problems in pregnancy and after childbirth. The EP committee has joined the newly formed National Perinatal Mental Health Clinical Reference Group to ensure that the experiences and views of patients inform and influence the planning and delivery of the specialised service.

The EP committee were pleased to be invited to contribute to this review, to be involved in the development of the research proposal and to provide patient and public involvement (PPI) advice throughout the research. The EP committee reviewed the draft research proposal and provided detailed feedback to the principal investigator. The EP committee has maintained involvement through contact with the principal investigator (JM), ad-hoc meetings, having an EP committee member sit on the Expert Clinical/Methodological Group and providing input into this report.

Search strategies for identification of studies

The review of effectiveness of interventions to prevent PND constituted the central platform for this report. The objectives of the individual RCTs and the data available from them determined what NMAs were feasible. The analysis of effectiveness determined the subsequent qualitative synthesis and economic analyses. The leading candidate interventions, demonstrated in terms of potential effectiveness, became the focus for the realist synthesis. This filtered approach recognised that it would not be feasible to conduct rich interpretive explorations across the wide heterogeneity of possible interventions and, therefore, interpretive resources were focused where they were most likely to yield insights on current and future interventions.

Search strategy for randomised controlled trials and systematic reviews

Search activities were as follows:

1. searches of electronic databases

2. searches of the internet

3. searches of specific websites

4. citation searches

5. reference lists of relevant studies

6. hand searches of relevant journals

7. scrutiny of references listed in reviews of the prevention of PND

8. suggestions from experts and those working in the field.

Inclusion and exclusion criteria for quantitative studies

Electronic databases

The search for the clinical effectiveness evidence was run with a qualitative filter to identify qualitative studies. The list of electronic bibliographic databases searched is presented in Appendix 1. The search was run again with a mixed-methods filter (devised with AB) to find papers that used quantitative and qualitative methodology. The numbers of qualitative studies and mixed-methods studies retrieved for the various databases searched are presented in Appendix 4.

Study selection criteria and procedures for the quantitative review

Two reviewers (JM and PS) independently screened the titles and abstracts to identify papers for possible inclusion. If no abstract was available, the full paper was retrieved for scrutiny. Full papers for RCTs were obtained if the abstract showed that the study fulfilled the inclusion criteria or it was unclear from the abstract whether or not the inclusion criteria were fulfilled. All full papers retrieved were independently reviewed by two reviewers. Papers were not excluded on quality at this selection stage. The full papers had to fulfil the inclusion criteria presented in Tables 2–5. Where there was no consensus following discussion about inclusion at the full-paper stage, a third reviewer or clinical expert (CLD, HS or SS-B) was consulted. The reasons for exclusion are presented in Appendix 5.

Study quality assessment checklists and procedures for the randomised controlled trials

Data extraction for randomised controlled trials

Data from the full papers were entered on to a specially designed, pre-piloted and tailored data extraction form, to summarise the intervention. The primary aim of the study was documented (PND prevention, antenatal well-being, birth outcomes, general health, general psychological well-being, infant outcomes or family outcomes). The intervention and comparison arms were described. The data extraction form indicating the main RCT characteristics is presented in Appendix 6.

Outcomes were recorded as maternal, neonatal and family outcomes, using mean [standard deviation (SD)] values when available and numbers and proportions of participants in specific outcome categories. The quality of the extracted data was checked (JM and PS).

Data synthesis of randomised controlled trials

A large number of RCTs and systematic reviews were eligible for inclusion according to our broad inclusion characteristics. We conducted a narrative description of the studies according to the level of preventive intervention (universal, selective or indicated), class of intervention and the context within which the RCTs were undertaken.

Methods of evidence synthesis

The extracted data and quality assessment variables were presented, for each study, in structured tables and as a narrative description. Both conventional RCTs, in which individual women were randomised to interventions, and cluster RCTs (CRCTs) were eligible for inclusion. Estimates of treatment effect and standard error of treatment effects from CRCTs were included in the analyses after allowing for the cluster design.

The reference treatment, for comparative purposes and for estimating intervention effects, was defined as usual care. Usual care in the UK, Australia, Canada, France, Norway and the USA was assumed to be sufficiently similar to be interchangeable and was collectively defined as ‘usual care’ for the purpose of the analysis.

The evidence was synthesised using a NMA. 128 A NMA (also known as a mixed-treatment comparison or a multiple treatment comparison) is an extension of a standard (pairwise) meta-analysis. It allows evidence from RCTs comparing different interventions to be combined to provide an internally consistent set of intervention effects while respecting the randomisation used in individual studies. The NMA enables a simultaneous comparison of all evaluated interventions in a single coherent analysis; thus, all interventions can be compared with one another, including comparisons not evaluated within individual studies. The only requirement is that each study must be linked to at least one other study through having at least one intervention in common. The analysis preserves the within-study randomised treatment comparison of each study and assumes that there is consistency across evidence. As with standard pairwise meta-analyses, treatment effects are assumed to be exchangeable across studies. In addition, it is assumed that treatment effects are transitive such that if the effect of intervention 2 relative to intervention 1 is d₂₁ and the effect of intervention 3 relative to intervention 1 is d₃₁, then the effect of intervention 3 relative to intervention 2 is d₃₂ = d₃₁ – d₂₁; this allows a synthesis of direct and indirect evidence about intervention effects and a simultaneous comparison between interventions. Evidence from RCTs presenting data at any assessment time up to 12 months were considered relevant to the decision problem.

Methods for the estimation of efficacy

Study selection criteria and procedures for the effectiveness review

A two-stage sifting process for inclusion of studies (title and abstract then full paper sift) was undertaken. Titles and abstracts of the qualitative studies were scrutinised by one assessor (AS) using the inclusion and exclusion criteria. No papers were excluded on the basis of quality at this stage. Full papers were obtained for potentially included studies and for where the abstract provided too little information. One-fifth of the total citations identified by electronic database searching (n = 2313) were checked for inclusion or exclusion by AB (n = 427).

Inclusion and exclusion criteria for qualitative studies

The PICOS process was used to clarify the inclusion and exclusion criteria (Box 6).

Study quality assessment checklists and procedures for qualitative studies

Studies meeting the inclusion criteria were evaluated by two reviewers (AS and AB) using the CERQual (Confidence in the Evidence from Reviews of Qualitative research) approach,135 which aims to assess how much certainty could be placed in the qualitative research evidence, and were rated as ‘very low’, ‘low’, ‘moderate’ or ‘high’. A summary assessment was made for each study, based on the methodological quality of each included study and the coherence of the review findings (the extent to which a clear pattern was identifiable across the individual study data). Coherence was assessed by examining whether or not the review findings were consistent across multiple contexts and incorporated explanations for variation across individual studies. Coherence was strengthened when individual studies contributing to the findings were drawn from a wide range of settings.

The methodological quality of individual studies was appraised using an abbreviated version of the Critical Appraisal Skills Programme (CASP) quality assessment tool for qualitative studies. 136 Two reviewers (AS and AB) independently applied the set of quality criteria to each included study.

Review findings were subsequently graded as ‘high’, ‘moderate’, ‘low’ or ‘very low’ according to the CASP assessment, the number and richness of the data in the studies, the consistency of the data within the studies across study settings and populations and the relevance of the findings to the review question.

Data extraction strategy for qualitative studies

Data extraction from included studies was undertaken by AS using a data extraction tool adapted and tailored for the qualitative review. A 20% sample of data extractions were checked by AB. When data for included studies were missing, reviewers attempted to contact the authors at their last known e-mail address.

Selective extraction of findings137 was undertaken when the data were pertaining to an optimal intervention to be delivered antenatally or postnatally to prevent PND. A framework for extraction was developed to elicit data extraction elements related directly to the review question. The data extraction elements for the data extraction for the studies are presented in Appendix 6. The level of extracted evidence included information on characteristics of the intervention, identified in the results and discussion sections and author comments and interpretation.

Data synthesis for qualitative studies

Qualitative meta-synthesis was undertaken by highlighting women’s and service providers’ issues around the acceptability of interventions, and elucidating evidence around regarding personal and social support strategies (PSSSs) applied by women, using the data extraction framework and thematic synthesis to aggregate the findings. 138 Evidence about interventions from women and service providers and evidence about PSSSs are presented separately (see Appendix 7).

Identification of key potential CLUSTERs

To exploit the potential of realist synthesis approaches requires rich conceptual and contextual data. Reporting limitations and the varied emphases of published reports make it unlikely that all relevant data are included in a single report of a study. However, the scale and expense of a RCT increases the likelihood that multiple research reports have been produced relating to the study of interest. Such reports may include supplementary qualitative work, process evaluations, student projects, pilot studies, feasibility studies and follow-up studies. All such papers may help us to understand the study context, mechanisms and outcomes. Therefore, a key task is to move from analysis of a single study report to a detailed examination of a cluster of related papers. Such forensic examination looks not only for directly related ‘sibling studies’, but also for tangentially related ‘kinship’ papers (i.e. papers that may represent replication of an existing programme in a different context, thus allowing for comparison and contrast). Finally, syntheses, analyses and theoretical papers may locate the study within a wider context of exemplars or case studies, thereby expanding the potential for comparison. Selection of clusters is necessarily limited by the resources available for analysis. In-depth analysis, as typically performed for realist synthesis, typically precludes the comprehensive and exhaustive approaches prescribed by systematic review methods. In selecting focal study clusters, the team considered both the likely success of the programme and the availability of sibling and/or kinship study reports. At this stage, the Preparing for Parenthood cluster was excluded, as even though it possessed several companion reports the trial did not demonstrate potential effectiveness.

Searching for CLUSTER documents

Searching for documents to populate a study cluster has, until recently, been viewed as essentially an unsystematic and arbitrary procedure. Conceptually it draws upon the long-established retrieval practices outlined in Bates’ seminal paper139 on ‘berry picking’ including ‘backward chaining’ (following up cited references) and ‘forward chaining’ (following up cited articles). Recent years have revealed a prodigious potential yield from supplementary documents. For example, a review by Jagosh et al. 140 revealed several clusters with an average of 12 reports per cluster. We used systematic methods previously developed by one of the authors of our study (AB) for implementing cluster to become CLUSTER searching, for which a full published description of the CLUSTER methods (Citations, Lead authors, Unpublished materials, Scholar searches, Theories, Early examples, Related projects) is openly available. 141 In essence, the research team undertook persistent pursuit of study links, contextual links and theoretical links from the source study or studies to other related reports which then, themselves, initiated a further cause for searching. CLUSTER searching is reliant on relatively rapid judgements on potential links between a referring document and its referent. 141 When papers shared a study identifier or acronym (e.g. PoNDER) or a RCT identifying number, such connections were easy to establish. However, more typically a sibling relationship between papers relies on similarities in authorship, study context and sponsoring institution. However, further checks involve pursuing cross-citation and co-citation so that a network of studies could be constructed.

Synthesis and construction of a theoretical model

For the synthesis stage we developed a rapid realist review approach, provisionally labelled as ‘best-fit realist synthesis’. This involves:

1. identification of a provisional ‘best-fit’ conceptual framework as a starting point for data analysis

2. population of the conceptual framework with ‘if–then’ statements from the identified articles

3. construction of pathways or chains from ‘if–then’ statements to surface potential mechanisms by which outcomes might be achieved

4. identification of existing theory underpinning individual mechanisms

5. development of a programme theory to explain how PND prevention programmes may work

6. testing of the programme theory with contextual data from included studies.

Quantitative review study characteristics

An overview of the 86 included RCTs is presented here. 61,121,123,129,146–150,153–229

Yield of systematic reviews

Twenty-three reviews were included (i.e. Austin et al. ,230 Bennett et al. ,231 Cuijpers et al. ,68 Dale et al. ,232 Dennis and Creedy,233 Dennis,234 Dennis,118 Dennis,235 Dennis and Kingston,236 Dennis et al. ,237 Dennis,238 Dodd and Crowther,239 Fontein-Kuipers et al. ,240 Howard et al. ,241 Jans et al. ,242 Lawrie et al. ,243 Leis et al. ,244 Lumley et al. ,245 Marc et al. ,246 Miller et al. ,247 Sado et al. ,248 Shaw et al. 249 and Sockol et al. 250), of which one provided an additional included study not identified in the searches. 191 The included reviews are summarised in Appendix 8.

Quantitative review study characteristics

Some studies are reported in multiple references, for example Armstrong et al. 1999,164,251,252 Brugha et al. 2000,188,253,254 Chabrol et al. 2002,158,255,256 Cooper et al. 2009,153,257 Dennis et al. 2009,205,258 Gamble et al. 2005,221,259,260 Harrison-Hohner et al. 2001,208,261 Ickovics et al. 2011,222,262 Lumley et al. 2006,147,263 MacArthur et al. 2002,146,264 Makrides et al. 2010,211,265 Morrell et al. 2000,199,266 Morrell et al. 2009,61,151,152 Petrou et al. 2006,174,267 Reid et al. 2002,200,268 Richter et al. 2014,203,269 Rotheram-Borus et al. 2011,226,270 Sen 2006,191,271 Stamp et al. 1995,195,272 Wisner et al. 2001,215,273 Wisner et al. 2004216,274 and Wolman et al. 1993. 204,275,276 Henceforth studies are referred to by the first identifying reference only.

Outcome assessment

The studies varied in their duration and assessment times: 6–8 weeks, 10–12 weeks, 3 or 4 months, 5 months, 6 months, 7 months, and 12 months postnatally.

Following the description of the overall study quality, the RCTs are described fully according to the level of preventive intervention in Chapters 5–7.

Quality of quantitative studies

Overall risk of bias of randomised controlled trials

Selection bias

Of all the 86 RCTs, 64 (74.4%) reported an adequate method for random sequence generation (low risk of bias), 16 (19.0%) were unclear about the allocation method (unclear risk of bias) and six (7.0%) had used a non-random process (high risk of bias). The greatest level of risk was associated with allocation concealment. Furthermore, 50 RCTs (58.1%) reported adequate treatment allocation concealment (low risk of bias), 27 (31.4%) were unclear (unclear risk of bias) and nine (10.5%) were at high risk of bias.

Performance bias

The nature of most of the interventions made blinding of participants and caregivers not possible, but it is unlikely that the lack of blinding could not have affected the results. Therefore, 73 RCTs (84.9%) were rated as being at low risk of performance bias for the assessment of blinding for participants and staff, 12 (14.0%) were rated as being at unclear risk of bias and one RCT (1.2%) was rated as at high risk of bias for failing to report blinding status of the women and study personnel. In 61 RCTs (71%), outcome assessors were reported to be blinded (low risk of bias), in 24 (28.0%) it was not clear if the outcome assessors were blinded (unclear risk of bias) and in one RCT the risk of bias was rated as high.

Attrition bias

The risk of attrition bias was assessed as low for 51 (59.3%) RCTs, and unclear for 27 RCTs (31.4%); eight (9.3%) RCTs were assessed as being at high risk for selective outcome and/or analysis bias.

Reporting bias and other bias

The risk of reporting bias was assessed as low in 74 RCTs (86%), unclear in eight RCTs (9.3%) and high in four RCTs (4.7%).

The risk of other bias (e.g. funding source or inappropriate analysis) was judged to be low for 54 RCTs (62.8%), unclear for 28 (32.6%) and high for four (4.7%).

Overall, the risks of bias were rated as higher for universal preventive intervention studies than for the selective and indicated preventive interventions; this was most notable for selection bias and attrition bias. The judgements about each risk of bias domain are presented in Tables 1–3 for each included study according to the level of preventive intervention (universal, selective or indicated) and summarised in Figure 3 for all included studies.

TABLE 1 - Risk of bias for included universal preventive intervention RCTs: summary judgments about each risk-of-bias item

Key: high, high risk of bias; ID, identification; ITT, intention to treat; low, low risk of bias; PP, per protocol; unclear, unclear risk of bias.

First author, year, reference number	Selection bias		Performance bias	Detection bias	Attrition bias	Reporting bias	Other bias
	Random sequence generation	Allocation concealment	Blinding of participants/personnel	Blinding of outcome assessors	Incomplete outcome data	Selective reporting of the outcome, subgroups, or analysis	Funding source, adequacy of statistical methods used, type of analysis (ITT/PP), baseline imbalance in important characteristics
Christie, 2011150	Low	Low	Unclear	Low	Low	Low	Low
Cooper, 2009153	Low	Unclear	Unclear	Low	Low	Low	Low
Doornbos, 2009207	Low	Unclear	Low	Low	High	Low	Unclear
Feinberg, 2008180	Unclear	Unclear	Unclear	Low	Low	Low	Low
Fujita, 2006228	Unclear	Unclear	Low	Low	High	Low	Unclear
Gao, 2010154	Low	Low	Low	Low	Low	Low	Low
Gjerdingen, 2002181	Low	Unclear	Unclear	Low	Unclear	High	High
Gunn, 1998225	Low	Low	Low	Low	Low	Low	Low
Harrison-Hohner, 2001208	Low	Low	Unclear	Unclear	High	High	High
Hayes, 2001182	Low	Low	Low	Unclear	Low	Low	Unclear
Ho, 2009183	High	High	Low	Low	Low	Low	Unclear
Hodnett, 2002197	Low	Low	Low	Unclear	Low	Low	Low
Kieffer, 2013198	Low	Low	Low	Low	Low	Low	Low
Kozinsky, 2012155	High	High	Unclear	Unclear	Unclear	High	High
Lawrie, 1998209	Low	Low	Low	Low	Low	Low	Low
Leung, 2012156	Low	Low	Low	Low	Low	Low	Low
Llorente, 2003210	Low	Low	Low	Unclear	Unclear	Low	Unclear
Lumley, 2006147	Low	Unclear	Low	Low	Unclear	Low	Low
MacArthur, 2002146	Low	Low	Low	Low	Low	Low	Low
Makrides, 2010211	Low	Low	Low	Low	Low	Low	Low
Mao, 2012157	Low	Low	Low	Low	Low	Low	Unclear
Matthey, 2004184	Unclear	Unclear	Low	Low	Low	Low	Unclear
Milgrom, 2011185	Low	Low	Low	Unclear	Unclear	Low	Low
Mokhber, 2011212	Unclear	Unclear	Low	Low	Unclear	Low	Unclear
Morrell, 2000199	Low	Low	Low	Low	Low	Low	Low
Morrell, 200961	Low	Low	Low	Low	Low	Low	Low
Norman, 2010123	Low	Low	Low	Low	Unclear	Unclear	Unclear
Priest, 2003217	Low	Unclear	Low	Low	Low	Low	Low
Reid, 2002200	Low	Low	Low	Low	Low	Low	Low
Rotheram-Borus, 2011226	Low	Low	Low	Low	Low	Low	Low
Sealy, 2009186	High	High	Low	Low	Unclear	Low	Unclear
Selkirk, 2006218	High	High	High	Low	High	Low	Low
Serwint, 1991227	High	High	Low	Low	Low	Low	Low
Shapiro, 2005187	Unclear	Unclear	Low	High	High	Low	Unclear
Shields, 1997219	Low	Unclear	Low	Low	Unclear	Unclear	Low
Songøygard, 2012129	Low	Low	Low	Low	Unclear	Unclear	Unclear
Waldenstrom, 2000220	Low	Low	Low	Low	Unclear	Low	Low

TABLE 2 - Risk of bias for included selective preventive intervention RCTs: summary judgments about each risk-of-bias item

Key: high, high risk of bias; ID, identification; ITT, intention to treat; low, low risk of bias; PP, per protocol; unclear, unclear risk of bias.

First author, year, reference number	Selection bias		Performance bias	Detection bias	Attrition bias	Reporting bias	Other bias
	Random sequence generation	Allocation concealment	Blinding of participants/personnel	Blinding of outcome assessors	Incomplete outcome data	Selective reporting of the outcome, subgroups, or analysis	Funding source, adequacy of statistical methods used, type of analysis (ITT/PP), baseline imbalance in important characteristics
Barnes, 2009149	Unclear	Unclear	Low	Low	Unclear	Low	Unclear
Brugha, 2000188	Low	Unclear	Low	Unclear	Low	Low	Low
Buist, 1999189	Unclear	Unclear	Low	Low	Unclear	Unclear	Unclear
Chabrol, 2002158	Unclear	High	Low	Unclear	Unclear	Low	Unclear
Cupples, 2011201	Low	Low	Low	Low	Low	Low	Low
Gamble, 2005221	Low	Low	Low	Low	Low	Low	Low
Hagan, 2004159	Low	Low	Low	Low	Low	Low	Low
Harris, 2002213	Low	Unclear	Low	Low	Unclear	Unclear	Unclear
Howell, 2012190	Low	Low	Low	Low	Low	Low	Low
Ickovics, 2011222	Low	Low	Low	Low	Low	Low	Low
Logsdon, 2005202	Low	Unclear	Unclear	Low	Unclear	Low	Unclear
Phipps, 2013160	Low	Low	Low	Low	Low	Low	Low
Richter, 2014203	Low	Low	Low	Unclear	High	Unclear	Unclear
Sen, 2006191	Low	Low	Low	Low	Low	Low	Low
Silverstein, 2011161	Low	Low	Low	Low	Low	Low	Low
Small, 2000223	Low	Low	Low	Low	Low	Low	Low
Tam, 2003162	Low	Low	Unclear	Unclear	Unclear	Low	Unclear
Walkup, 2009192	Low	Low	Low	Low	Unclear	Low	Low
Wolman, 1993204	Low	Low	Low	Unclear	Unclear	Low	Low
Zlotnick, 2011163	Low	Low	Low	Unclear	Low	Low	Low

TABLE 3 - Risk of bias for included indicated preventive intervention RCTs: summary judgements about each risk-of-bias item

Key: high, high risk of bias; ID, identification; ITT, intention to treat; low, low risk of bias; PP, per protocol; unclear, unclear risk of bias.

First author, year, reference number	Selection bias		Performance bias	Detection bias	Attrition bias	Reporting bias	Other bias
	Random sequence generation	Allocation concealment	Blinding of participants/personnel	Blinding of outcome assessors	Incomplete outcome data	Selective reporting of the outcome subgroups or analysis	Funding source adequacy of statistical methods used type of analysis (ITT/PP) baseline imbalance in important characteristics
Armstrong, 1999164	Low	Low	Low	Low	Low	Low	Low
Austin, 2008165	Low	Unclear	Low	Low	Unclear	Unclear	Unclear
Crockett, 2008166	Unclear	Unclear	Low	Unclear	Low	Low	Unclear
Dennis, 2009205	Low	Low	Low	Low	Low	Low	Low
El-Mohandes, 2008167	Low	Low	Low	Low	Low	Low	Low
Ginsburg, 2012168	Unclear	Unclear	Low	Unclear	Unclear	Low	Unclear
Gorman, 1997169	Unclear	Unclear	Unclear	Low	Low	Low	Low
Grote, 2009170	Low	Unclear	Low	Unclear	Low	Low	Low
Harris, 2006206	Low	Low	Low	Unclear	Unclear	Unclear	Unclear
Heh, 2003193	High	High	Low	Low	Low	Low	Unclear
Lara, 2010194	Low	Low	Low	Unclear	High	Low	High
Le, 2011171	Low	Low	Low	Unclear	Low	Low	Low
Manber, 2004229	Unclear	Unclear	Unclear	Unclear	Unclear	Low	Unclear
Marks, 2003224	Low	Low	Unclear	Unclear	Low	Low	Unclear
McKee, 2006172	Unclear	Unclear	Low	Unclear	High	Low	Unclear
Morrell, 200961	Low	Low	Low	Low	Low	Low	Low
Mozurkewich, 2013214	Low	Low	Low	Low	Low	Low	Low
Munoz, 1998173	Low	Low	Low	Unclear	Low	Low	Low
Petrou, 2006174	Low	Low	Low	Low	Low	Low	Low
Rahman, 2008148	Low	Low	Low	Low	Low	Low	Low
Stamp, 1995195	Low	Low	Low	Low	Low	Low	Low
Tiwari, 2005175	Low	Low	Low	Low	Low	Low	Low
Vieten, 2008121	Unclear	Unclear	Low	Low	Unclear	High	Unclear
Webster, 2003196	Low	Low	Low	Low	Unclear	Low	Low
Weidner, 2010176	Low	High	Low	Low	Unclear	Low	Low
Wilson, 2013177	Low	High	Low	Unclear	Unclear	Low	Low
Wisner, 2001215	Low	Low	Low	Low	Low	Low	Low
Wisner, 2004216	Low	Low	Low	Low	Low	Low	Low
Zlotnick, 2001178	Unclear	Unclear	Low	Unclear	Low	Low	Low
Zlotnick, 2006179	Low	Unclear	Low	Unclear	Low	Low	Low

FIGURE 3.

Risk-of-bias graph for all included RCTs: authors’ judgements about each risk-of-bias item. Key: ITT, intention to treat; NA, not applicable; PP, per protocol.

Quality of systematic and other reviews

No quality assessment was undertaken for the systematic reviews.

Qualitative studies level of preventive intervention

Among the 21 studies (27 citations):

• Fourteen were studies of a universal preventive intervention. Twelve studies reported qualitative data on the perspectives and attitudes of those who had received universal preventive interventions for PND277–289 (of these, two also reported perspectives and attitudes of service providers on universal preventive interventions287,288 and two studies reported only on the perspectives and attitudes of service providers to preventive interventions). 290,291

• Four studies presented data from those who had received a selective preventive intervention292–298 (with one study additionally presenting data relating to an indicated population). 296–298 Of these, one study also reported perspectives and attitudes of service providers on selective preventive interventions. 296–298

• Three studies presented data from those who had received an indicated preventive intervention. 253,256,299,300 One study, with a separate citation,301 additionally reported on the perspectives of and attitudes of service providers on indicated preventive interventions.

These data are synthesised in Chapters 5–7. The remaining 29 (23 studies) citations about PSSSs that women believed helped prevent PND are synthesised in Chapter 8, the realist synthesis and are presented separately (see Appendix 7).

Qualitative review study characteristics

Qualitative review study characteristics: personal and social support strategy studies

A total of 23 studies (n = 29 citations) reported qualitative data on perspectives and attitudes of women who had not experienced PND, regarding PSSSs that they believe helped them to prevent the condition. 7,302–325 This included five citations from three intervention studies which included PSSS evidence. 286,292,296–298

Quality of the qualitative intervention studies

All studies met the requirement to report either qualitative research or qualitative data within mixed-methods studies, indicated in Table 4. All included studies also adequately described the context and aims of the study. Few (29%) demonstrated evidence of researcher reflexivity (i.e. awareness of the researcher’s contribution to the construction of meanings throughout the research process and an acknowledgement of the impossibility of remaining ‘outside’ one’s subject matter while conducting research); among those which did, these descriptions were brief. A number of studies illustrated that reflexivity in the research process had been incorporated, such as making changes to the interview guide as necessary and responding to participants’ wishes. All 21 studies provided adequate descriptions of recruitment methods, just over half (n = 13) provided adequate descriptions of data collection methods, although such descriptions tended to be brief. The study methods used involved interview methods in nine studies, supplemented by other methods such as focus groups and observation in three studies. Qualitative data came from open-ended questions as part of a questionnaire in three studies. Two studies used focus groups and one study used online messages. Twelve studies provided an adequate description of data analysis methods, and 13 studies provided sufficiently in-depth, detailed and rich data. The absence of detail in the remaining studies may have been, in part, because of limitations imposed by journal reporting requirements.

Certainty of the review findings: intervention studies

The CERQual approach137 was used to assess the certainty of the review findings, graded as ‘very low’, ‘low’, ‘moderate’ or ‘high’. A summary assessment was based on the CASP quality assessment finding,136 the number and richness of the study data, the consistency of the data across study settings and populations, and the relevance of the finding to the review question. There were 37 findings in women’s evidence: nine were assessed as of moderate certainty, 25 as low and three as of very low certainty. For service providers’ evidence, there were 25 findings: one finding was assessed as being of moderate certainty, 18 as of low certainty and six as of very low certainty. No findings were assessed as high certainty.

Overview of main findings from qualitative intervention studies (all levels)

Quality of the qualitative personal and social support strategy studies

As shown in Table 12, all included studies reported either qualitative research or qualitative data within mixed-methods studies. All included studies adequately described the context and aims of the study. Sixteen of the 23 studies demonstrated evidence of researcher reflexivity. As in the intervention studies, few PSSS studies made explicit reference to researcher reflexivity, and in those which did descriptions were often brief. Most studies adequately described recruitment methods (n = 21) and data collection methods (n = 21), although such descriptions tended to be brief. Eighteen studies used interview methods, four used focus groups and one study used an online survey. Eighteen of 23 of studies provided an adequate description of data analysis methods, and 20 of the 23 studies provided sufficiently in-depth, detailed and rich data.

Qualitative studies further analysis by level of preventive intervention: universal, selective and indicated

Further quantitative and qualitative results are presented in Chapter 5 (37 universal preventive interventions trials, 14 qualitative studies), Chapter 6 (20 selective preventive interventions trials, four qualitative studies) and Chapter 7 (30 indicated preventive interventions trials, three qualitative studies). One study presented two levels of intervention and analysis. 61

Description of qualitative studies of universal preventive interventions

There were 14 studies relating to 15 citations reporting qualitative data on universal preventive interventions for PND. 277–291 Twelve studies relating to 13 citations reported the perspectives and attitudes of women who had received an intervention. 277–289 Four studies reported perspectives and attitudes of service providers of universal preventive interventions. 286,287,290,291

The qualitative studies related to five of the seven intervention classes:

1. psychological (n = 1)282

2. social support (n = 2)281,287

3. midwifery-led interventions (n = 7)277,283–285,289–291

4. organisation of maternity care (n = 1)288

5. CAM or other (n = 3). 278–280,286

For ease of reference the universal preventive interventions have been given short-version descriptive labels (Table 13).

Characteristics and main outcomes of randomised controlled trials of universal preventive interventions of psychological interventions

Of the 37 universal preventive interventions, six studies evaluated a psychological intervention,61,153–157 including promotion of parent–infant interaction,153 psychoeducation,155 IPT,154,157 CBT-based interventions61,157 and PCA. Comparators included usual care in specific countries61,153,154,156,157 and educational information. 155 Three studies provided the intervention using a group format,154–156 whereas two involved individual sessions61,153 and one combined both group and individual sessions. 157 Two trials provided interventions in the home setting,61,153 whereas in the other four trials the intervention was provided in the antenatal setting. 154–157 Three trials provided the intervention in the antenatal period only,155–157 whereas one trial initiated the intervention postnatally61 and two trials provided the intervention across the perinatal period both during pregnancy and following childbirth. 153,154 The interventions were provided by different health-care providers including community workers,153 midwives,154 health visitors,155 psychologists156 and obstetricians,157 with the number of intervention contacts ranging from two154 to 16153 and with the duration of contact ranging from 1 to 2 hours.

A summary of the characteristics and main outcomes is provided in Table 14. In the psychological intervention trials, PND was assessed using various measures including the EPDS,61,153,154,156,157 General Health Questionnaire,154 Leverton Questionnaire,155 the Short Form questionnaire-36 items (SF-36) mental component summary (MCS),328 the Clinical Outcomes in Routine Evaluation-Outcome Measure (CORE-OM),151 the PHQ-9157 and the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders (SCID). 157 The State–Trait Anxiety Inventory (STAI) was the only anxiety outcome reported151 and the Perceived Stress Scale (PSS) the only stress outcome. 156 Family outcomes included the Dyadic Adjustment Scale,61 the Parenting Stress Index (PSI),151 the Relationship Efficacy Measure,156 the Satisfaction with Interpersonal Relationships Scale154 and mother–infant interaction (maternal sensitivity and intrusiveness). 153 The three infant outcomes reported were the Ainsworth Strange Situation Assessment of Infant Attachment,153 the Behaviour Screening Questionnaire61 and the Checklist for Autism in Toddlers. 61 General health and other outcomes included the Short Form questionnaire-12 items (SF-12), the SF-36 physical component summary (PCS), the Short-Form questionnaire-6 Dimensions (SF-6D)61 and the Subjective Happiness Scale. 156

Description and findings from qualitative studies of universal preventive interventions of psychological interventions

A description of the qualitative study evaluating a psychological intervention is provided in Table 15.

Characteristics and main outcomes of randomised controlled trials of universal preventive interventions of educational interventions

Eight included studies evaluated an educational intervention for the universal prevention of PND, none of which were conducted in the UK. 180,182–187 Two main types of educational interventions were identified: education on preparing for parenting180,181,184,185,187 and various advisory booklets on PND. 182,183,186 Comparisons were made with usual care in specific countries180–184,186 and educational information. 185 Four trials provided the educational intervention via group format,180,181,184,187 while the remaining four trials involved individual sessions. 182,183,185,186 Only one trial provided the intervention in the home setting. 186 Four trials provided the intervention in the antenatal period only,181,182,185,187 whereas two trials initiated the intervention postnatally183,186 and two trials provided the intervention across the perinatal period both during pregnancy and following childbirth. 180,184 The interventions were provided by psychologists,180,181,184,185 midwives182 and nurses. 183,186 The number of contacts ranged from two to eight and the duration of contact ranged from 30 minutes to 4 hours.

A summary of the characteristics and main outcomes is provided in Table 16. No qualitative studies were found for educational interventions as a universal preventive intervention.

Characteristics and main outcomes of randomised controlled trials of universal preventive interventions of social support

Of the 37 RCTs of universal preventive interventions, four (11%) evaluated a social support intervention,197–200 two of which were conducted in the UK. 199,200 Several types of social support were identified including support in labour197 and self-help support. 200 Comparisons were made with usual care in specific countries197,199,200 and educational information. 198 One intervention involved a group session,200 two studies involved individual sessions197,199 and one involved both group and individual sessions. 198 One study took place in the home setting. 199 None of the studies were undertaken in the antenatal period only, two were in the postnatal period only,199,200 two were in a combination of both antenatal and postnatal periods198 and one was at the stage of labour. 197 As with other types of included interventions, these were provided by different health-care providers: community workers,198 midwives,200 nurses197 and support workers. 198 The number of contacts ranged from one to 14 but duration of contact (10–378 minutes) was reported in only one study. 199

A summary of the characteristics and main outcomes is provided in Table 17.

Description and findings from qualitative studies of universal preventive interventions of social support

Of the two qualitative studies of social support included in the universal preventive interventions category, one was conducted in the UK287 and one in Canada. 281 One intervention was an online discussion group281 and the other a postnatal support worker intervention. 287 Further details are provided in Table 18.

Characteristics and main outcomes of randomised controlled trials of universal preventive intervention of pharmacological agents or supplements

Of the six trials that evaluated a specific supplement or drug for the universal prevention of PND, none were conducted in the UK. 207–212 Several types of pharmacological agents or supplements were identified including docosahexaenoic acid (DHA) at different doses,207,210,211 calcium,208 norethisterone ethanate209 and selenium. 212 All six studies compared the interventions with usual care in specific countries. 207–212 All six studies involved individual sessions. 207–212 Three studies took place in the antenatal period208,211,212 and two in the postnatal period;209,210 one combined both antenatal and postnatal periods. 207 Included interventions were all delivered by the provider. A summary of the characteristics and main outcomes is provided in Table 19. No qualitative studies were identified of pharmacological agents or supplements aimed at a universal population.

Characteristics and main outcomes of randomised controlled trials of universal preventive interventions of midwifery-led interventions

Of the five studies146,217–220 evaluating midwifery-led interventions for the universal prevention of PND, two were conducted in the UK. 146,219 Several types of midwifery-led interventions were identified, including midwifery redesigned postnatal care,146 midwife-led debriefing or counselling after childbirth,217,218 midwife-managed care219 and team midwife care. 220 Comparisons were made with usual care in specific countries. 146,217–220 All six trials involved individual sessions. 146,217–220 None of the trials provided the intervention only antenatally; three initiated the intervention postnatally,146,217,264 and two trials initiated the intervention during the pregnancy and continued it postnatally. 219,220 The provision of the midwifery care varied in the number of contacts, with duration ranging from 15 minutes to 1 hour. A summary of the characteristics and main outcomes is provided in Table 20.

Description and findings from qualitative studies of universal preventive interventions of midwifery-led interventions

Seven qualitative reports were identified of women’s experiences of midwifery-led interventions. 277,283–285,289–291 All seven reports related to the CenteringPregnancy initiative (Table 21).

Characteristics and main outcomes of randomised controlled trials of universal preventive interventions of organisation of maternity care

Of the five studies147,150,225–227 evaluating the organisation of maternity care for the universal prevention of PND, only one was conducted in the UK. 150 Several types of organisation of maternity care interventions were identified, including planned health visitor visits,150 early contact with care provider,225 primary care and community care strategies,147 programmes for HIV, alcohol and mental health226 and early contact with care provider. 227 Comparisons were made with usual care in specific countries. 147,225–227 All five studies involved individual contacts. 147,150,225–227

None of the studies were undertaken in only the antenatal period, three in only the postnatal period150,225,227 and two used a combination of both antenatal and postnatal periods. 226,227 The interventions were provided by different health-care providers including health visitors,150 GPs,225 primary care nurses, community development workers,147 community health workers, peer mentors,226 paediatric house officer or nurse practitioners. 227 The number of contacts varied greatly. A summary of the characteristics and main outcomes is provided in Table 22.

Description and findings from qualitative studies of universal preventive interventions of organisation of maternity care

One qualitative study288 reported women’s experiences of interventions aimed at a universal population involving the organisation of maternity care (Table 23).

Characteristics and main outcomes of randomised controlled trials of universal preventive interventions of complementary and alternative medicine or other

None of the three studies123,129,228 evaluating the CAMs for the universal prevention of PND were conducted in the UK. Several types of CAMs interventions were identified including baby massage228 and exercise. 123,129 Comparisons were made with usual care in specific countries129,228 and educational information. 123 All three studies involved individual sessions. One was undertaken in the antenatal period only129 and two in the postnatal period only. 123,228 The provider of these interventions was a massage instructor228 or physical therapist. 123,129 The number of contacts varied and the length of contact was 1 hour in two studies. 123,129 A summary of the characteristics and main outcomes is provided in Table 24.

Description and findings of qualitative studies of universal preventive interventions of complementary and alternative medicine or other

Three qualitative studies278–280,286,288 reported women’s experiences of interventions aimed at a universal population involving the CAMs or other intervention (Table 25).

Results for universal preventive interventions for Edinburgh Postnatal Depression Scale threshold score at 6 weeks postnatally

Data were available from five studies presenting the EPDS threshold score at 6 weeks postnatally. 197,208,211,219,220 The results for the five universal preventive intervention trials presenting an EPDS threshold score are combined here. A NMA compared the effects of support in labour,197 midwife-managed care,219 DHA,211 calcium208 and team midwife care220 relative to usual care on EPDS threshold. Figure 5 presents the network of evidence. There were five intervention effects (relative to usual care) to estimate from five studies.

FIGURE 5.

Universal preventive interventions EPDS threshold score at 6 weeks postnatally: network of evidence.

Figure 6 presents the odds ratios of each intervention relative to usual care and Figure 7 presents the probabilities of treatment rankings. The total residual deviance was 10.04, which is compared with the total number of data points, 10, included in the analysis. This implies a good fit of the model to the data. The between-study SD was estimated to be 0.26 (95% CrI 0.01 to 0.72), which implies mild heterogeneity of intervention effects between studies.

FIGURE 6.

Universal preventive interventions EPDS threshold score at 6 weeks postnatally: odds ratios for all treatment comparisons. Key: OR, odds ratio.

FIGURE 7.

Universal preventive interventions EPDS threshold score at 6 weeks postnatally: probability of treatment rankings (ranks 1–6).

For all interventions except midwifery team care the odds ratio was less than 1, suggesting a beneficial effect compared with usual care. However, none of the comparisons were statistically significant at a conventional 5% level (see Figure 6). The interventions with the highest probabilities of being the best were midwife-managed care and calcium (probability 0.43 and 0.36, respectively).

Midwifery team care was associated with an increased odds ratio compared with usual care (1.39, 95% CrI 0.65 to 3.01) and had a 74% chance of being the least effective among the six interventions (see Figure 7).

Results for universal preventive interventions for Edinburgh Postnatal Depression Scale threshold score at 3 months postnatally

A NMA was used to compare the effects of calcium,208 booklet on PND,186 exercise129 and early contact with care provider225 relative to usual care on EPDS threshold. Data were available from four studies comparing five interventions. Figure 8 presents the network of evidence. There were four intervention effects to estimate from four studies. 129,186,208,225

FIGURE 8.

Universal preventive interventions EPDS threshold score at 3 months postnatally: network of evidence.

Figure 9 presents the odds ratios of each intervention relative to usual care and Figure 10 presents the probabilities of treatment rankings. The total residual deviance was 7.04, which is compared with the total number of data points, seven, included in the analysis. This implies a good fit of the model to the data. The between-study SD was estimated to be 0.23 (95% CrI 0.00 to 0.74), which implies mild heterogeneity between studies in intervention effects.

FIGURE 9.

Universal preventive interventions EPDS threshold score at 3 months postnatally: odds ratios for all treatment comparisons. Key: OR, odds ratio.

FIGURE 10.

Universal preventive interventions EPDS threshold score at 3 months postnatally: probability of treatment rankings (ranks 1–5).

The odds ratios of calcium, a booklet on PND and exercise were less than 1, suggesting a beneficial effect compared with usual care. Early contact with care provider had an odds ratio greater than 1, suggesting a worsening effect (see Figure 9). However, only the effect of calcium was statistically significant at a conventional 5% level. The interventions with the highest probabilities of being the best were calcium and booklet on PND (probability 0.48 and 0.45, respectively).

Results for universal preventive interventions for Edinburgh Postnatal Depression Scale threshold score at 6 months postnatally

A NMA was used to compare the effects of DHA,211 CBT-based intervention,61 PCA-based intervention,61 primary care and community care strategies147 and early contact with care provider225 relative to usual care on EPDS threshold. Data were available from four studies comparing six interventions. Figure 11 presents the network of evidence. There were five intervention effects to estimate from four studies. 61,147,211,225

FIGURE 11.

Universal preventive interventions EPDS threshold score at 6 months postnatally: network of evidence. Dashed lines represent three-arm trials.

Figure 12 presents the odds ratios of each intervention relative to usual care and Figure 13 presents the probabilities of treatment rankings. The total residual deviance was 7.04, which is compared with the total number of data points, seven, included in the analysis. This implies a good fit of the model to the data. The between-study SD was estimated to be 0.22 (95% CrI 0.00 to 0.71), which implies mild heterogeneity of intervention effects between studies.

FIGURE 12.

Universal preventive interventions EPDS threshold score at 6 months postnatally: odds ratios all treatment comparisons. Key: OR, odds ratio.

FIGURE 13.

Universal preventive interventions EPDS threshold score at 6 months postnatally: probability of treatment rankings (ranks 1–6).

For all interventions except primary care and community care strategies the odds ratio compared with usual care was less than 1, suggesting a beneficial effect. However, none of the comparisons were statistically significant at a conventional 5% level (see Figure 12). The interventions with the highest probabilities of being the best were CBT-based intervention and PCA-based intervention (probabilities 0.38 and 0.29, respectively) (see Figure 13).

Results for universal preventive interventions for Edinburgh Postnatal Depression Scale threshold score at 12 months postnatally

A NMA was used to compare the effects of CBT-based intervention,61 PCA-based intervention61 and midwifery redesigned postnatal care146 relative to usual care on EPDS threshold. Data were available from two studies comparing four interventions. Figure 14 presents the network of evidence. There were three intervention effects to estimate from two studies. 61,146

FIGURE 14.

Universal preventive interventions EPDS threshold score at 12 months postnatally: network of evidence.

All three interventions were associated with a beneficial effect, although the results were statistically inconclusive (Figure 15). The probability of the intervention being the best was 0.43, 0.32 and 0.25 for midwifery redesigned postnatal care,146 CBT-based intervention61 and PCA-based intervention,61 respectively (Figure 16).

FIGURE 15.

Universal preventive interventions EPDS threshold score at 12 months postnatally: odds ratios for all treatment comparisons. Key: OR, odds ratio.

FIGURE 16.

Universal preventive interventions EPDS threshold score at 12 months postnatally: probability of treatment rankings (ranks 1–4).

Summary of results from network meta-analysis for universal preventive interventions for Edinburgh Postnatal Depression Scale threshold score

In general, the intervention effects were inconclusive, although calcium was associated with a statistically significant benefit relative to usual care at 3 months. Intervention effects tended to vary over time. The interventions most likely to be the best among those evaluable at each assessment were:

• at 6 weeks postnatally, midwife-managed care219 and calcium208 (the included studies were of unclear and high risk of bias, respectively)

• at 3 months postnatally, booklet on PND186 and calcium208 (the included studies were both at high risk of bias)

• at 6 months postnatally, CBT-based intervention61 and PCA-based intervention61

• at 12 months postnatally midwifery redesigned postnatal care,146 CBT-based intervention61 and PCA-based intervention. 61

However, there was considerable uncertainty associated with the results and none of the probabilities of being the best intervention exceeded 0.5.

A weakly informative prior distribution was used for the between-study SD because there were insufficient studies with which to estimate it from the sample data alone. This prior distribution was chosen to ensure that, a priori, 95% of the study-specific odds ratios were within a factor of 2 of the median odds ratio for each treatment comparison. The sensitivity analysis is presented for completeness in Appendix 11.

Summary of results from network meta-analysis for universal preventive intervention studies for Edinburgh Postnatal Depression Scale mean scores

Not all studies provided information about intervention effects at each time, making it difficult to draw inferences across all interventions at each time. In general, the intervention effects were inconclusive and the CrIs were wide. Intervention effects tended to vary over time. The interventions most likely to be the best among those evaluable at each assessment were:

• 6–8 weeks postnatally: selenium212 (the risk of bias for this study was unclear so the benefit of selenium estimated in this NMA should be treated with some caution)

• 3 months postnatally: midwifery redesigned postnatal care146

• 6 months postnatally: CBT-based intervention61 and PCA-based intervention61

• 12 months postnatally: midwifery redesigned postnatal care,146 CBT-based intervention61 and PCA-based intervention. 61

Overall summary of results for universal preventive interventions for Edinburgh Postnatal Depression Scale threshold and Edinburgh Postnatal Depression Scale mean scores

Table 26 indicates the results of the NMAs for the EPDS threshold scores and EPDS mean scores at all assessment times. The results indicate that the universal preventive intervention with the best-quality evidence and the most enduring effect were midwifery redesigned postnatal care,146 CBT-based intervention61 and PCA-based intervention. 61

The qualitative data indicated that women appreciated the benefits from IPT, the reassurance and normalisation of social support, and the support received from peers while taking part in midwifery-led interventions and group-based CAM interventions.

Description of qualitative studies of selective preventive interventions

The qualitative synthesis identified four studies in the selected preventive interventions group, within three of the seven intervention classes:

1. psychological (n = 1)296–298

2. educational (n = 1)253,254

3. midwifery-led interventions (n = 2). 292–294

For ease of reference the selective preventive interventions have been given short-version descriptive labels (Table 27).

Characteristics and main outcomes of randomised controlled trials of selective preventive interventions of psychological interventions

Of the six included selective preventive intervention trials evaluating a psychological intervention,158–163 none were conducted in the UK. Three types of psychological interventions were evaluated: psychoeducational therapy,162 IPT160,163 and CBT. 158,159,161 Comparisons were made with usual care in specific countries158,159,161–163 and educational information. 160 One trial provided the intervention in a group format159 and five trials incorporated individual sessions. 158,160–163 None of the interventions were provided in the home setting. One trial provided the intervention in the antenatal period only,160 whereas three trials initiated the intervention postnatally159,161,162 and two trials provided the intervention across the perinatal period from pregnancy to after childbirth. 158,163 Interventions were provided by a variety of service providers. The number of contacts ranged from one to six (mean 4.3) and contact duration ranged from 25 minutes to 2 hours.

A summary of the characteristics and main outcomes is provided in Table 28.

Description and findings from qualitative studies of selective preventive interventions of psychological interventions

There was one US-based study reporting qualitative data on selective preventive interventions for PND. 296–298 The IPT intervention for teenagers promoted support from peers and clinicians, and participants were able to gain practical skills and felt empowered (Table 29).

Characteristics and main outcomes of randomised controlled trials of selective preventive intervention of educational interventions

Of five included trials of a selective preventive intervention evaluating an educational intervention,188–190,192,271 two were conducted in the UK. 188,191 Two main types of interventions were identified: education on preparing for parenting188,189,192,271 and a booklet on PND and social worker telephone call. 190 Comparisons were made with usual care in specific countries. 188,189,192,271 One study evaluated the effect of group sessions,188 two studies evaluated the effect of individual sessions only190,192 and two studies evaluated a combination of individual and group sessions. 189,191 One trial provided the intervention in a home setting. 192 Three trials provided the intervention in the antenatal period only,188,189,191 whereas one trial initiated the intervention in the postnatal period190 and one trial provided the intervention across the antenatal and postnatal periods. 192 The interventions were provided by a variety of service providers, with the number of contacts ranging from 1 to 25 (mean 9.6 contacts) and the duration varying between 1 and 2 hours. A summary of the characteristics and main outcomes is provided in Table 30.

Description and findings from qualitative studies of selective preventive interventions of educational interventions

One qualitative study of an educational intervention was included in the indicated preventive interventions category. 253,254 This study was linked to the trial of education on preparing for parenting. 188 Further details are provided in Table 31.