Evaluation of trained volunteer doula services for disadvantaged women in five areas in England: women's experiences

Disadvantaged childbearing women experience barriers to accessing health and social care services and face greater risk of adverse medical, social and emotional outcomes. Support from doulas (trained lay women) has been identified as a way to improve outcomes; however, in the UK doula support is usually paid-for privately by the individual, limiting access among disadvantaged groups. As part of an independent multi-site evaluation of a volunteer doula service, this study examined women's experiences of one-to-one support from a trained volunteer doula during pregnancy, labour and the post-natal period among women living in five low-income communities in England. A mixed methods multi-site evaluation was conducted with women (total n = 137) who received the service before December 2012, using a combination of questionnaires (n = 136), and individual or group interviews (n = 12). Topics explored with women included the timing and nature of support, its impact, the relationship with the doula and negative experiences. Most women valued volunteer support, describing positive impacts for emotional health and well-being, and their relationships with their partners. Such impacts did not depend upon the volunteer's presence during labour and birth. Indeed, only half (75/137; 54.7%) had a doula attend their birth. Many experienced volunteer support as a friendship, distinct from the relationships offered by healthcare professionals and family. This led to potential feelings of loss in these often isolated women when the relationship ended. Volunteer doula support that supplements routine maternity services is potentially beneficial for disadvantaged women in the UK even when it does not involve birth support. However, the distress experienced by some women at the conclusion of their relationship with their volunteer doula may compromise the service's impact. Greater consideration is needed for managing the ending of a one-to-one relationship with a volunteer, particularly given the likelihood of it coinciding with a period of heightened emotional vulnerability.

• Support from doulas (trained lay women) has been associated with improved outcomes; however, doula support is usually paid-for in the UK, limiting access among disadvantaged groups.
• Few studies have explored doula support in settings where midwives are the lead health professionals.
What this paper adds • Women from low-income communities using a volunteer doula service alongside routine maternity services reported predominantly positive impacts that did not depend upon volunteers attending labour.
• Women described feelings of loss when the relationship ended.
• Managing the ending of a one-toone relationship with a volunteer requires greater consideration given its potential to compromise impact.

Introduction
In the UK, most women access maternity care through the National Health Service; this is free at the point of access. Midwives work across hospital and community settings, co-ordinate the care provided during pregnancy, birth and the early postnatal period and are the lead healthcare professionals for women whose pregnancies are considered low risk. Women may also receive other statutory services, e.g. from General Practitioners, health visitors and social services. Disadvantaged women (including those with complex social needs such as social deprivation, lone parenting, substance misuse, mental health problems, domestic abuse, asylum seekers and refugees) are less likely to access routine services and face increased risk of poorer maternal and child health outcomes (CEMACH 2009, Downe et al. 2009, Hodnett et al. 2010, Centre for Maternal and Child Enquiries, 2011, O'Hara & Mccabe 2013. The National Institute for Health and Care Excellence (NICE) in England and Wales recommended that service provision for pregnant women with complex social needs be better integrated both within the NHS and between the NHS and those services provided in the community by not-for-profit organisations (described in the UK as the voluntary or third sector) (NICE, 2010). This fits with a move in highincome countries towards using lay health workers (i.e. those with some training, but no formal professional training or qualification) to engage minority communities and support those with complex needs (Glenton et al. 2013). Recognising the limited evidence base, NICE identified two research questions that related to this: What effect does involving third [voluntary] sector agencies in providing support and co-ordination of care for vulnerable women have on outcomes? Is intervention and/or family support provided by statutory and third [voluntary] sector agencies effective in improving outcomes for women and their babies? (NICE, 2010) The research reported here examined a voluntary sector service where disadvantaged childbearing women are allocated a volunteer 'doula' (the term adopted by the service) with the aim of enhancing support and well-being, and improving the uptake of health and social services. The doulas are volunteers from the local community who receive accredited training, funded by the service; as such they are considered lay rather than professional. Training covers preparation for and support during labour and birth, breastfeeding, child protection, domestic abuse awareness, cultural diversity and communication skills. Salaried service staff match a volunteer (and, sometimes, a backup volunteer) to each woman according to needs and practicalities; and facilitate an initial meeting between the woman and volunteer and mentor the volunteer throughout the support period, typically from the sixth month of pregnancy until 6-week post-partum. Service policy stipulates that doulas and women do not have continued contact beyond the ending of the support period.
In common with models of doula support in previously published research (Sosa et al. 1980, Hodnett et al. 2007, Steel et al. 2015, the volunteers offer emotional support, information and physical support, but do not provide clinical care. The volunteer doulas differ from traditional schemes in two main ways. First, support extends over a long period rather than being focused on birth and the immediate post-partum; the birth may or may not be attended by the doula. Second, the support offered is more diverse and seeks to optimise women's use of both health and social care services; thus the role includes working closely with existing services, facilitating communications between the woman, her partner and health and social care providers, and signposting to other services, including voluntary and community organisations. In these respects, the closest similar model is the communitybased doulas, an extended doula model which has largely focused on supporting young mothers or those from ethnic minorities (e.g. Breedlove 2005, Gentry et al. 2010, Wen et al. 2010, Akhavan & Edge 2012. Support in the scheme evaluated here can include: home visits; telephone contact; giving information about services and accompanying to appointments; going for walks and trips to cafes (to reduce social isolation); giving information about pregnancy, labour, birth and looking after the baby; providing physical and emotional support during labour and birth; giving practical help with baby equipment; and breastfeeding support.
Previous research has shown doula support to be associated with more positive feelings about labour, increased feelings of control and confidence as a mother and less post-natal depression and anxiety (Hofmeyr et al. 1991, Wolman et al. 1993, Langer et al. 1998, Gordon et al. 1999, Scott et al. 1999. However, research gaps remain. Several studies focused on intrapartum in-hospital support. A recent critical review (Steel et al. 2015) identified the relative absence of research examining the outcomes for women receiving doula support in home or community settings. The review, which focused on 'fee-forservice' doulas, also noted that, despite the focus of doula care being on social and emotional support, research has focused on medical outcomes (i.e. pregnancy and birth outcomes). Alongside the relative dearth of qualitative evidence around recipients' experiences of support is a lack of research into how change is achieved; a notable exception being a grounded theory study identifying the use of several problem-solving strategies used by community-based doulas working with adolescent mothers (Gentry et al. 2010). In addition, there is a paucity of UK evidence, where doula support is offered alongside midwifery care.
We conducted an independent multi-site evaluation, informed by Realistic Evaluation (Pawson & Tilley 1997), which was funded by the National Institute of Health Research. The full report is available (Spiby et al., 2015). One of the aims of the evaluation was to examine the health and psychosocial impacts for women who used the volunteer doula service. Analysis of the service databases suggested some clinical outcomes of doula supported women were improved relative to the local population; the caveats around those findings are discussed elsewhere (Spiby et al., 2015). This paper focuses on the experiences of the women who used the service; specifically, the areas of impact and the nature of the relationship that may offer insights into how such outcomes occur.

Settings
The doula service was originally set up in site A in 2006 and subsequently in 2011 rolled-out to four other sites (W, X, Y and Z); all of which are lowincome communities. The services are predominately run by voluntary sector organisations. Volunteer doula support is provided free of charge to women and is additional to routine statutory and voluntary services. Women may self-refer but are typically referred by another statutory or voluntary agency, usually due to: being unsupported and potentially birthing alone; experiencing health or social problems; or having particular concerns about labour and birth. At two sites, services are restricted to women from ethnic minority groups and a third serves an area with a very large ethnic minority population.

Ethics and governance
Approval for the study was obtained from the West Midlands Research Ethics Committee (reference 12/ WM/0342) and governance permissions were obtained at each research site.

Eligibility
All women who had used the service and whose support had ceased prior to the period of data collection (December 2012-April 2013) were potentially eligible. Exceptions were those whose personal circumstances (e.g. stillbirth or certain welfare concerns) meant that contact might increase stress or vulnerability.

Procedure
Women were invited to complete a questionnaire and/or be interviewed. Questionnaires were completed with the assistance of a researcher or interpreter (by telephone) or self-completed (by post). Interpreter services were favoured over written translation due to the large number of languages used and because service staff indicated that literacy barriers were not limited to English language.
Service staff approached women using the recruitment procedure shown in Figure 1 and completed anonymised monitoring logs detailing the dates of contact, reasons for non-approach and reasons for not sending out research packs. Reminder postcards were sent out 3 weeks after the initial packs. Women were able to ask questions about the research before deciding whether to participate. All women indicating interest in being interviewed were provided with further information and written informed consent was secured prior to interview. Interviews were audiorecorded and transcribed.

Development of data collection materials
Following a Realistic Evaluation perspective, literature review and early discussions with key informants were used to develop topics of interest and a priori hypotheses concerning 'what works for whom, in what circumstances' (Pawson & Tilley 1997); key informants included service staff and reference panels comprised of volunteer doulas and women who had used the service. The topics and hypotheses were subsequently explored by questionnaire and interviews with participants. No validated questionnaires exist that would enable evaluation of all aspects identified for investigation. A questionnaire was developed and piloted with the women's reference panel. The questionnaire included both open and closed question formats. Due to length, women using assisted telephone completion were asked a reduced set of questions. A semi-structured interview topic guide was developed, to explore in greater detail women's experiences of some of the issues raised by key informants, including how the volunteer role was similar to and contrasted with support from family, partner and professionals.

Analysis
A mixed methods evaluation was used, whereby the method was considered secondary to the research question, reflecting a pragmatic perspective (Johnson et al. 2007, Morgan 2007. Quantitative questionnaire data were analysed using descriptive statistics and chi-squared with Yates' continuity correction using SPSS version 20 (Spss Ibm Corp, 2011). Qualitative data (including open-ended questionnaire comments and transcription data) were analysed using content analysis (Grbich 1999). Anonymous participant identifiers were assigned in the format: data source (Q for questionnaire and I for interview), identification number and study site. The open text questionnaire responses were tabulated to show horizontally all of an individual's responses to the questions and vertically all of the responses received to any question. This facilitated coding of themes on a question-byquestion basis, identification of disconfirming responses and the exploration of linked patterns between questions. The transcripts from the interviews were read and reread to gain a detailed familiarity with the overall accounts, and then systematically coded manually both deductively to identify themes related to survey questions and a priori hypotheses and inductively to identify emerging themes (Elo & Kyng€ as 2008). These themes were grouped and collapsed into higher order conceptual Where permission was obtained (via service staff or expression of interest when completing questionnaire), research team contacted women directly to provide further details about group interviews.
Interviews were held in community venues identified by service staff and were audio-recorded.
Shopping voucher sent to woman on receipt of completed questionnaire and/or provided at attendance of group interview, to thank for participation.
Service staff telephone woman using last known contact details (using bilingual staff or interpreter where required, as indicated by service records) No reply (following three attempts by call or text message) Speaks with woman and introduces the research; i) asks willingness for research pack to be sent or offers assisted completion; ii) asks willingness for research team to make contact about group interviews  themes with subthemes. The findings of the qualitative and quantitative analyses were integrated to provide a comprehensive narrative of women's experiences.
Impacts presented here include: emotional health and well-being; supporting partners and women's relationships with their partners; endings and loss. Insights into the nature of the relationship that may inform how these impacts occur are also presented.

Questionnaire response rate
In total, 627 women had used the service. Of these, 578 (92.2%) were sent a postal questionnaire for selfcompletion or were contacted by an interpreter or researcher for assisted telephone completion (see Table 1). Reasons for not making contact or sending the questionnaire were women's circumstances (e.g. stillbirth) and failure to make telephone contact with women who required an interpreter or did not have address details held by the services.
Questionnaires were completed by 136 women; this represented 21.7% of women who had used the service. One in eight questionnaires were completed by telephone; the majority using an interpreter (see Table 1). Most women who were interviewed (11/12) also completed a questionnaire.

Sample characteristics
Sample characteristics were gathered by questionnaire and are reported in Table 2. This was an ethnically diverse sample with 33 countries of birth and 29 main languages listed; 41.0% did not have English as a main language. Reflecting the service's emphasis on women in situations of disadvantage, including a lack of support, 52.9% reported not having a supportive partner at the time of the pregnancy and 16.8% reported having no supportive friends or family at all. Less than half of the women (40.7%) were primiparous. Site A's service database indicated that multiparous women and older women were overrepresented among questionnaire respondents. The majority of women had been introduced to the service between 2010 and 2012; earlier introductions (n = 23) were limited to the original site, reflecting the service's histories.

Description of the volunteer support intervention
The stages, intensity and nature of volunteer doula support are shown in Table 3. Support in all three stages of the childbearing episode (i.e. antenatal, intrapartum and post-natal support) was most common (47.8%), followed by support during pregnancy and the post-natal period, without intrapartum support (26.5%). Of the 122 women whose support commenced during pregnancy, only 75 (61.5%) had their birth attended by a volunteer. This largely reflected women's preferences with just nine women reporting that they had wanted the doula there but that it had not been possible: because the birth happened sooner than anticipated (n = 5); because only one birth partner was allowed (n = 3); or because the doula was unavailable (n = 1).

Impacts of volunteer support
Impact: Emotional health and well-being. The qualitative data illustrated the significance of volunteer support for emotional health and well-being and this was not dependent on the doula being present for labour and birth. Benefits were particularly evident for women with little other support, but were also found for women who had involved partners or mothers, particularly those women with previous negative experiences of childbearing. Many described the ways in which change occurred, offering insights into mechanisms. The volunteer was someone to talk to and to Many women commented on how volunteer support helped them to feel more in control of their maternity care through becoming more aware of their choices; influenced their beliefs in their own physical abilities around birth and confidence for parenting by supporting their choices; and facilitated communication with health professionals, helping to navigate services. Such mechanisms were found both for firsttime mothers and mothers who had previously experienced a difficult birth: Gained confidence and belief in myself to deliver naturally and once my baby was born to get out the house with two babies. (Q334A)  Many spoke of their sadness about the ending of a close relationship. Some felt 'a little discarded' (Q380Y) by this 'temporary friendship' (I337A): I found it really hard actually, I kept asking if I could keep in touch with her . . . but we couldn't . . . once a friend they become a friend don't they and that's it. (I319A) There was a day she told me that I'm not allowed to get in contact with her, that is not how they do their services, I cried . . . oh, I really miss her. (I366Y) And is not fair according to [service] policy, that when you finish the last day that's it . . . She was more than a doulalike family. (Q336X) By contrast, other women found that the support had ended at the right time: The ending was in the right time, after I felt confident with my baby. (Q408W) This was particularly likely for women who primarily wanted information from their doula, rather than emotional support, and women at the one site with an extended post-natal support period of 3 months.
We hypothesised that endings would be facilitated by having greater preparation. Key informants identified various ways in which doulas prepared women for the ending of the service such as providing an account of their time together or photos. Women for whom a memento had been provided were not less likely to feel that support had ended too soon (31.0% vs. 37.5%; v 2 = 0.24, df = 110, P = 0.63). The relationship between having something provided and wanting to stay in touch with their volunteer reached borderline significance (72.5% vs. 52.4%; v 2 = 3.78, df = 110, P = 0.05). The finding that mementos did not appear to facilitate endings or reduce feelings of loss may suggest that these acts reflected the quality of the relationship rather than preparation per se.
Women proposed two ways to improve endings: timing the ending to woman's needs (e.g. following operative birth), or permitting some contact beyond the ending of support; for example, a one-to-one informal meeting, or a reunion attended by several women and their volunteers. Some women framed this in terms of wanting to be able to thank the volunteer by showing her the long-term impact of her support: Understanding the relationship Understanding the relationship: how the volunteer is viewed. Women were asked to choose all that applied from a list describing how they viewed their volunteer. Most saw her 'as a friend' (88/118; 74.6%); other answers were 'like a professional' (32.2%); 'like a family member' (31.4%); 'like a sister' (21.2%); 'like a mother' (17.8%); 'like an advocate' (17.8%); 'someone like me' (16.9%); and 'like a role model' (14.4%).
Most of those viewing the volunteer as 'like a family member' had wanted to stay in touch (mother: 90.5%; sister: 91.7%; friend 69.0%; professional 59.5%). Viewing the doula as like a mother appeared strongly linked to whether that role was missing in the woman's own network. None of the 21 women likening the volunteer to a mother had a supportive mother available during their pregnancy and no one with a supportive mother described the role in this way. Women with supportive family or friends nonetheless valued their volunteer's support; volunteers were better informed about pregnancy and birth, talked through options and supported the woman's choice in a non-directive way, whereas family and friends may have their own needs and agenda.
During interview discussions women contrasted volunteers' support with health professionals'. They valued the greater accessibility and continuity offered by volunteers, considered 'the one constant person' (I315A). Volunteers were largely viewed as focused completely on the woman ('just there for you', I341Y) with no competing agenda, promoting trust. Many women felt that they could ask their volunteer about anything, including beyond the 'medical things' (I486W), whereas they sometimes felt embarrassed or lacked confidence to ask health professionals who were perceived to be busy or dismissive.
Understanding the relationship: timing of support. We hypothesised that the volunteer support may not 'work' where a match happened late in the antenatal period and there was not time to establish a relationship. One-third (38/115; 33.0%) felt the relationship would have been different if they had met sooner and 22.6% (26/115) felt that the relationship would have been different if they had met later. Some women felt that meeting later would not have influenced the relationship because they met relatively late anyway, just shortly before the birth. Overwhelmingly, women felt that the relationship would have been better for meeting sooner; either to gain the benefits of support earlier in pregnancy or establish the relationship sooner, ensuring the opportunity to develop 'trust' (Q332A), get to know each other (Q448A) and 'bond' (Q423A). Consistent with this, some women reported feeling less comfortable with the backup doula because of lacking the opportunity to develop a relationship.
Negative experiences. A small proportion of women reported negative experiences. Fifteen of 129 women (11.6%) reported that the service had not helped them in the way they had hoped. Rating their experience of support from zero (very poor) to five (very good), 11.4% (15/132) rated at three and 2.3% (3/132) rated less than three. Most commonly it was the volunteer's unreliability or inability to provide continuity that was criticised. Some women had been disappointed at the limitations of the service (e.g. not assisting with household chores or providing care for older children) and some felt inhibited about asking for more support, knowing that volunteers were unpaid. Indeed, several women, including those reporting positive experiences overall, expressed feelings of guilt about accepting support from a volunteer without the ability to reciprocate.

Discussion
Most women reported positive impacts on their emotional well-being; including combating feelings of depression, having fears allayed, and building confidence and self-esteem. While similar benefits have been reported elsewhere (Hofmeyr et al. 1991, Wolman et al. 1993, Langer et al. 1998, Gordon et al. 1999, Scott et al. 1999, a key finding of this study is that such benefits did not depend upon doulas being involved in the labour and birth. Benefits appeared to be achieved through listening by someone who was non-judgemental and non-directive, relief of isolation, information provision, supporting women's choices and help navigating statutory and other services. These findings resonate with Gentry et al. (2010) who through interviewing adolescent mothers supported by community-based doulas identified the use of problem-solving strategies including active listening, assuring, affirming, advising and advocating.
Women also described the mechanisms by which woman-partner relationships were strengthened; including through the sharing of roles, alleviating concerns and promoting communication. The need to involve fathers in pregnancy, childbirth and the transition to parenthood is increasingly recognised by national UK and international policy (Steen et al. 2012). The current research suggests that volunteer doula services may offer a route to supporting involvement, consistent with reports of the Ounce Home Visiting and Doula Program in the US (The Ounce 2014). Research is needed on perceptions of doula support from the perspectives of partners and other family members (Steel et al. 2015) and how these family relationships may interact with the impacts of the support (Wen et al. 2010) Few women reported negative experiences or dissatisfaction although we recognise that this may partly reflect self-selection sampling bias and that women are often reluctant to be critical of their care (Green 2012). While there were instances of disappointment with the lack of assistance with household chores, as has been reported with lay workers in the context of health visiting (Mackenzie 2006), dissatisfaction was mainly related to perceiving the volunteer as unreliable or not having as much contact with the volunteers as they wished; something that women felt was harder to negotiate when support was delivered by a volunteer.
Understanding how women viewed their volunteers offered insights into how support 'worked', from a theoretical perspective (Pawson & Tilley 1997). Women frequently likened the volunteer to a family member or friend, consistent with the literature on volunteers and lay workers in the context of childbearing (Taggart et al. 2000, Perkins & Macfarlane 2001, Meier et al. 2007, Hazard et al. 2009, Gentry et al. 2010. Friendship was a central theme here and we note the overlaps between the current volunteer role and other community-based support programmes, such as those that use volunteer befrienders for women who may find it difficult to access or engage with services (Coe & Barlow 2013). For some women, however, the concept of friendship was challenged by the unidirectional and unbalanced nature of this relationship; an observation lacking in the doula literature.
Few studies have explored doula support in settings where the midwife is the lead health professional. Here, support from volunteers was contrasted with health professionals' with distinctive features of doula support being continuity, not feeling time pressured, feeling able to 'ask anything', feeling their choices were supported and seeing the doulas as more reliable and trustworthy. These findings resonate with studies of lay support for disadvantaged childbearing women in high-income countries; including, community-based doulas in the USA (Gentry et al. 2010), home visits in Australia (Taggart et al. 2000) and the USA (Sheppard et al. 2004), and infant feeding support in the UK (Beake et al. 2005). The greater continuity afforded by doulas compared with midwives has been reported elsewhere in a Swedish study (Lundgren 2010).
While participants were largely favourable towards the volunteer doula support and valued the continuity provided, it was striking that women commonly reported feelings of loss around the ending of support, which could constitute a negative impact. Volunteer support was valued regardless of whether women had support from their friends or family. The aspect of support often valued most highly was the one-to-one relationship. Its ending could be particularly difficult for some women, particularly those who viewed the volunteer as like a mother or where there were continuing practical needs, for instance, following an operative birth. Even women who felt well-prepared to move on independently and did not have continuing support needs could still feel saddened by the absence of opportunity for any contact with the volunteer in the future.
These findings highlight the challenges noted elsewhere in the volunteer and lay worker literature around ways of working that hinge on a close relationship between worker and recipient and the need to consider further the management of emotional relationships and boundaries (Perkins & Macfarlane 2001, Heslop 2006, Mitchell & Pistrang 2011, Glenton et al. 2013, Gillard et al. 2014, Simpson et al. 2014). These challenges are not limited to relationships with volunteer and lay workers. Similar experiences have been reported with caseload midwifery with women reporting 'midwife grief' and feeling lost or abandoned at the end of the period of support (Walsh 1999).
It is feasible that such endings may compromise the impact of the period of support. In social work, concerns have been expressed that endings may reinforce previous negative separation experiences (Huntley 2002). In psychotherapy, it is recognised that abrupt endings and forced endings have the potential to be harmful (Gelso & Woodhouse 2002). A recent systematic review of befriending in mental health (Thompson et al. 2015) argued that experiencing some of the qualities of friendship accompanied by an enforced ending could lead to the intervention failing, calling for clearer expectations for support recipients about the nature of what is being offered.
Continuing doula support beyond 6-week postpartum should be considered, especially as this coincides with a time of peak incidence of postnatal depression (Cox et al. 1993). There was some indication that endings may have been easier at the one site where post-natal contact extended until 12 weeks after birth although sample sizes precluded definitive comparisons. Regardless of the length of post-natal support, the ending itself still requires planning and appropriate management, with support from service staff, as required. Several women suggested changing the service to offer an informal meeting to provide an update, groupbased, if necessary. Other evaluations of peer support have recommended using more teamwork, using goals and being problem-focused to minimise dependency in a one-to-one relationship (Perkins & Macfarlane 2001, Repper & Watson 2012; such ways of working may help to enable a transition from the one-to-one relationship but it is unknown how this would influence the impact of support.

Strengths and limitations
This is the largest independent evaluation of trained volunteer doula support in the UK and our findings reflect those of another independent evaluation of one doula service (Granville & Sugarman 2012). Questionnaire data were complemented by interviews, which offered opportunities for more detailed exploration, including the ways in which the volunteer role was similar to and contrasted with support from family, partner and professionals. A strength of our evaluation was the representation of women of non-English-speaking background; however, the questionnaire was only completed by 21.7% of women who had used the service, posing some concerns around sampling bias and transferability of findings. A low response was anticipated because support recipients were in situations of disadvantage with high mobility and in groups traditionally hard to engage in research. In addition, some recipients had accessed the service several years previously and could no longer be contacted. It was not possible to determine from the information provided by the services the extent to which participants were representative in terms of time since using the service and we acknowledge that there is potential impact for memory bias that was not explored here. A higher response rate would be necessary to explore fully the influence of the ending of the relationship on the overall impact of a volunteer doula service.
Efforts to maximise responding included approach via a known service (also essential due to confidentiality) and assisted questionnaire completion. However, any positive impacts from these efforts was possibly limited by language needs being underrecognised by the services, who documented the need for an interpreter, rather than the main language(s) spoken and it appeared that some women may have been sent written information that did not meet their language needs. Unfortunately, fewer data were available for those women using assisted completion because of the need to ensure that the questionnaire length remained acceptable.

Conclusion
The UK NICE guidance for the care of Pregnant women with Complex Social Factors (NICE, 2010) calls for models that overcome barriers and facilitate access to improve women's outcomes. It would appear that volunteer doula services have the potential to make a contribution to this. Of note, the benefits reported by women did not always involve direct support during the labour and birth. An approach akin to friendship and based on building trust, listening and enabling appears to be fundamental; in some circumstances, this can be strengthened by actively supporting involvement of family, including partners. Critically, the ending of the close one-to-one relationship carries the potential for feelings of loss and distress which could undermine the benefits experienced. The timing and management of endings warrant further exploration, particularly given the potential for coinciding with a period of heightened vulnerability for mental health problems. Further longitudinal research is needed to gather women's views and experiences through the period of support, and the ending, to further elucidate the mechanisms by which positive impacts of doula support are achieved and may be threatened.