Psychological interventions for postnatal depression: cluster randomised trial and economic evaluation. The PONDER trial

Depression

Mental health is considered to be ‘a state of well-being in which the individual realises his or her own abilities, can cope with the normal stresses of life, can work productively and fruitfully, and is able to make a contribution to his or her community’. 10

Conversely, mental ill health covers mental health problems, strain, impaired functioning associated with distress, symptoms and diagnosable mental disorders such as schizophrenia and depression. 11 Each year, more than a quarter of European adults will experience mental ill health of one form or another,12 most commonly depression.

Postnatal mental health

Most women feel exhausted after the birth of their baby and will be tearful and feel low because of the exhaustion. There continues to be some debate about the classification of postnatal mental health conditions and whether PND exists as a unique diagnosis or whether depression occurring postnatally is just coincidental. 13 However, both of the internationally used classification systems no longer provide a separate category of PND. In effect, the presence of depression is determined by the same set of criteria, regardless of timing or context. When mood states were measured in a sample of pregnant women and a group of matched non-childbearing women,14 there were no differences between the two groups in rates of major or minor depression after the babies were born, but the postnatal women had more symptoms of depression. 14 Women do experience postnatal distress and less satisfaction in their relationships at this time, especially with their partners. 14,15 It has been proposed that the depressive symptoms or distress that some women experience are an appropriate response following childbirth and so should not be confused with clinical depression. 16 Any emphasis on depression might draw attention away from the social and cultural context of parenting and the changes and losses that accompany the birth of a baby and consequently the sharing of the responsibility for the distress that the women experience. 16

There is some evidence that women are more vulnerable to depression within the first 6 postnatal months, not just the first few postnatal weeks. 17 Because of issues of context and in particular the welfare of the infant, and other family members, health professionals need to be aware of the postnatal onset of depression, puerperal psychosis, post-traumatic stress disorder, panic disorder and relapse of other illnesses such as schizophrenia. 18

Within the first days after the birth of a baby, 39–85% of women feel more emotional than usual, weepy, irritable and anxious and have insomnia and a low mood because of what is called postnatal ‘blues’ or baby ‘blues’. 14 Women can be given information about symptoms and reassurance that postnatal blues resolve quickly within a few days, without treatment, and can be advised about self-help. However, there is no evidence for the effectiveness of these measures. 18

At the other extreme, puerperal psychosis is a severe mental illness affecting one or two per thousand women soon after delivery. 19 Women with a history of a postnatal mood disorder carry a very high risk of recurrence. The dramatic symptoms are severe depression with a risk of suicide or even infanticide. 20 Mania, hallucinations or delusions require urgent psychiatric treatment, often as an inpatient. This very small but important risk of suicide and infanticide in some severely depressed mothers manifests itself in violent methods, more often than in the population generally, sometimes before and sometimes after 6 weeks postnatally. 21 Although exceptionally serious, statistically suicide is rare and most HVs will never encounter maternal suicide.

In 1992 the International Classification of Diseases (ICD-10)22 first included an optional supplementary code for diseases that occur during and complicate pregnancy, childbirth or the puerperium (the O99 code can be applied to any form of mental disorder).

The American Psychiatric Association Diagnostic and Statistical Manual of Mental Disorders (DSM-IV)23 makes provision for a postpartum-onset specifier code that can be added to a diagnosis of manic depression, bipolar disorder or major depression among others, provided onset is within 4 weeks of childbirth. No reason for the 4-week timing is provided and the ICD does not suggest a time period for onset. The diagnosis of depression (irrespective of the gender of the sufferer or the timing of the episode) relies on the presence of at least five of the following symptoms for at least 2 continuous weeks:23

• depressed mood

• loss of interest or pleasure

• significant increase or decrease in appetite

• insomnia or hypersomnia

• psychomotor agitation or retardation

• fatigue or loss of energy

• feelings of worthlessness or guilt

• diminished concentration

• recurrent thoughts of suicide.

Depression is far more common than psychosis at any time as well as in the context of pregnancy and childbirth. Depression can last for up to 1 year after delivery in about 4% of all mothers20 or a quarter of mothers who become depressed, and may last even longer. 24 But the relevance of this is not clear as not many studies have followed up women for long enough to determine the depression duration and no study has described in a standardised way the course of depression (in women) with respect to whether or not the episode occurred within the context of childbirth.

The proportion of postnatal women who might be depressed varies between 11%20 and 22%25,26 depending on the sample of women, the criteria used and the time of assessment postnatally. 14 Based on a meta-analysis of estimates from 59 studies internationally, the average prevalence of depression postnatally is 13%. 27 A later meta-analysis estimated that 14.5% of women may have a new episode of major or minor depression during the first 3 postnatal months, with 6.5% having a new episode of major depression. 28 The same review estimated a prevalence of 6.5–12.9% for major and minor depression at any time during the first postnatal year, and a 1–5.9% prevalence of major depression. 28

Consequences of depression

Depression is a public health problem with financial costs to the gross domestic product associated with sickness absence,29 mainly through lost productivity. There is a high risk of relapse. 30 The UK spends 12% of its total health expenditure on mental health,11 and the cost for antidepressant prescriptions is around £401 million. 31 The costs of the stigma and discrimination associated with mental ill health remain intangible.

Depression can lead to more deaths from suicide each year than there are deaths from road accidents. Suicide rates and mental health states vary across European countries, reflecting their diverse traditions, cultures, situations and religious variations in reporting suicide. 11 The number of deaths from psychiatric illness is underestimated as suicidal deaths may not be classified as such, to spare family feelings. 32 The sixth report of the confidential enquiries into maternal deaths in the UK, Why Mothers Die,21 reported suicide as the most common cause of maternal death for women in the first year after childbirth.

The natural history of PND varies among women, but around one-third develop a chronic problem with long-term adverse consequences. Although there is little evidence to date, there is a belief that women’s depression may affect their partners,33 who become depressed,34–36 thereby reducing their ability to cope with supporting the mother or caring for the new infant or other children.

There has been growing concern from the literature on the evidence of the effects that depression might have on the cognitive37 and emotional development of children38 and the attachment of infants to their mothers, particularly for boys, possibly well beyond infancy. 39 Boys whose mothers are depressed in the first year may have particular problems with reading. 40 Infants are highly sensitive to the quality of their interpersonal contacts, which are most often provided by the mother in the first few months of life. 39 This could be because the baby has a rapidly developing brain in the first 6 months of life and is heavily dependent on external stimulation and therefore particularly vulnerable during this sensitive period. 39,40 It is also possible that the association between PND and the development of infants represents a complex two-way interaction. 41 Also, mothers with depression are more likely to report parenting stress, negative perceptions of their infant’s behaviour and hostile feelings towards their infant. 42

Aware of the link, the European Union’s project on building mental health in infants, children and adolescents recognised the need to address PND. 11 Similarly, guidance from the National Institute for Health and Clinical Excellence (NICE) on treating depression in children and young people also referred to the need for parents’ own depression to be treated in parallel. 43

Causes of depression

The factors that can contribute to depression include an individual’s personal experiences, biological or inherited tendencies, social support factors, and economic and environmental factors. 11 For example, people with mental health problems are more likely than the general population to live in rented housing and to say they are dissatisfied with their accommodation. 44 Because of the growing concern over mental health, policy initiatives have been developed, recognising that health-care interventions alone are not the only solution.

There is no consensus about the cause of PND, but there is an association with risk in women who have a number of psychosocial risk factors. A meta-analysis of 59 studies27 used regression analyses to evaluate the relative contributions of several postnatal variables to the development of PND (Table 1). The strongest predictors are related to antenatal anxiety or depression, lack of social support and stressful life events. Weaker predictors are neuroticism, negative cognitive attributional style and obstetric variables. The suggestion that women having a traumatic delivery, by emergency Caesarean section, might be more likely to become depressed45 may be true only for women who have a previous history of a depressive disorder. 46 For the general population of women, complications such as forceps or emergency Caesarean section are not associated with depression. 46 A link between Caesarean section and PND was not established in a meta-analysis of suitable studies. 47

Recognising that it is a simplification, O’Hara and Swain synthesised all of the risk factors that emerged from the meta-analysis27 to present a prototype of a pregnant women at risk of PND, as most likely to:

• occupy a lower social stratum

• have experienced stressors during pregnancy

• have had a more difficult than normal pregnancy or delivery

• be experiencing marital difficulties

• experience her partner as providing little social support

• perceive others in her social network as not supportive

• have a history of psychopathology

• show evidence of being at least mildly depressed, anxious and worried.

A prospective study48 of women recruited antenatally found that those who were depressed postnatally felt that the practical and emotional support provided by their partners was inadequate. The depressed women felt that they could not talk freely with their partners, they were not there for them when they needed them and they were not able to rely on them for childcare help as much as they would have liked. In general, they felt that their partners made their lives less easy. 48

Assessment and detection

There is a general problem with the detection of depression in primary care. 49 For many women with PND their problem will not have been fully recognised in routine clinical practice. 24,26,50 Because the onset of PND may be gradual, it is not easy to distinguish it from the fatigue and emotional liability that most mothers feel when adjusting to the demands of a new baby and recovering from childbirth. 50 It is also not easy to detect depression, partly because some women are not willing to disclose their true feelings. 51 Some women feel that they become depressed as a result of feeling exhausted, unwell, unsupported or isolated as mothers, with no time for themselves. 52 These women may not try to access professional or other support, either because they feel that their problem is not so bad or they ought to deal with it on their own or because they do not have anyone to ask for support. 52 Some women may feel that there is a stigma attached to being depressed and they may feel embarrassed or ashamed to seek help for what they might regard as a sign of personal inadequacy or an admission of failure on their part. 53 They may not regard professional intervention as relevant or may not want to be labelled as an unfit mother. 53

The EPDS is one of the mood assessment instruments most widely used in clinical practice. 54 It was not developed as a diagnostic test. 55 The EPDS is not adequate to confirm PND without a clinical interview to assess a mother’s mood, depressive symptoms and suicidal thoughts, and explore her relationship with the baby. 56

The National Screening Committee commissioned a review57 to evaluate the evidence of the validity of the EPDS as a screening tool; the most effective intervention for PND; and the size of the beneficial or adverse effects for interventions. The report stated that:

At present, it is not recommended to the National Screening Committee that screening for postnatal depression be introduced . . . the introduction of isolated screening programmes which are not part of a research project will not add to the evidence base which is agreed to be insufficient to justify the introduction of screening.

Until more research is conducted into its potential for routine use in screening for PND the NSC recommends that the EPDS should not be used as a screening tool.

It may, however, serve as a checklist as part of a mood assessment for postnatal mothers, when it should be used alongside professional judgement and a clinical interview. The professional administering it should have training in its appropriate use and should not use it as a pass/fail screening tool.

The difficulties of the EPDS have been openly discussed;58–61 many HVs do find the instrument valuable whereas others highlight the limitations. Some PCTs endorsed the systematic use of the EPDS by HVs and established a system of cascade HV training in its use. 62 Other PCTs were mindful of the criticism of the EPDS following the review commissioned by the National Screening Committee. This prompted some PCTs to restrict the use of the EPDS by HVs.

The guidance from the National Collaborating Centre for Primary Care18 on the postnatal care of women and their babies proposes that it is good practice to ask women who have had a baby how they are feeling emotionally, but cautions that the use of the EPDS is not acceptable to some women.

Pharmacological treatment for depression in primary care

There is not enough evidence from well-controlled and reported trials about the costs and benefits of different interventions for depression. Within the UK, depression in primary care is usually treated with either tricyclic antidepressants (TCAs) or, more recently, the newer non-tricyclic drugs or selective serotonin reuptake inhibitors (SSRIs). 29 The SSRIs are believed to be as effective as TCAs but less toxic in overdose. 29 Many people are uncertain about taking drugs for depression, believing that they are addictive or that drug treatment is not appropriate for what is seen as a reasonable reaction to adverse events. 29 There are certainly concerns about the risk of adverse neonatal outcomes following exposure to antidepressants during pregnancy. 63

Antidepressants are effective for postpartum depression. 64 However, it is not known which specific class of antidepressants or which individual antidepressant is most helpful; which is the best prevention for high-risk women; or what impact antenatal treatment or excretion of antidepressants via breast milk might have on the cognitive and emotional development of exposed infants. 64 Because there is insufficient information on the overall effectiveness of antidepressant drugs in PND, there are very little data upon which to base decisions about the safety of breastfeeding while taking these medications. 65 Women with PND prefer not to take antidepressants66 and so compliance is not good. Physicians either prescribe a reduced, potentially non-therapeutic dose, advise women not to breastfeed or delay offering treatment until the woman has finished breastfeeding. 67

An American expert panel68 reached a majority consensus on the appropriateness of including antidepressants (specifically SSRIs) and non-pharmacological treatments for women with severe depressive symptoms. For milder symptoms the panel gave equal endorsement to other treatment modalities or preferred psychotherapy over antidepressant medication. 68

Psychological interventions for mental health problems in primary care

Partly in response to concerns about antidepressants,29 over the past 25 years there has been a move towards increasing the availability of psychological interventions. 69

Psychological interventions include counselling and psychotherapy, and it can be difficult to distinguish between the two. Both cover a range of modalities, the most common being cognitive behavioural therapy (CBT) and psychoanalytic, psychodynamic, interpersonal and client-centred, non-directive approaches.

The British Association of Counselling (BAC)70 presents an ethical framework for good practice in counselling and psychotherapy, including values, principles and personal moral qualities. The BAC refers to counselling as:71

. . . the skilled and principled use of relationships which develop self-knowledge, emotional acceptance and growth, and personal resources.

. . . concerned with addressing and resolving specific problems, making decisions, coping with crises, working through feelings and inner conflict, or improving relationships with others.

Some patients with depression actively choose counselling over antidepressants. 72 The availability of counselling will depend upon the number of effectively trained practitioners.

In primary care a range of professionals can offer psychological interventions, including counsellors, community psychiatric nurses (CPNs), clinical psychologists, HVs and social workers. 69,73

In primary care, generic brief counselling74 or psychological interventions using non-directive psychotherapy are as effective as routine GP care, or perhaps more effective. 75–77 Generic counselling is as effective as antidepressants, although those taking antidepressants may recover more quickly. 72 In the short term psychological symptoms of patients who receive counselling may improve more than symptoms in those who have usual GP care. 69 Primary care patients may prefer brief psychotherapy to usual GP care75–77 and, given the choice, patients who choose counselling over antidepressants may improve more than those who have no strong preference. 72 Interventions using CBT also appear to be cost-effective in primary care78 and possibly helpful in preventing relapse. 30

Health visitors’ detection and treatment of postnatal depression

There is some evidence of the effectiveness of HVs in using psychological approaches to support women with PND. In a pioneering small randomised trial in Edinburgh and Livingston,1 HVs were asked to administer the EPDS to all women at around 6 weeks postnatally. Those who had a raised EPDS score were interviewed by psychiatric interview at 13 weeks postnatally, and those identified as depressed were allocated to an intervention group (IG) (n = 26) or a control group (CG) (n = 24). The IG were offered a postnatal one-to-one non-directive type counselling intervention of eight 1-hour weekly sessions by 17 HVs, whilst the CG received routine care. Outcomes included the Goldberg Standardised Psychiatric Interview and EPDS after 13 weeks. Although the HVs providing the intervention continued to visit the CG women, the statistically significant result was that 69% of the IG women (n = 18) recovered compared with 38% of the CG women (n = 9). In the absence of stronger evidence, the findings of this trial have been widely implemented throughout the UK.

The Lewisham primary prevention programme was one of the more important, small studies, which was not a randomised controlled trial. 6,79 The study compared outcomes for women screened antenatally as ‘vulnerable’ (using the Leverton questionnaire) in ‘Preparing for Parenthood’ (for first-time mothers) or ‘Surviving Parenthood’ (for second-time mothers) against routine primary care. The allocation to group was not random but by the baby’s date of birth and it was flawed because of the lack of concealment. HVs were asked to make contact with the women as soon as possible, in mid-pregnancy. There were five antenatal group sessions, beginning at 24 weeks, and six postnatal sessions, led by a clinical psychologist and a HV. At 3 months the women were interviewed, in part using the Present State Examination (PSE). Among the more vulnerable women, for those who had been offered the service, 19% (n = 48) were depressed compared with 40% of those who had not been offered the service. There was a significant reduction in EPDS for first-time mothers (n = 21) at 3 months compared with control subjects (n = 24), but no difference at 3 months for second-time mothers and no difference at 1 year for invited women. The authors concluded that some depression following childbirth can be prevented by brief psychological interventions, which can be incorporated within existing systems of antenatal classes and postnatal support groups, and pointed out that first-time mothers may be more likely to accept an invitation and attend meetings. 6

Following the Edinburgh trial1 Holden and Elliott wanted to give HVs the chance to take part in a training programme to adopt strategies for detecting PND and for early interventions. 4 To test whether the Edinburgh intervention, which appeared successful within a small trial, could be effective in routine HV practice they set up a three-centre study in Edinburgh, North Staffordshire and Lewisham, south-east London.

Health visitors were invited to a minimum of seven 2-hour training sessions and were asked to administer the EPDS to women, normally at the child health clinic, with a home visit for non-attenders. The preventive strategies included antenatal visits and education about PND, the realities of parenting and the potential benefit of support groups. 4 The study used an EPDS cut-off score of 12 so that each HV would counsel about three women on their caseload over the study period, using non-directive counselling (NDC). The HVs did not wish to be regarded as counsellors and preferred the term ‘listening visits’ to the term ‘non-directive counselling’. 4 In the North Staffordshire arm, the median EPDS score changed from 7 at baseline to 5 post training. 4 There were reported improvements in counselling skills and an increase in HVs’ mental health assessments, recording of symptoms and referrals to mental health services. Elliott et al. 7 suggested that the training and intervention should be evaluated using a rigorous research design. The study was not a randomised trial and the limited reporting suggests that it was probably subject to selection bias. It showed the potential role for HVs in using a structured approach for delivering an intervention following a brief training in psychological counselling for PND.

Another study with postnatal women, which was not a randomised controlled trial, explored the effect of care from HVs who were trained to detect PND using the EPDS and to manage PND using counselling and cognitive behavioural techniques, such as problem-solving. In total, there were 30 women who received routine primary care before the training and who became historical control subjects and 70 women who were seen after the HV training. 5 The study, which was not rigorously controlled, or reported, found a significant reduction in EPDS scores after the training. 5

Support for the role of health visitors in perinatal mental health

Health visitors have been working in multidisciplinary teams for some time in the area of prevention and the early identification of maternal depression and support for affected women. 80–88 A series of proposals and guidance has offered backing for the role of HVs in perinatal mental health. 79

NICE asked the National Collaborating Centre for Mental Health (NCCMH) to develop a clinical guideline on the treatment and management of mental health problems in the antenatal and postnatal period. 89 Before this, the National Service Framework (NSF) for Mental Health90 set priorities for the way that services were to be provided, four of which were relevant to the role of HVs. Standard one related to mental health promotion and emphasised the need to build capacity and capability in primary care by supporting staff through continuing professional development. Standards two and three referred to primary care and access to services. The NSF proposed protocols to be implemented for the management of PND, anxiety disorders and those needing referral to psychological therapies. The NSF recognised the role of HVs with training who could use routine contact with new mothers to identify PND and treat its milder forms. The NSF seventh standard90 related to actions to reduce suicides, by ensuring that staff would be competent to assess the risk of suicide among individuals at greatest risk. This standard was relevant to HVs, as maternal suicide was cited as the largest cause of maternal death in the first postnatal year. 21 The later review of the NSF prioritised investing more in mental health promotion. 91

The Scottish Intercollegiate Guidelines Network (SIGN) evidence-based guideline on PND and puerperal psychosis emphasised the role for HVs in the detection and management of PND. 92

The Department of Health published guidance in September 2003, Into the Mainstream, Implementation Plan: Mainstreaming Gender and Women’s Mental Health, for developing services for perinatal depression, which supported the role of HVs. 93

In the UK, the Sainsbury Centre for Mental Health, the Local Government Association, the NHS Confederation and the Association of Directors of Social Services produced a joint policy paper. 94 The report presented a vision for 2015, which minimised public fear, stigma and discrimination for people with mental health problems, shifted resources to primary care, invested in the mental health workforce and extended the availability of psychological therapies to people with a range of mental health problems.

Given the absence of a national policy on PND, HVs in many PCTs developed their own local policies,95 with differing strategies and integrated care pathways (ICPs) for the detection and management of the depression. 83 Some PCTs developed protocols for GPs for the management of PND, with information on treatment options and criteria for referral to the community mental health team. It is appropriate for HVs to refer some women to mental health services rather than offer support themselves. These circumstances include women who have obsessive compulsive disorder, eating disorders, post-traumatic stress disorder and panic disorder, as well those who have psychosis or suicidal plans, and other situations in which a HV feels very concerned. 7 This approach has been supported by Department of Health policy. 90,93

Health visitors’ professional support

Support for the role of HVs in perinatal mental health came from the HVs’ professional body. In 2000 the Community Practitioners’ and Health Visitors’ Association (CPHVA) established a Postnatal Depression and Maternal Mental Health (PDMMH) network for HVs to enhance perinatal services for women and their families. The PDMMH network facilitated the exchange of information on the development of ICPs, conferences, resources, publications and multicultural work. The CPHVA ran workshops about the use of the EPDS as part of a full mood assessment and advertised courses in identifying and managing perinatal depression.

An audit of the CPHVA membership was published in the network newsletter in June 2003. 96 The results suggested that 85% of PCTs had formal mechanisms for managing PND; 55% had a lead professional for perinatal mental health (72% of these being HVs); and 85% of PCTs were using the EPDS to some degree, but only 70% of these (sic) had received training in its proper use.

The generic role of health visitors

Health visiting relies on a sound interpersonal process and establishing a relationship with a client. The use of interpersonal skills and communication skills lie at the core of health visiting. 97 Whether regarded from a medical or psychosocial perspective, PND is acknowledged as an important health problem, and a key area of HVs’ work given their established role and unique personal contact with postnatal women. The following explains the historical, generic role of HVs and presents the context and rationale for their role in postnatal maternal health.

Health visiting has its origins in Salford, Manchester, where the Ladies’ Sanitary Reform Association first began home visiting to offer a universal service, with some focus on maternal and child welfare. 98,99 Since 1962 HVs have been qualified nurses, with special experience in child health, health promotion and health education, employed as part of the NHS community health service. They work with GPs and other primary health-care team workers (practice nurse, district nurse, midwife) and other community-based health and social care professionals, based within the GP surgery or practice premises or local health centre.

The Council for the Education and Training of Health Visitors100 identified the four main principles of health visiting as the philosophy underpinning practice:

1. the search for health needs

2. stimulation of the awareness of health needs

3. influencing policies that affect health

4. facilitating health-enhancing activities.

As policies within the NHS and in child health surveillance services have changed over time,101 so has the role of HVs. 99 In the 1990s HVs were encouraged to change the way that they worked, to offer a more targeted, needs-based service, rather than a universal service. The work of contemporary HVs is mainly around primary preventive activities on a broad range of health issues. However, recently there has been a strong drive for HVs to focus increasingly at the level of secondary prevention, targeting vulnerable children102 and using more community-based public health approaches. 103

The review of the British literature on health visiting104 indicated that HVs’ work can fall into the following categories:

• individuals and groups with special needs

• children with special needs

• elderly

• homeless families

• mothers with PND

• prevention of sudden infant death syndrome

• traveller families, vulnerable families and families in poverty

• child protection, domestic violence, childhood injury

• child health services, child health surveillance.

Health visitors are concerned with all aspects of a woman’s health and the health and welfare of her child and family. HVs maintain a ‘caseload’ of individual clients and part of a HV’s role is to visit families with new babies, in their homes, as part of routine child health surveillance. Therefore, every family with a child under 5 years has a named HV who can advise parents on everyday infant and childcare difficulties and immunisation programmes, as well as signposting families to other sources of health support, for example housing, financial benefits or specialist services. Some HVs also work in corporate teams with HVs sharing the caseload and so families have access to different HVs.

The standard HV contact times for women with a baby are around 4 weeks antenatally, at a new birth visit and in well-baby clinics. Routine contacts for assessment of infants’ developmental progress are being phased out.

The effectiveness of health visiting

There has been wide discussion over the evidence of the effectiveness of the work of HVs. 97,104,105 One of the first systematic overviews of home visiting106 indicated that there were positive outcomes in children’s mental development, mental health and physical growth; reductions in mother’s anxiety, depression and tobacco use; and improvements in maternal employment and nutrition, among others. There are very few reports of UK-based research in health visiting. 107 The review of articles on the effectiveness of home visiting in relation to child and maternal outcomes107 found evidence to suggest that home visiting programmes for parents of young children can have an effect on improvements in:

• various dimensions of parenting108

• some child behaviour problems

• cognitive development, especially for some groups of children

• childhood accidental injury rates109

• the detection and management of PND. 1

There was no evidence that home visiting increased the uptake of immunisations or hospital admissions. 107 As is often the case, the review indicated a need to address methodological limitations of trials in this area to provide, in particular,104 a clear theoretical framework; clear descriptions of the intervention content, intensity, timing and duration; process measures; long follow-up times; a client perspective and assessment of satisfaction.

The role of the health visitor in black and ethnic minority communities

There are specific mental health issues affecting Asian and other non-indigenous women bringing up children in the UK. For example, the suicide rate among women who are born in South Asia and live in England is higher than that in the general population. When they are providing supportive care, HVs need to consider cultural practices about childrearing.

English language is not an issue for some second-generation immigrant women who speak Punjabi and Urdu, and some HVs have used the EPDS with English-speaking women from Asian and other ethnic minority groups. Some women who have recently entered the UK from Pakistan are unable to speak English, and there are growing numbers of Arabic-speaking women as well as Kosovans, Kurdish people and asylum seekers from other mid-European countries and elsewhere. For Pakistani women, link workers are employed who speak Punjabi and can read Urdu. There are also interpreters and link workers who speak other relevant languages. Aside from any language difficulty, literacy for women from Pakistan is more likely to be an issue.

There are no effective, validated, culturally sensitive tools for many women who have English as a second language. As well as literacy and language difficulties, immigrant women may also be isolated, and so women who are vulnerable to PND may be missed. Some work is beginning to develop and validate linguistically and culturally competent tools for use in primary care, using link workers and health professionals to identify psychological distress and assist the early detection of women from South Asian communities who speak Bengali, Gujarati, Punjabi and Urdu, as well as those from other ethnic minority communities. This may be useful in instances in which it is not possible to detect PND in other ways.

A Punjabi Postnatal Depression Screening Questionnaire (PPDSQ) was developed by a consultant psychologist in Bradford City PCT and the University of Bradford. 110 Also, the CPHVA has supported the development of a pictorial method for women who have English as a second language, to detect those who may be depressed. This work is undergoing a pilot validation study.

Antidepressants to treat postnatal depression

The Cochrane review of antidepressant drug treatment for PND aimed to compare the effectiveness and safety of different antidepressants with other forms of treatment. 67 The Cochrane review included only one trial of fluoxetine,3 which was rated for methodological quality as category A. This was a community-based, randomised, double-blind controlled trial of 87 depressed postnatal women in Manchester that had four treatment groups:

1. fluoxetine 20 mg with one session of cognitive behavioural counselling (CBC) by a psychologist

2. fluoxetine with six sessions of CBC

3. placebo with one session of CBC

4. placebo with six sessions of CBC.

There was improvement in all groups and fluoxetine was more effective than placebo and six sessions of CBC were more effective than one session of CBC. Fluoxetine and CBC were equally effective for non-psychotic depression in postnatal women. However, 101/188 (54%) eligible women refused to take part, mainly due to reluctance to take the drug, and there was a 30% dropout rate with 61/87 women who agreed to participate, completing.

The review concluded that women with PND can be treated equally effectively with fluoxetine or a course of CBC in the short term and that there should be more, longer-term studies comparing different antidepressants and psychosocial interventions. 67

Psychosocial interventions

The withdrawn Cochrane review to assess the effect of caregiver support for postpartum depression included the Manchester trial3 of fluoxetine and cognitive behavioural-type counselling and the Edinburgh trial, which was a study of a HV one-to-one NDC intervention. 1 The review indicated that it would be premature to make practice recommendations based upon only two small trials. It indicated that future research should consider lay support; home visits, phone calls or group sessions; and the prevention and treatment of PND, including outcomes of symptoms, hospital admission rates and long-term maternal and infant and family well-being. Additionally, an economic evaluation would be necessary to determine the relative efficiency of the provision of care.

The Edinburgh trial tested one-to-one NDC visits,1 whilst the Manchester trial tested CBT3 and the Cambridge treatment trial2 tested CBT against person-centred therapy and psychodynamic psychotherapy. The Cambridge trial was the largest trial of psychological interventions for women with depression postnatally. Primiparous women were screened ‘in the early postpartum period’ to identify those who met DSM-III-R criteria for current major depressive disorder. Women were offered therapy in their homes from 8 to 18 weeks postpartum or routine care in four groups, as in Table 6. The EPDS and Structured Clinical Interview for DSM-IV Disorders (SCID) were administered at 18 weeks. In the short term there was a 25–35% reduction in EPDS in the IG compared with about 4% in the CG.

The trial found that there was a significant improvement following NDC, CBT or psychodynamic therapy immediately after treatment compared with routine care. 2 By 9 months there had been recovery in the CG so that there was little difference between the reduction in their EPDS scores and the reduction in the scores for the women who had received counselling or psychotherapy. 2 The improvements were well maintained in the CBT group up to 18 months and, using the percentages of women who dropped out of treatment early, the CBT appeared to be the most acceptable treatment.

The trial also found a significant benefit in mothers’ reports of relationship problems with their infants at 4 months postnatally. 163 Mothers’ reports also indicated evidence of a benefit from NDC at 18 months for emotional and behavioural problems.

On the basis that it works for major depression, the efficacy of interpersonal psychotherapy (IPT) was tested in a randomised controlled trial149 of 120 women in Iowa, USA, who met the DSM-IV criteria for major depression using a modified SCID. Most of the women were white and well educated. The IG women (n = 48) were offered 12 × 1-hour sessions of IPT, which were carried out by 10 experienced psychotherapists who were required to complete a 12-session course of IPT with a postpartum depressed woman ‘at a satisfactory level of competence’. The CG was a waiting list control (n = 51) who were phoned every 2 weeks. In total, 20% of women withdrew from IPT. After 12 weeks all scores except the Dyadic Adjustment Scale (DAS) were significantly better in the IG than the CG, as assessed by non-blinded raters.

The Australian study comparing CBT, group counselling and one-to-one counselling against routine primary care used an allocation method of drawing slips of paper from a bag and was therefore not free from bias. This study found improvements in all three groups compared with control, and the greatest benefit appeared to be associated with individual counselling. 157

In contrast, the very small Australian study of nurse-delivered CBC for women with depression found no significant difference immediately post treatment, but there may have been some effect at the 6-month follow-up. 161

A very small study156 in Cardiff, Wales, examined the effectiveness of a brief ‘psychoeducational group’ intervention for women who scored 12 or more on the EPDS. In total, 23 women were allocated to the intervention of eight 2-hour group meetings run by HVs. These covered education about childcare and accessing social support, cognitive behavioural techniques and relaxation techniques. Compared with the women in the CG who received routine primary care, the mean EPDS scores decreased significantly in the IG.

The very small controlled study162 in Goteborg, Sweden, included 20 IG women and 21 CG women who scored 12 or more on the EPDS at 2 months and again at 3 months interviewed with the Montgomery-Asberg Depression Rating Scale (MADRS) and diagnosed as having major depression. The IG women were offered 6-weekly 1-hour counselling sessions in the home or clinic by the child health nurse, who had received four half-day training sessions in NDC, whereas the CG received routine care. In total, 80% of women (12/15) in the IG showed no major depression after six sessions compared with 25% (4/16) of CG women (p < 0.01). Seriously ill women were excluded and the randomisation process was not described.

Informational support

The small Taiwanese study155 of 70 women, offering information about PND, appeared to show some positive impact but it was not well reported.

Postnatal support groups

In the small Taiwanese study151 30 women who had a Beck Depression Inventory (BDI) score of 9 or more at 3 weeks attended a support group of five to six mothers with their infants. A total of 30 CG women completed two assessments. In the IG, 10 women (33.3%) remained depressed, compared with 18 women (60%) in the CG, as measured by a BDI score of less than 10, indicating a possible benefit from the intervention.

Psychoeducational visits with partners

The very small study158 of psychoeducational clinic visits for depressed women, with partner support in four of them, found lower EPDS scores in the women whose partners attended the support group, as well as less psychological morbidity in the partners themselves, compared with those who did not attend.

Pram-walking exercise programme

In a very small Australia study women who had an EPDS score of 12 or more at baseline were allocated at random to either an exercise group (n = 9), who were encouraged to attend two pram-walking sessions per week, or a social support group (n = 10), who met once per week. 150 The women were a well-educated group (who had a pram) and the analysis was not by intention to treat. The mean EPDS scores in the pram group were significantly lower than those in the social support group after 6 weeks, but there were no significant changes in social support in either group.

Telephone support

The pilot study153 of telephone-based peer support with 42 women found significant reduced EPDS scores in the IG at 4 months postnatally compared with the usual care group.

Massage

The two very small massage studies154,160 were both poorly reported but appeared to demonstrate some difference between the IG and CG.

Primary aim

The primary aim of the trial was to reliably estimate any differences in outcomes for postnatal women attributed to special training for HVs, delivered at GP practice (cluster) level, in systematically identifying depressive symptoms and delivering experimental psychological sessions, based on either cognitive behavioural principles166 or person-centred principles,167 compared with HV usual care (control).

Secondary aim

The secondary aim was to establish the relative cost-effectiveness of the HV training from an NHS perspective, relative to HV usual care.

Overview of design

The study was a prospective pragmatic randomised cluster trial with clusters allocated at random to one of two experimental psychological approaches or HV usual care (control arm), stratified by number of expected births per year. It was a pragmatic trial of the effectiveness of an intervention offered under normal conditions, excluding as few women as possible and accommodating non-acceptance of the offered intervention and other co-interventions.

Pragmatic trials aim to establish the relative value of interventions, as they would be provided in routine care settings, to increase the external validity without adversely affecting the internal validity. Hence, the trial aimed to answer a real-life clinical question in a real-life clinical situation. 169 This means that the interventions reflected the clinical variation that exists in routine primary care contexts. 69 In the CG, routine care included all of the support that women would normally access, from HVs, GPs and elsewhere.

The clusters randomised in the trial were GP practices, and the HVs who worked with GPs, and held a caseload of families registered with the GP practice, were approached to take part. The intervention arm HVs were trained in delivering the psychological sessions to which their practice had been randomised.

Collaborating HVs approached pregnant eligible women aged over 17 years who were on their caseload to take part in the study. Women who consented and who had a live baby were sent a 6-week postal questionnaire. All women with a 6-week postal EPDS score ≥ 12 were regarded as at-risk women and were included in the main trial of the two psychological approaches, cognitive behavioural approach (CBA) and person-centred approach (PCA), compared with HV usual care. These two contrasting approaches were explored because earlier studies had found promising evidence of a positive effect from offering sessions with a person-centred (non-directive) approach1,2 and also from offering sessions that incorporated a cognitive behavioural component. 3,152,156,157,159

The IG at-risk women were invited for an interview using the SCAN. 168 Those who were found to be moderately or severely depressed were asked to state their preference for psychological sessions, an SSRI, or both. The EPDS does not provide sufficiently fine-tuned information to differentiate different levels of severity of clinical depression and does not assess depressive psychosis or mania, all of which are capable of being assessed in a standardised way using SCAN.

The IG at-risk women, all of whom had a postally administered EPDS score ≥ 12, were reassessed at 8 weeks postnatally by a face-to-face HV administration of the EPDS. Women were eligible for the psychological sessions to which their practice (cluster) had been randomised according to the HVs’ management protocol if they had an 8-week EPDS score ≥ 12. The intervention therefore comprised the package of HV training to develop skills in the assessment of postnatal women and the provision of psychological sessions, plus the option of an SSRI if a woman’s SCAN outcome indicated moderate or severe depression. In addition, there was a change to the original protocol so that HVs were able to provide the intervention to women irrespective of their EPDS score if the clinical assessment by the HV indicated that they might benefit from the intervention sessions.

All women were followed up at 6, 12 and 18 months postnatally, using postal questionnaires. The primary outcome was the proportion of at-risk women with a 6-month EPDS score ≥ 12. The trial is illustrated diagrammatically in Figure 1, and full details of the methods are explained below.

FIGURE 1.

Diagrammatic representation of overview of trial. CBA-F, cognitive behavioural approach face-to-face group; CBA-P, cognitive behavioural approach postal group; CG, control group; EPDS, Edinburgh Postnatal Depression Scale; PCA-F, person-centred approach face-to-face group; PCA-P, person-centred approach postal group; SCAN, Schedule for Clinical Assessment in Neuropsychiatry.

Estimates of intracluster correlation coefficient

We used the ICC (ρ) estimate of 0.006 derived from 6-month EPDS scores by GP practice,146 which indicated little clustering by practice. 170 Among the interested practices, the annual rate of births ranged from 50 to 150 per practice, with an average of 78 births. To estimate the number of clusters required and the numbers of women to be recruited per group to have 90% power at the 5% two-sided level of significance, to detect a 15% absolute difference in the proportions of intervention and control women with a 6-month EPDS score ≥ 12 (35% versus 50%) we made the following assumptions. If 12–14% of all women would be eligible for the intervention,27 the average cluster size would be 6–8, 50% of at-risk women in the CG would have a 6-month EPDS score ≥ 12, and 50% of women would consent (Table 7). Within the IG there would consequently be 80% power to detect a 15% difference in the proportions of women with a 6-month EPDS score ≥ 12 (i.e. 42.5% versus 27.5%), between the two approaches (CBA versus PCA), as statistically significant at the 5% two-sided level.

We assumed an average cluster size of six women, which would require a total of 519 women recruited from 87 practices over 1 year. Assuming a 20% loss to follow-up at 6 months,33 we required 649 women in total and a recruitment phase of 15 months. The sample size calculation was based on a two-sided statistical test and assumed an allocation ratio of one PCA group to one CBA group to one CG. This was so that, first, the outcomes for all of the intervention clusters could be compared with the outcomes for all of the CG clusters and, second, that the outcomes for the two IGs could be compared.

Cluster not individual allocation

In general it is preferable to randomise at the individual participant level. However, an individual woman could not be the unit of randomisation because it was not possible to ask the HVs to provide the control usual care to one woman and then to provide an intervention to another without some contamination of the CG.

Thus, although cluster randomisation is less statistically efficient than individual randomisation,171 a cluster allocation was chosen to avoid major sources of bias and minimise contamination between groups, particularly as blinding was not possible. 170 The GP practice (cluster) was the unit of randomisation and intervention, because that is where the intervention was delivered, even though the effect was to be evaluated by measuring outcomes in individual women and analysing at the individual level, adjusting for clustering at the GP practice level. 165 To avoid any contamination between clusters, HVs for women in the control arm were not to be trained in the skills needed for the experimental intervention, which was not to be offered at any time in the control practices.

Cluster random allocation

To minimise any imbalance across the IGs, the details of practices were used to separate the clusters into three strata for random allocation. The three strata were based on the expected number of births per year, according to whether the HV caseload in the practice was small (less than 70), medium (70–100) or large (more than 100). Once stratified, the clusters were coded for allocation.

A computer-generated random list, using a program from the University of Southampton Medical Statistics and Computing Department, was prepared by Professor Mike Campbell, who was blind to the identity of the collaborating HVs and the GPs with whom they worked. This sequence was used to allocate clusters, stratified by the expected number of births per year, to one of the two main experimental groups, the CBA or the PCA, or to the CG. The two main experimental groups had an equal probability of face-to-face plus postal or postal-only EPDS administration, in a ratio of 1 CBA-F:1 CBA-P:1 PCA-F:1 PCA-P:2 control clusters. The random allocation took place in February and March 2003. To minimise selection bias the HVs were asked to consent to take part before allocation and were therefore blind to the allocation at the time of their consent. After the allocation some HVs expressed their strong disappointment about being in the CG but nevertheless continued to collaborate in the trial.

Inclusion and exclusion criteria

Cluster level recruitment

To facilitate recruitment trial information was disseminated before securing funding, aiming to capitalise on existing networks of GPs, HVs and primary care research and development leads interested in research in the region. There was notable support of Trent Focus personnel in the former Trent region, the acting research and development manager in Broxtowe and Hucknall PCT and the Barnsley research fellow in primary care.

Cluster level consent

Following approval from the Trent Multicentre Research Ethics Committee (MREC) in February 2003, at least one GP in each practice was asked to sign the consent form to indicate that they understood the HVs’ research role and that the GP practice would support their collaboration. The HVs and GPs then signed a consent form stating their duties before they volunteered the women in their practice as a cluster. 172,173

Once they had consented, the HVs were given details of the group allocation and, when relevant, the training cohorts, and a trial file with the HV protocol to guide the recruitment of women, administration of the EPDS and provision of information to the research office, according to allocation to group.

As not all of the clusters consented at the same time the randomisation process was repeated when there were sufficient clusters with enough HVs to comprise a training cohort for each psychological approach, plus the equivalent number of control HVs, to allow random allocation at one time. Because of this there were five successive training cohorts, to accommodate the training needs of the HVs who were recruited at different times.

Individual level recruitment

The protocol stated that IG women found to be at risk of PND would be offered an intervention as part of the trial. To avoid selection bias HVs were asked to invite all eligible women antenatally to consent to take part in the trial, before the development of any depression.

The HVs were asked to log the details of pregnant women and, to avoid disclosing personal details to the research office at a preconsent stage, they were asked to post the women a research information leaflet (RIL) and a consent form at 32–36 weeks antenatally. Women gave or sent their signed forms to the HVs, who were asked to send the originals and fax (or photocopy and post) the updated consented women’s log each week to the research office for monitoring the consent rate and preparing the 6-week questionnaire for posting.

Individual level consent

Normally consent to take part in a trial is obtained before randomisation, to reduce the possibility of selection bias. People in a randomised controlled trial would normally be asked to consent to be part of an experiment, having been informed of their chance of being allocated to the experimental intervention. Although the clusters consented before randomisation it was impossible to obtain women’s consent before the clusters were randomised.

There are ethical issues surrounding consent in cluster trials. 172,174 A balance was needed to retain women’s autonomy; to provide women with enough information to avoid increasing their concern, losing the good will of disappointed CG women and, because of lack of incentives, jeopardising the recruitment rate; and to avoid non-random selection bias and differential consent affecting the validity (SJL Edwards, 2002, personal communication). 173

It is recommended that consent should be sought at all possible levels, and for the trial these were the sponsor, the MREC, the PCTs, health-care professionals, as ‘guardians’, and individual women acting independently of the guardian.

The trial followed the Medical Research Council (MRC) guidance on cluster recruitment. 175 All women received explicit information about the trial and the same consent process was used for all women. The RIL explained that women were being asked to take part in a study in which one-third of HVs would continue to provide their usual care in up to 30 practices, whilst the other HVs would provide one of the two kinds of support being researched for women found to be at risk of PND. CG women were asked only to return their postal questionnaires, and they would still receive routine care. 171,173 Monitoring the rate of consent in each cluster assessed the scope for potential bias. The HVs were asked to document the women’s main reasons for choosing not to participate.

Baseline measurements at cluster level

The cluster level characteristics of PCT, number of GP partners, number of full-time equivalent HVs, practice population, expected births per year and Index of Multiple Deprivation (IMD)176 were collected and compared to establish the comparability of cluster characteristics, as well as the representativeness among other practices within the region and nationally.

Baseline measurements at individual level

The probability of an imbalance on important prognostic variables may be greater for a cluster randomised controlled trial than an individual randomised controlled trial. 171 The measurement of baseline individual level variables also allows an assessment of how the randomisation process has worked to produce directly comparable groups. Furthermore, this comparison can be used to indicate where any baseline adjustments, if any, should be made, depending upon how great the differences might be, as well as the predictive relationship of a characteristic with the primary outcome. We therefore measured sociodemographic baseline variables. The HVs collected details of women who were ineligible or who declined to take part in each cluster – whether a woman’s baby was her first, a girl or a multiple birth. They also collected information on whether a woman had had PND previously, had English as a first language, lived alone or lived in rented accommodation and on the type of housing. This information was compared with the details of consenting women and was used for describing the generalisability of the results. The MRC guidelines indicated that completely anonymised personal data could be used for such a purpose. 175

Six-week measurements

Cluster level intervention

The cluster level intervention comprised the package of HV training in assessing women and identifying depressive symptoms (using the EPDS and clinical assessment) and the delivery of either CBA166 or PCA167 psychological sessions to at-risk women.

The manualised HV training addressed therapy allegiance and prepared the HVs to provide an appropriate, pragmatic, distinctive, derivative approach, delivering critical elements from CBT or person-centred therapy, not psychotherapy.

The common areas for both training approaches were to enable HVs to acquire further generic skills in developing helpful relationships, such as positive regard and empathy. The cognitive behavioural training emphasised a normalising rationale and the identification of unhelpful patterns of behaviours, perceptions or thoughts in the woman’s life, in order to help the woman to change these herself. 166 The person-centred training used the three principles of the actualising tendency, a non-directive attitude and the necessary and sufficient conditions of change. 167 Further details of the HV training are provided in the section on training of the intervention group HVs.

Individual level intervention

The proportion of women with a 6-week EPDS score ≥ 12 was very important, as this was the group of at-risk women whose outcomes were compared across all IGs and the CG at 6 months postnatally. It was not possible to use the group of women eligible for the intervention with an EPDS score ≥ 12 on two occasions in the comparison as the HV 8-week EPDS scores were not available for the CG women.

At-risk women were interviewed at 7 weeks postnatally using the SCAN, which incorporates algorithms using ICD-10 criteria for depression to determine depression severity (none, mild, moderate or severe). 168 Those with an ICD-10 SCAN classification of moderate or severe depression were asked to state their preference for psychological sessions, an SSRI, or both.

The HV readministered the EPDS to at-risk women face-to-face at 8 weeks postnatally. Women with an 8-week EPDS score ≥ 12 were offered the psychological sessions as cluster randomised. These consisted of a 1-hour CBA or PCA session, focusing on the needs of the mother, once per week for up to 8 weeks, commencing around 8 weeks postnatally. Because of the limitations of the EPDS some HVs also offered the sessions to women, irrespective of their 6-week or 8-week EPDS scores, if their assessment indicated that they were likely to benefit.

Primary outcome: at-risk women

The primary outcome was the difference between the proportion of at-risk women scoring ≥ 12 on the 6-week EPDS administered by postal questionnaire and the proportion scoring ≥ 12 at 6 months postnatally, to assess persisting risk of PND.

Four-month follow-up: at-risk women

A 4-month follow-up was proposed for IG women who received an intervention to compare outcomes immediately at the conclusion of the intervention sessions. However, the 4-month follow-up was not possible to achieve, mainly because it was impractical to send a further questionnaire, which may have imposed a further burden on already unwell women and may have served as a deterrent to returning the 6-month postal questionnaire (the primary outcome) 8 weeks later. For some women whose intervention did not begin exactly at 8 weeks postnatally, the outcome would have been assessed before the end of the intervention. It may also have led to a further degree of incomparability when only the women who had received an intervention were asked to complete a questionnaire at 4 months.

Secondary outcomes

A range of other outcomes was measured to capture the potential benefit from the intervention in different dimensions of health and use of services. The secondary outcomes were measured in at-risk women at 6, 12 and 18 months postnatally by postal questionnaires. Standard instruments were used as the main and secondary outcomes, with supplementary questions.

Infant outcomes

To examine infant development outcomes from the perspective of both parents, on the 18-month questionnaire women and their partners completed questions on their toddler’s growth and development and on concerns about toddler development, the modified Behaviour Screening Questionnaire (BSQ) and the Checklist for Autism in Toddlers (CHAT). The scores on the infant outcomes were rescaled onto a 1–100 scale. The HVs extracted infant immunisation data from the GP records.

Primary outcome

The primary outcome was the proportion of at-risk women with a 6-week EPDS score ≥ 12. The primary comparison was between those at-risk women in the combined clusters randomised to intervention and those women in practices randomised to provide HV usual care (control). The secondary comparison was to determine any differences between the proportions of women with a 6-week EPDS score ≥ 12 for the two main IGs.

A marginal generalised linear model, with coefficients estimated using generalised estimating equations,196 with robust standard errors and an exchangeable autocorrelation matrix in STATA v8197 was used to analyse the outcomes and allow for the clustered nature of the data. For binary outcomes, such as EPDS score < 12 or ≥ 12, a logit link with a binomial distribution for the outcome was used.

Secondary outcomes

For continuous outcomes, such as mean EPDS score, an identity link with a normal distribution for the outcome was used. Estimates for the group coefficients from these regression models were reported along with their associated 95% confidence intervals. In all of the analyses both a simple unadjusted model and a model to adjust the outcome comparisons for individual level covariates, for example lives alone, history of PND, life events and baseline (6-week) EPDS score, were fitted.

The exchangeable correlation structure corresponds to an equal correlation model, meaning that the correlations of outcomes with a cluster (GP practice) were constant.

For the other secondary maternal outcomes, that is, the CORE-OM, STAI, SF-12, PSI and DAS, the mean values were compared between the intervention and control groups at 6, 12 and 18 months using similar models.

The family outcomes collected from women’s partners at 6, 12 and 18 months were compared between the intervention and control groups. For the infants, outcomes at 18 months were compared between the intervention and control groups.

The primary analysis was for the at-risk women with an EPDS score ≥ 12 who had completed a 6-month EPDS score (n = 418). The analysis was also reported for the cohort of all women who consented and who completed a 6-week and a 6-month EPDS (n = 2659).

There is no general consensus on what procedure to adopt to allow for multiple comparisons198,199 Following this we have reported unadjusted p-values and confidence limits. However, because of multiple hypotheses testing, some caution should be applied in the interpretation of the p-values we have reported, particularly for the various secondary outcomes and end points.

Training Reference Group

The implementation of the psychotherapeutic HV training programmes and intervention was informed by the documented methodological prerequisites for comparative psychotherapy research. 200,201 The major requirement was to minimise any biasing effect of any of the researchers’ allegiances to either of the therapeutic approaches. To enhance the rigour and effectiveness of training for both psychotherapeutic approaches, to maximise the comparability of the programmes and to ensure that the trial would be considered by advocates of each method to have been a credible and fair test of that method, a Training Reference Group (TRG) was established before the trial, at the end of 2002. This comprised experienced academically based psychotherapy trainers from England and Scotland, including representatives of both the cognitive behavioural and person-centred approaches.

The TRG considered the potential for bias and distortion of results of comparative studies attributed to researchers’ loyalties to their preferred therapeutic method. 202 Two of the practical recommendations to try to minimise any potential impact of a researcher’s therapy allegiance unfairly influencing the effect size of the therapy compared were to include a mix of researchers who represent different therapy allegiances and to arrange for the people providing the therapy to be supervised by those representing the same intervention mode.

Training manuals

The two main psychotherapist trainers (TR and KT) were specialists with experience in practice as trainers and supervisors. They prepared a manual for each HV to keep throughout the trial, and a separate trainer’s manual. The manuals were to include the theoretical basis for the relevant psychological approach and the training plan so that, if necessary, the training could be replicated elsewhere. The manuals were drafted in January 2003 and final changes were made in February 2003.

Principles and standards for training for the intervention

The TRG held two verification meetings at the University of Sheffield in November 2002, chaired by Professor David Shapiro. The following summarises the main principles and standards for the HV training and manuals.

Recognising the training employed in previous trials in Edinburgh – ‘a brief training inthe principles of person-centred counselling’1 – and in Manchester, using CBT,3 the training in the planned trial was to prepare HVs to provide a brief, derivative intervention, not psychotherapy, mainly for pragmatic reasons. That is, the outcomes to be compared between the IG and CG would be associated with a brief training in delivering critical elements derived from one of the two therapies (cognitive behavioural therapy and person-centred therapy). The training had to be delivered at an appropriately pragmatic level to enthuse HVs and develop their skills, rather than to develop their theoretical knowledge, recognising the preference of HVs to support women with psychological difficulties rather than to become mental health workers. It was not intended that the HVs should regard themselves (or be regarded) as therapists, whose training takes much longer than 8 days. Therefore, the terms ‘person-centred approach’ and ‘cognitive behavioural approach’ were to be used consistently, to avoid the use of the term ‘therapy’, ‘counselling’ or ‘counsellors’.

Comparable training

The TRG was asked to verify that the training manuals for both intervention arms were comparable, with an appreciation of the differing ethos and styles of the two psychological approaches.

For the CBA the purpose of the training was to prepare HVs to provide a simple, easily communicated intervention related to the phenomenology of PND. The basis of the training was the worksheet approach, in which HVs learned to carry out a problem-focused assessment in five key areas of a woman’s experience and then select and use appropriate PND-specific worksheets for each woman.

For the PCA the purpose of the training was to prepare HVs using key principles and issues in such a way that they would be able to help the women to accept and ameliorate their depressive process.

The training preparations aimed to make the training experience equal, as far as possible, for all IG HVs. To enhance comparability it was agreed that all training cohorts should ideally be no larger than 12 (to allow for four small groups of three HVs working together during the training day) and no fewer than eight HVs. The training used the term ‘client’ or ‘woman’ rather than ‘patient’. Key qualities of the training environment were that they should be uninterrupted and secure and congruent with what would normally be expected of a training environment, with a pragmatic consideration of reproducibility and deliverability in the NHS.

Appropriate training

The TRG emphasised the avoidance of unfamiliar language and jargon, for example the term ‘negative automatic thoughts’, to avoid putting off HVs and to provide accessible, distinguishable, theoretically congruent and reproducible models with key skills. As well as being appropriate for the HVs, the intervention was planned to be appropriate for the women, with little time or energy to do too much homework.

Clinical supervision

The HVs also needed access to clinical supervision and support, for example when dealing with distressing information from a client, such as negative thoughts towards the baby. Regular formally structured reflective practice sessions using role-play were offered for HVs who may not have had the opportunity to work with affected women. HVs also attended peer-supervisory sessions.

Prior beliefs of health visitors

Because of the random allocation of clusters to the groups (control, CBA or PCA), the HVs would not be able to choose any preferred option. Therefore, there was the potential for incongruence between HVs’ personal predispositions or beliefs and one of the approaches. To be able to check the balance of the randomisation in terms of HVs’ prior beliefs, there was a prerandomisation measurement before random allocation of all of the HVs’ attitudes and levels of interest and motivation in counselling, using the Opinions on Psychological Problems (OPP)203 within a pre-trial questionnaire. The OPP is a self-report, two-part questionnaire developed to measure how people view the causes of and treatments for psychological problems. The first part measures how people view the causes of psychological problems. The second part, the treatment section, includes 47 questions about people’s views on what may help psychological problems. The questions are grouped under the following headings:

• psychodynamic

• humanistic/interpersonal

• behavioural

• cognitive

• organic

• social/economic

• naïve.

Each question has six responses and is scored on a 6-point scale from –3 to + 3, where + 3 means agree strongly and –3 means disagree strongly. The HVs were asked to indicate their level of agreement with each of 37 statements. The OPP in the HV questionnaire included only 37 questions as the naïve questions were not included. The second part of the OPP was included in a post-trial questionnaire for intervention and control group HVs, when the intervention was complete, to assess any post-trial differences between the three main groups.

Pre-trial health visitor questionnaire

To gather personal views and experience within the pre-trial questionnaire, to establish baseline practice, HVs were asked to answer truthfully, without discussing any of their answers. Codes were used so that all answers could be treated completely confidentially. The questionnaire included questions about:

1. Skills – details, the title, length, year and location of training, qualifications achieved for training on identifying or supporting women at risk of postnatal depression.

2. Assessing and identifying women – at baseline, how much confidence they had in identifying and supporting women who may be at risk of postnatal depression and how, over the previous 6 months, they had assessed and identified women who might have been at risk of PND.

3. Use of the EPDS – personal use of the EPDS over the previous 6 months and whether used universally with every postnatal woman; at how many weeks, when administered; and whether used selectively and how they had decided to use it.

4. General experience – the number of years since they qualified and how many years they had worked as a HV.

5. Experience with PND – how they had supported women whom they felt were suffering from PND in the previous 6 months.

Introductory training day

The IG HVs were invited to attend an introductory day during the week before contact with the psychotherapist trainers, to introduce the features and problems that women report and the different ways of understanding PND and covering risk issues (Table 9). These were run by Jane Morrell (principal investigator, trained as a HV, with group work and presentation skills) and Jan Cubison (Sheffield Community Health Maternal and Mental Health Services). Before the introductory day was finalised, in January 2003, JM attended the 1-day training in London on Perinatal Depression: Detection in Primary Care, which was organised by the CPHVA to train HVs in the use of the EPDS.

Each HV was provided with their own, named introductory training day manual, which they were asked not to share with anyone else, or copy, as one measure to reduce contamination when HVs may have come into contact with other HVs or their clients from another arm of the trial. They were also each given a copy of a scored EPDS, several copies of blank EPDS forms, laminated cards and sheets with questions to assist the clinical interview, and their own copy of the PowerPoint slides.

The day covered the use and meaning of the term ‘postnatal depression’, prevalence, factors associated with PND, consequences and a summary of the research about treatments for PND, including the pioneering work in Edinburgh on HV listening visits. 1 It also covered the limitations of research into PND and the need for the trial. The service perspective was presented, covering Why Mothers Die21 (confidential enquiry into maternal deaths) and the relevant standards within the NSF for mental health. 90

The clinical perspective covered core features of depression and core features of PND, distinguished from puerperal psychosis. The training on the clinical interview covered risk management and assessing risk to mothers, infants and children.

The practical skills development focused on how to recognise PND, describing the review commissioned by the National Screening Committee,57 the SIGN publication on PND. 92 The development and appropriate use of the EPDS were reviewed, covering its strength, limitations and recommended practice. The HVs were all asked to interview another HV, to gain experience in the practical use of the tool.

The HV protocol for the administration of the EPDS at 6 weeks postnatally within the trial was explained in detail, with a copy for each HV included within the introductory day manual. This indicated four main actions: first, when there was urgent concern about a woman (suicide risk, risk to the baby, risk to others and severe impairment); second, when there was a positive response to question 10 on the EPDS (indicating the risk of self-harm); third, the need to repeat the administration of the EPDS at 8 weeks postnatally if the score reached or exceeded the threshold of 12 when first administered at 6 weeks postnatally; and, finally, what to do if the EPDS score was below the threshold score of 12.

Five-day training

There were four CBA training cohorts and five PCA training cohorts (Table 10).

Evaluation of the health visitor training

Intervention monitoring

When participants do not accept an intervention, the estimated treatment effects are diluted in an intention to treat analysis. The HVs in the IG were asked to complete a monthly intervention monitoring form to provide an update on which women had been offered and accepted intervention sessions, how many sessions had been audiotaped, and whether Agnew Relationship Measures (ARMs) had been completed.

Intervention process evaluation

Reviews of psychotherapy studies indicate that there is little evidence for the differential efficacy of therapies, partly because of the quality of the research evidence. 204 This is referred to within psychotherapy research as the equivalence paradox. It is assumed that, although different models may have specific effects, for example cognitive behavioural therapy operates by identifying and changing negative cognitions, all therapy models contain common factors or non-specific effects, such as warmth, reassurance, instillation of hope and feelings of support. To help understand how a therapy works, several process evaluation methods can be used, for example to look at a client’s change in mood, such as the Session Evaluation Questionnaire or the Session Impact Rating Scale. 205

Preparation of local co-ordinators

To enhance their research role the LCs were able to attend training to develop their research skills and appreciation in critical appraisal, health economics, literature searching, qualitative research, research governance, statistics, use of NUD*Ist and NVivo qualitative software and use of the Statistical Package for Social Scientists (SPSS). The LCs also attended the Society for Reproductive and Infant Psychology conference in Sheffield in 2004. The LCs’ main training was in preparing them to undertake the SCAN interviews.

SCAN interview

The purpose of the SCAN168 interview was to establish the baseline severity of depression among those women with a 6-week EPDS score ≥ 12. A secondary aim of the SCAN interview was to examine the identification of depressive symptoms by the EPDS.

Identification of women for SCAN interview

The 6-week postal EPDS score determined which women were invited for a SCAN interview. All women with a 6-week EPDS score ≥ 12 were invited for interview, to determine the presence and severity of any depression (none, mild, moderate or severe). Further women who had borderline EPDS scores of 9–11, plus a random sample of those who scored between 0 and 8, were also invited for a SCAN interview, as part of the assessment of the performance of the EPDS (Figure 2).

FIGURE 2.

Procedure for SCAN interviews. CBA-F, cognitive behavioural approach face-to-face group; CBA-P, cognitive behavioural approach postal group; EPDS, Edinburgh Postnatal Depression Scale; PCA-F, person-centred approach face-to-face group; PCA-P, person-centred approach postal group; SCAN, Schedule for Clinical Assessment in Neuropsychiatry; SSRI, selective serotonin reuptake inhibitor.

Each day, after all of the questionnaires of the IG women were scored manually in the research office, all women who scored 9 or more on the EPDS were identified for an interview, plus a selection of those who scored less than 9. The SIs were given each woman’s name (separately from her PCT, practice and id code to protect privacy), address and her baby’s date of birth. The SI contacted each woman to arrange the interview time, unless the woman’s HV stated a preference for contacting the woman first. For women for whom the SI could not get a reply on the phone, or for those who did not have a phone, the SI wrote invitation letters asking them to contact her. The replies from the women were grouped as consented, declined, missed or assumed declined, if the SI had unsuccessfully tried to make contact many times.

Before her SCAN interview, each woman was asked to sign a consent form. If the outcome was moderate or severe the woman was given an information sheet after the SCAN interview, which explained that the interview indicated that she might be feeling depressed. The sheet gave brief details about psychological therapies and antidepressants and indicated that the woman should discuss her planned care with the HV or an antidepressant could be prescribed by her GP.

Application to the Multicentre Research Ethics Committee

In December 2002 the MREC application included a draft RIL, consent forms and questionnaires for the committee to consider. These drafts were to be further developed once the trial began, having been reviewed by the service users (consumer involvement) and the wording amended.

In January 2003 the MREC requested clarity on:

• the CG HV continuing to provide care currently given

• the use of the EPDS as part of the initial assessment and as an outcome

• the recruitment of women

• the protocol for women with severe depression

• the recruitment of subjects for interview

• infant outcomes

• visits to women who scored ≥ 12 at 6 weeks postnatally

• the Punjabi Postnatal Depression Screening Questionnaire (PPDSQ)110

• safeguards, to avoid approaching women who had a stillbirth or perinatal death.

Administration

Trial Advisory Group

The Trial Advisory Group (TAG), chaired by Professor Michael Barkham, met every 3 months throughout the duration of the trial. The purpose of the group was:

• To provide overall supervision for the trial on behalf of the trial sponsor (the Department of Health by way of the HTA programme)

• To ensure that the trial was conducted to the rigorous standards set out in the MRC guidelines for good clinical practice.

• To concentrate on the progress of the trial against the project plan, adherence to the protocol, participant safety and the consideration of new information of relevance to the research question.

• To provide advice, through its chair, to the principal investigator, the HTA programme and the host institution on all aspects of the trial.

In addition:

• Membership should include an independent chair, at least two other independent members, one or two principal investigators and, when possible, a consumer representative. Involvement of independent members provides protection for both trial participants and the chief investigator.

• Observers from the HTA programme and the host institution should be invited to all TSC meetings.

• Responsibility for calling and organising TSC meetings lies with the principal investigator.

• There may be occasions when the trial sponsor will wish to organise and administer these meetings for particular trials. In the HTA programme’s case this is unlikely, but they reserved the right to convene a meeting of the TSC in exceptional circumstances.

Cluster level recruitment and consent

Health visitors and GPs from 241 clusters expressed an interest in the trial, of whom 141 did not consent and participate in the study. There were 30 clusters allocated to CBA, 32 to PCA and 38 to control. Within the intervention clusters there were 13 allocated to CBA-F, 17 to CBA-P, 15 to PCA-F and 17 to PCA-P. Figure 3 illustrates the recruitment and allocation of clusters to each of the five groups, and the flow of the clusters up to 18 months’ follow-up, that is, the clusters in which women returned a questionnaire and contributed to overall data.

FIGURE 3.

Diagrammatic representation of cluster participation. CBA-F, cognitive behavioural approach face-to-face group; CBA-P, cognitive behavioural approach postal group; PCA-F, person-centred approach face-to-face group; PCA-P, person-centred approach postal group.

Baseline measurements at cluster level

The IMD 2004176 was used to compare the characteristics of the recruited GP clusters with those of the GP practices in the former Trent region and also with those of GP practices in England. The IMD measures deprivation for special census geographies called super output areas (SOAs). It combines indicators across seven domains into a single deprivation score and rank. The domains are income deprivation; employment deprivation; health deprivation and disability; education, skills and training deprivation; barriers to housing and services; living environment deprivation; and crime. Graphs were created using the functionality of geographical information systems (GIS) by coding IMD 2004 data to SOAs. These areas allow socially similar areas to be grouped together, allowing for more realistic patterns to emerge in the spatial data; it also allowed a GP practice to be tied to this measure. It is assumed that a practice placed in a SOA has the characteristics of that spatial unit rather than those of neighbouring ones. The source of the GP locations was an NHS website. 211 An automated extraction tool lifted and restructured the address and other information from the website before formatting and importing into a GIS.

Description of participating clusters

The clusters in which women contributed to the study are described below. The mean practice population was 7664 and the mean number of expected births per year was 79 (based on the actual number of births in the previous 2 years). The number of women in each cluster ranged from 1 to 140 (Table 12). Figures 4 and 5 indicate the distribution of registered practice populations and the total number of expected births per year in recruited clusters respectively. Figure 6 indicates the distribution of the number of women recruited per cluster.

TABLE 12 - Characteristics of GP practices in the PoNDER study

IMD, Index of Multiple Deprivation.

Expected births per year were based on the actual number of births in each practice in the previous 2 years.

Descriptor	Minimum	Maximum	Mean	SD
Total registered practice population	2300	19,000	7664	3662
Expected births per practice per yeara	21	157	79	34
Total number of women recruited to the study	1	140	42	29
IMD 2004 score for sample cases	3	63	24	15

FIGURE 4.

Distribution of registered populations for GP practices in the study.

FIGURE 5.

Distribution of total expected births in GP practices in the study.

FIGURE 6.

Distribution of number of women consented per practice in the study.

Individual level recruitment and consent to the trial

Exclusion criteria applied

In total, 89.3% of all antenatal women were eligible to take part and 10.7% were ineligible. The main reasons that women were not eligible were that they were changing GP practice or address (4.8%), they were aged less than 18 years (1.4%) or they had mental health issues (1.1%), which meant severe and enduring mental health problems such as bipolar disorder or schizophrenia (Table 14).

Baseline measurements of consented women at individual level

Of all of the eligible pregnant women, 53.3% consented. Among those for whom details were available, the greatest proportion of women who did not consent said that they were not interested (76%), with others saying that they were too busy (7%) or not replying (11%). Table 15 presents the characteristics of the women who consented.

In total, 15.1% of women who consented to take part had had PND previously versus 14.1% of all antenatal women. Table 16 presents the characteristics of women who both consented and returned a 6-week questionnaire.

Individual level follow-up

The overall participant flow chart (Appendix 1, Figures 27 and 28) illustrate the number of women in the study at baseline and follow-up. Of the 7649 eligible women in all clusters, 4084 (53.4%) consented and 3449 returned a 6-week questionnaire (88% return rate). A total of 2875 women (72%) returned a 6-month questionnaire; 2029 women (61%) returned a 12-month questionnaire; and 1097 women (56%) returned an 18-month questionnaire (Appendix 1, Figure 27).

A total of 595 women returned a 6-week questionnaire and scored ≥ 12 on the EPDS to become the at-risk women. The follow-up of the at-risk women is illustrated in Figure 7 and Appendix 1, Figure 29.

FIGURE 7.

Flow chart for at-risk women by group (n = 595). CBA, cognitive behavioural approach; EPDS, Edinburgh Postnatal Depression Scale; PCA, person-centred approach.

Primary outcome for at-risk women

Sensitivity analysis: imputation of missing 6-month EPDS data for at-risk women

In total, 595 at-risk women had a 6-week EPDS score ≥ 12 and 418 of these women also had 6-month follow-up EPDS scores. The results for these 418 women were therefore available for the primary statistical analysis. A sensitivity analysis was performed to impute the missing 6-month EPDS scores for the 177/595 (29.7%) at-risk women who were lost to follow-up. Two types of missing data imputation were performed:

• last observation carried forward (LOCF)

• regression imputation.

For most postnatal women symptoms of depression will naturally reduce over time, as seen in the CG at-risk women, among whom 54.4% (87/147) no longer had an EPDS score ≥ 12 at the 6-month follow-up. LOCF imputation represents the worst-case scenario, in which, for example, a 6-week EPDS score of 12 would be carried forward to be used as a woman’s missing 6-month EPDS score. This woman would therefore still be regarded as being in the above-threshold group of at-risk women.

Regression imputation is more logical for this group of postnatal women, as it better reflects the natural reduction in symptoms of depression over time. A regression imputation, based on 2659 women who returned both a 6-week and a 6-month EPDS, produced the following model (Figure 8):

EPDS 6 months = 2.287 ( SE 0.135 ) + 0.526 ( SE 0.17 ) × EPDS 6 weeks ( R 2 = 26.9 )

FIGURE 8.

Scatter plot of the relationship between the 6-month and 6-week EPDS scores (n = 2659) with the regression imputation line of best fit.

Using regression imputation, a woman with a 6-week EPDS score of 12 and a missing 6-month EPDS score would have a regression-imputed 6-month EPDS score of 8.6 [i.e. 2.287 + (12 × 0.526)]. Only women with a 6-week EPDS score ≥ 19 would have a 6-month EPDS score ≥ 12.

Table 20 shows the results of the sensitivity analysis for the imputation of missing data by LOCF and regression imputation compared with the primary statistical analysis. Using LOCF the results changed markedly and the observed treatment effect was smaller and not statistically significant. Using regression imputation the results were similar to those of the primary analysis although the treatment effect was smaller (odds ratio of 0.72 versus 0.62) and not statistically significant, with a p-value of 0.089 (compared with p = 0.036 for the observed data). The results for the regression-imputed model adjusted for covariates were statistically significant and very similar to the observed data (odds ratio of 0.62 versus 0.60).

TABLE 20 - Primary outcome: proportions of at-risk women with a 6-month EPDS score ≥ 12, control vs intervention, after LOCF and regression imputation of missing scores

EPDS, Edinburgh Postnatal Depression Scale; Int, intervention; LOCF, last observation carried forward imputation.

Adjusted for 6-week EPDS score, lives alone, history of postnatal depression, any life events.

Regression imputation based on the following model from 2659 women who completed both a 6-week and a 6-month EPDS: EPDS_{6 months} = 2.287 (SE 0.135) + 0.526 (SE 0.17) × EPDS_{6 weeks} (R² = 26.9).

	6-month EPDS score ≥ 12		Valid n	Difference (%)	95% CI	Odds ratio, int to control	95% CI	p-value
	Control group	Int group
Primary analysis
n	67	92
%	45.6	33.9		11.7	0.4 to 22.9	0.62	0.40 to 0.97	0.036
						0.60a	0.38 to 0.95a	0.028 (n = 409)a
Total n	147	271	418
LOCF
n	111	225
%	58.1	55.7		2.4	–6.6 to 12.7	0.90	0.62 to 1.31	0.58
						0.87a	0.59 to 1.27a	0.47 (n = 582)a
Total n	191	404	595
Regression
n	74	126
%	38.7	31.2		7.5	–1.3 to 16.4	0.72	0.49 to 1.05	0.089
						0.62a	0.41 to 0.96a	0.032 (n = 582)a
Total n	191	404	595

Six-month secondary outcomes for at-risk women

Twelve-month secondary outcomes for at-risk women

Of the 741 questionnaires not sent at 12 months, 597 (81%) were not sent because the women had not reached the 12-month postnatal follow-up time. Twelve-month outcomes were therefore available for 94 CG and 167 IG at-risk women.

Twelve-month EPDS scores for at-risk women: intervention versus control group

At 12 months, among the at-risk women who had an EPDS score on their returned 6-month questionnaires, the mean EPDS score was 10.6 (SD 6.2) for women in the CG and 8.1 (SD 5.6) for women in the IG (Table 25). This difference was statistically significant (p < 0.001). After adjusting for 6-week variables, the difference remained statistically significant (p = 0.005). Figure 9 shows the reduction in mean EPDS score over time in the intervention and control groups from 6 weeks to 12 months. Most of the benefit is gained at 6 months and is then maintained up to 12 months postnatally.

TABLE 25 - Twelve-month secondary outcomes for at-risk women: control vs intervention, unadjusted and adjusted

CORE-OM, Clinical Outcomes in Routine Evaluation Outcome Measure; DAS, Dyadic Adjustment Scale; EPDS, Edinburgh Postnatal Depression Scale; MCS, mental component summary; PCDI, parent–child dysfunctional interaction; PCS, physical component summary; PSI, Parenting Stress Index.

EPDS, SF-12, SF-6D, CORE-OM all adjusted for 6-week score, lives alone, history of postnatal depression, any life events.

Better health represented by a lower score in CORE-OM, EPDS, STAI. Better health represented by a higher score in DAS, PSI, SF-12 and SF-6D.

12-month outcome	Control			Intervention			Unadjusted			Adjusteda
	n	Mean	SD	n	Mean	SD	Difference	95% CI	p-value	Difference	95% CI	p-value
EPDS	94	10.6	6.2	167	8.1	5.6	–2.6	–4.3 to –0.9	0.003	–2.4	–4.1 to –0.7	0.005
SF-12 PCS	92	54.1	7.5	161	53.9	7.0	–0.2	–2.0 to 1.7	0.857	–0.4	–2.0 to 1.3	0.650
SF-12 MCS	92	40.8	10.9	161	44.9	10.9	4.1	1.3 to 6.9	0.004	3.8	1.0 to 6.6	0.008
SF-6D	93	0.7	0.1	165	0.8	0.1	0.0	0.0 to 0.1	0.007	0.0	0.0 to 0.1	0.013
CORE-OM well-being	94	1.4	0.9	167	1.1	0.9	–0.3	–0.5 to –0.1	0.013	–0.3	–0.5 to –0.1	0.013
CORE-OM risk	94	0.1	0.4	167	0.1	0.2	–0.1	–0.1 to 0.0	0.191	–0.0	–0.1 to 0.0	0.353
CORE-OM symptoms	94	1.1	0.8	167	0.8	0.7	–0.2	–0.5 to –0.1	0.010	–0.2	–0.4 to –0.0	0.021
CORE-OM functioning	94	1.2	0.8	167	0.9	0.8	–0.3	–0.5 to –0.1	0.005	–0.3	–0.5 to –0.1	0.011
CORE-OM total score	94	1.1	0.7	167	0.7	0.6	–0.3	–0.4 to –0.1	0.005	–0.2	–0.4 to –0.0	0.015
State anxiety (STAI)	93	45.0	13.2	162	40.7	12.6	–0.4	–8.1 to –0.7	0.019	–4.2	–7.8 to –0.5	0.025
Trait anxiety (STAI)	89	43.2	11.4	162	39.3	10.8	–4.0	–7.3 to –0.6	0.019	–3.9	–7.2 to –0.5	0.023
DAS Likert	89	3.3	1.4	158	3.8	1.4	0.5	0.1 to 0.8	0.011	–0.2	–0.6 to 0.1	0.217
DAS	89	19.2	3.2	154	18.2	3.0	–0.9	–1.8 to –0.1	0.027	0.0	–0.7 to 0.8	0.965
PSI parenting distress	93	38.6	9.5	158	42.5	10.0	3.9	1.4 to 6.4	0.002	3.9	1.5 to 6.3	0.002
PSI PCDI	94	54.3	7.1	163	55.9	6.1	1.6	–0.3 to 3.1	0.055	1.6	0.0 to 3.2	0.047
PSI difficult child	91	47.6	8.4	164	50.5	7.4	3.0	1.0 to 5.0	0.003	2.9	1.2 to 4.7	0.001
PSI total stress	90	140.7	21.4	156	148.7	19.2	8.1	3.1 to 13.0	0.001	8.1	3.6 to 12.6	0.000

FIGURE 9.

Mean EPDS scores for at-risk women at 6 weeks, 6 months and 12 months by intervention and control group.

Eighteen-month secondary outcomes for at-risk women

Eighteen-month outcomes for at-risk women: intervention versus control

Of the 2113 questionnaires not sent at 18 months, 1879 (88.9%) were not sent because the women had not reached the 18-month postnatal follow-up time. The EPDS was not administered at 18 months as it is not validated for use beyond 12 months. There were some statistically significant differences between the CG and the IG at 18 months for some of the PSI subscales. These are presented in Table 26. Figures 10–13 illustrate the changes in mean scores for at-risk women from 6 weeks to 18 months for the secondary outcomes SF-12 MCS, SF-12 PCS, SF-6D and CORE-OM respectively.

TABLE 26 - Eighteen-month secondary outcomes for at-risk women: control vs intervention, unadjusted and adjusted

6W, 6 weeks; 18M, 18 months; CORE-OM, Clinical Outcomes in Routine Evaluation Outcome Measure; MCS, mental component summary; PCDI, parent–child dysfunctional interaction; PCS, physical component summary; PSI, Parenting Stress Index.

EPDS, SF-12, SF-6D, CORE-OM all adjusted for 6-week score, lives alone, history of postnatal depression, any life events.

Better health represented by a lower score in CORE-OM, EPDS and STAI. Better health represented by a higher score in PSI, SF-12 and SF-6D and SF-36.

Outcome	Control			Intervention			Unadjusted			Adjusteda
	n	Mean	SD	n	Mean	SD	Difference	95% CI	p	n	Difference	95% CI	p
SF-36 PCS 6W	143	48.5	10.9	265	50.1	9.4
SF-12 PCS 18M	46	52.1	12.9	86	56.0	8.6	4.1	1.1 to 7.1	0.007	86	2.9	–0.1 to 6.0	0.062
SF–36 MCS 6W	143	29.4	9.2	265	29.1	8.0
SF-12 MCS 18M	46	40.2	7.1	86	39.5	5.4	–0.8	–2.9 to 1.3	0.463	86	–0.7	–2.7 to 1.3	0.509
SF-6D 6W	142	0.59	0.08	268	0.60	0.1
SF-6D 18M	47	0.73	0.15	87	0.78	0.15	0.04	–0.00 to 0.09	0.063	87	0.03	–0.01 to 0.07	0.122
CORE-OM well-being 6W	146	2.1	0.74	269	2.0	0.7
CORE-OM well-being 18M	47	1.4	1.0	88	1.2	1.0	–0.2	–0.6 to 0.2	0.309	88	–0.1	–0.4 to 0.2	0.523
CORE-OM risk 6W	145	0.2	0.33	269	0.14	0.3
CORE-OM risk 18M	47	0.2	0.4	88	0.1	0.3	–0.1	–0.2 to 0.1	0.328	88	–0.0	–0.1 to.01	0.670
CORE-OM symptoms 6W	146	1.6	0.6	269	1.6	0.6
CORE-OM symptoms 18M	47	1.0	0.9	88	0.8	0.8	–0.2	–0.5 to 0.1	0.192	88	–0.1	–0.4 to 0.1	0.212
CORE-OM functioning 6W	146	1.6	0.6	269	1.5	0.6
CORE-OM functioning 18M	47	1.1	0.9	88	0.9	0.8	–0.2	–0.5 to 0.2	0.290	88	–0.1	–0.3 to 0.1	0.497
CORE-OM total score 6W	146	1.4	0.5	269	1.4	0.5
CORE-OM total score 18M	47	1.0	0.8	88	0.8	0.7	–0.2	–0.4 to 0.1	0.228	88	–0.1	–0.3 to 0.1	0.374
State anxiety (STAI) 18M	46	42.7	14.5	85	40.8	14.5	–1.7	–6.9 to 3.5	0.521	85	–1.3	–5.8 to 3.2	0.578
PSI parenting distress 18M	47	38.8	9.9	84	43.3	10.1	4.6	0.8 to 8.4	0.017	84	4.6	1.3 to 8.0	0.006
PSI PCDI 18M	46	53.5	7.6	87	55.2	6.3	1.4	–1.0 to 3.8	0.243	87	1.2	–0.9 to 3.4	0.263
PSI difficult child 18M	47	46.0	8.6	85	49.1	8.0	3.3	0.4 to 6.2	0.027	85	3.2	0.3 to 6.0	0.028
PSI total stress 18M	46	138.1	23.2	82	147.2	21.2	9.3	1.1 to 17.4	0.026	82	9.0	1.6 to 16.4	0.017

FIGURE 10.

Mean SF mental component summary (MCS) score for at-risk women from 6 weeks to 18 months, by intervention and control.

FIGURE 11.

Mean SF physical component summary (PCS) score for at-risk women from 6 weeks to 18 months, by intervention and control.

FIGURE 12.

Mean SF-6D scores for at-risk women from 6 weeks to 18 months, by intervention and control.

FIGURE 13.

Mean CORE-OM total score for at-risk women from 6 weeks to 18 months, by intervention and control.

Figures 14 and 15 illustrate the changes in mean state anxiety and PSI scores for at-risk women from 6 months to 18 months.

FIGURE 14.

Mean state anxiety scores for at-risk women from 6 months to 18 months, by intervention and control.

FIGURE 15.

Mean Parenting Stress Index total scores for at-risk women from 6 months to 18 months, by intervention and control.

Remission and relapse to 18 months

The results from the unvalidated measure of at-risk women’s self-reported health for women who scored ≥ 18 on the 6-week EPDS are shown in Figure 16. This illustrates changes in symptoms of depression from the time of a baby’s birth to 18 months postnatally on a scale from 0 to 4, where 0 indicates no symptoms and 4 indicates severe symptoms of depression.

FIGURE 16.

Remission and relapse scores over 18 months for at-risk women, intervention vs control group.

Threshold score of 12

There were 860 SCAN interviews performed for the study, 355 with at-risk women. Of these 355 at-risk women the outcome for 18.6% (66/355) was mild depression and for 14.1% (50/355) was moderate or severe depression. That is, among all the at-risk women who had a SCAN interview, 32.7% (116/355) had an outcome of depression (Tables 27 and 28). Apart from depression, the SCAN indicated depersonalisation syndrome, generalised anxiety disorder, nightmares, non-organic insomnia or panic disorder in some women, some of these in conjunction with depression. In total, 38.3% (136/355) of at-risk women and 19.8% (170/860) of all women who were interviewed had some outcome on the SCAN (Table 29). The SCAN outcome was no depression for 219 women with a range of EPDS scores from 12 to 22. A total of 80 women with an outcome of mild depression had EPDS scores ranging from 5 to 25. Similarly, 52 women who had a SCAN outcome of moderate depression (either alone or with another outcome) had EPDS scores ranging from 7 to 27. Of the five women who had an outcome of severe depression, EPDS scores ranged from 7 to 25.

Using a threshold of 12 (score ≥ 12), the sensitivity of the EPDS (the proportion of depressed women who scored ≥ 12 on the EPDS) was 0.866 (CI 0.808 to 0.923) and the specificity (the proportion of non-depressed women who scored ≤ 11 on the EPDS) was 0.671 (CI 0.637 to 0.705) (Table 30). The sensitivity for detecting moderate or severe depression using the threshold of 12 was 0.926 (CI 0.856 to 0.996), whereas the specificity was 0.622. The positive predictive value [proportion of women above the threshold of 12 on the EPDS (n = 355) who had an outcome of depression (n = 116)] was 32.7%.

Threshold score of 13

Using a threshold of 13 (score ≥ 13) (Tables 31–34) the sensitivity of the EPDS was 0.791 and the specificity was 0.755 (Table 33). The sensitivity for detecting moderate or severe depression using the threshold of 13 was 0.852, whereas the specificity was 0.705 and the positive predictive value was 37.3% (106/284).

At-risk women who returned a 6-month questionnaire

A total of 404 women who scored ≥ 12 on the 6-week postal EPDS were at-risk women, of whom 274 (67.8%) returned a 6-month questionnaire.

Health visitor administration of 8-week EPDS to at-risk women

The HV protocol stated that HVs should repeat the administration of the EPDS face-to-face at 8 weeks postnatally for all at-risk women to determine which women were eligible for the intervention. Of all of the 404 at-risk women, 70.8% (286/404) had an 8-week EPDS score and for 29.2% (118/404) the score was missing. Of those who had an 8-week EPDS score, for 60.5% of women (173/286) the score was < 12 and for 39.5% (173/286) the score was ≥ 12 (Table 35). Of the missing scores, 48% (57/118) were missing because the HV-administered face-to-face 6-week EPDS score was < 12. A further 27% (32/118) were missing for reasons to do with the women being absent or declining, and 20% (24/118) were absent for reasons to do with the HV being unavailable.

Health visitor psychological intervention sessions offered to at-risk women

For the 395 at-risk women for whom data were available, 50% (197/395) were offered a psychological intervention session and therefore 50% (198/395) were not offered a session. In total, 31% of all at-risk women (121/395) received at least one psychological intervention session and, of those offered, 39% (76/197) declined (Table 35). There were 259 intervention sessions delivered in the CBA group and 242 intervention sessions delivered in the PCA group.

Figure 17 illustrates that, of the at-risk women who returned a 6-month EPDS, 46% (125/274) were offered intervention sessions and 29% (80/274) accepted the intervention sessions. Among the 32% (130/404) of women who did not return a 6-month EPDS score and therefore were no longer included in the trial, 55% (72/130) were offered sessions and 32% (41/130) received sessions. Figure 17 also illustrates that HVs offered sessions to women who had no 8-week EPDS score and who had an 8-week EPDS score < 12. Among all of the at-risk women, of whom according to the SCAN interviews most were not depressed, 197/404 (48.88) were offered sessions and 61% (121/197) accepted.

FIGURE 17.

Flow chart of at-risk women: 8-week EPDS assessment and intervention sessions offered.

At-risk women receiving other support

About 22% (84/380) of the women had also been prescribed antidepressants, but not all of the women took these. Around 40% (148/374) were also receiving support apart from the HV. Most frequently (19.5%) the women received support from a GP (73/374); 8% of women were also in receipt of other mental health services and 6% attended a postnatal support group. Of possibly the greatest concern were the seven women who had an 8-week EPDS score ≥ 12 but who were not offered the psychological intervention sessions by the HV. Table 36 indicates that these seven women were supported by a GP, counsellor or mental health worker or were not classified as depressed.

Preference for psychological intervention or antidepressants

There was no evidence that women preferred an antidepressant to the HV psychological intervention.

Six-month EPDS outcome for all women

Six-month EPDS outcome for all women: intervention versus control

At 6 months, among all of the women who had returned both a 6-week and a 6-month questionnaire, 16.4% in the CG scored ≥ 12 on the EPDS versus 11.7% in the IG. The absolute difference was 4.7% (95% CI 0.7 to 8.6). This effect was statistically significant (p = 0.003). After adjusting for covariates – 6-week EPDS score, living alone, previous history of PND and any life events – the point estimate of the odds ratio for the IG effect was relatively unchanged (at around 0.67) and this effect remained statistically significant (Table 37).

Figure 18 illustrates the change in EPDS score over time in the control and intervention groups for at-risk women and all women.

TABLE 37 - Six-month EPDS outcome: proportion of all women with an EPDS score ≥ 12 at 6 months, by intervention or control (n = 2659)

Int, intervention.

n = 2659, adjusted for 6-week EPDS score.

n = 2624, adjusted for 6-week EPDS score, lives alone, history of postnatal depression, life events.

6-month EPDS score	Int group	Control group	All	Absolute difference (%)	95% CI	Odds ratio, int to control	95% CI	p-value
< 12, n (%)	1540 (88.3)	764 (83.6)	2304 (86.6)
≥ 12, n (%)	205 (11.7)	150 (16.4)	355 (13.4)	4.7	0.7 to 8.6	0.67	0.51 to 0.87	0.003
						0.68a	0.52 to 0.88a	0.004a
						0.67b	0.52 to 0.86b	0.002b
Total, n	1745	914	2659

FIGURE 18.

Mean EPDS scores for at-risk women and all women from 6 weeks to 12 months by intervention and control group.

Six-month secondary outcomes for all women

Twelve-month secondary outcomes for all women

Of the 741 questionnaires not sent at 12 months, 597 (81%) were not sent because the women had not reached the 12-month postnatal follow-up time. Twelve-month outcomes were therefore available for 593 CG and 1118 IG women.

Eighteen-month secondary outcomes for all women

Secondary outcomes for women’s partners

Eighteen-month infant outcomes

Intervention process monitoring