Patient and public involvement

Midwives and obstetricians in the Stockholm hospitals had been stressing the need for extra labour support for migrant women. A political decision to introduce CBDs, replicating the model designed and implemented by midwives in Gothenburg, was made by Stockholm County Council (Stockholms Läns Landsting, SLL) in 2016. SLL funding was first made available to the community organisation Mira to begin to implement the CBD programme in a Södertälje, a city south of Stockholm. At that time, we approached SSL to discuss the need for robust evaluation and it was agreed to use an RCT design to introduce and evaluate the programme in areas of Stockholm where the programme was not yet available. As the CBD model to be tested replicated the model previously developed, refined and successfully implemented over several years in Gothenburg, further testing the feasibility of the model was not considered necessary. Furthermore, no funding was made available to conduct an initial pilot as evaluation outcomes were required by the funders within a restricted time frame before the eventual roll-out of the model for all migrant women in need of support.

A number of antenatal clinics were approached for their high density of migrant women giving birth according to the Swedish Maternal Health Care Register.26 Five antenatal clinics in Stockholm were chosen initially as study sites for recruiting women to the trial, and a sixth was added after 6 months when the recruitment rate was fairly low. Since recruitment commenced, women outside the study sites may now also contact the research team if they are interested in participating. Simultaneously, the CBD programme was rolled out in the remaining antenatal clinics in Stockholm. The non-profit organisation Mira is responsible for the organisation and management of CBDs, including their training, and the research team is responsible for the evaluation. The SLL finances Mira as well as the research project.

Women speaking the five languages of the participants have been involved in different aspects of the study design as a researcher (AE) or research assistants (NT and others) and have commented on the outcome measures and questionnaires, have taken part in the recruitment procedures and data collection, and continuously informed migrant communities and stakeholders about the study. Women’s and staff’s positive and negative experiences of the intervention are being assessed.

Community-based doulas

Female CBDs are recruited, trained and employed by Mira, using well-established processes.17 Based on experiences from the programme in Gothenburg and Södertälje, a total of 23 CBDs are employed for the five migrant groups included in the trial, ensuring the possibility of back-up when needed. Education is provided by the same midwife overseeing all training and includes eight full days of theoretical and practical training that covers topics; such as anatomy and physiology, strategies for providing effective continuous labour support, comfort measures (breathing and relaxing techniques, providing massage, suggesting positions during labour and birth), obstetric interventions, practical strategies for facilitating communication/interpreting to enhance understanding between women and midwives and the CBDs roles and boundaries (table 2). As part of her training, a CBD to be assists three women during labour and birth before being authorised and assigned to study participants. An experienced midwife provides supervision of CBDs during the practical training. The CBDs are employed on an hourly basis to enable flexibility in timely provision of support to women in labour. The CBDs also have additional employment duties for Mira and provide labour and birth support to women not included in the study; however, they give priority to the study participants.

Participants, inclusion criteria and recruitment

Recruitment to the study commenced on 28 February 2018 and data collection is expected to be finalised by mid-2020. Inclusion criteria are nulliparous or multiparous pregnant Somali, Arabic, Polish, Russian and Tigrinya-speaking women, who are not able to communicate fluently in Swedish, are 18 years or more, in gestational week 25–35, and with no contraindications for vaginal birth. Exclusion criteria are a planned caesarean birth and not giving consent for access to the birth records.

Midwives at six antenatal clinics identify eligible women and inform them briefly about the study, if necessary with the assistance of interpreters, and ask them whether they would be interested to receive further information about what participation would mean. If the woman is interested, the midwife fills in a logbook with inclusion criteria and the woman’s contact details. Thereafter, trained bilingual research assistants collect the logbooks and contact each woman by telephone to provide more detailed study information and collect informed consent for participation. They also conduct a baseline interview, and then open the next envelope in a prenumbered sequence, with the allocation code (see under Randomisation and blinding) and inform the woman about which group she has been allocated to.

The comparison: usual intrapartum care

Regardless of their trial allocation, all women receive usual intrapartum care as provided at their chosen hospital of birth. This includes the support from a midwife and usually from one auxiliary nurse working in the labour ward. The midwife and the auxiliary nurse are responsible for between one and three women in labour. All women are encouraged to bring their partners and/or other support person to accompany them in labour, and language interpreting is offered according to the routines of the hospital. Most interpreting is provided over the phone rather than in person.

The intervention:CBD support

Women allocated to receive CBD support, in addition to usual care, are contacted by a CBD speaking their language, and arrangements are made for them to meet twice prior to the birth to get to know each other and discuss the woman’s wishes regarding support during labour and birth, and what the CBD can offer. Each woman then contacts her CBD when she goes into the first stage of labour, and the CBD meets up with her at the hospital and stays with her throughout labour and birth, in addition to any other support people she may have, such as her partner. The CBD is trained to provide continuous support (emotional, physical and information support) throughout labour and birth, and to facilitate interpretation and mediate communication between the woman and her companion and the caregivers. For instance, to prevent anxiety, she is continuously present by the woman’s side creating a calm atmosphere, providing reassurance and helping the woman to focus on managing contractions one at a time; physical support is provided by means of massage or assisting the woman to find a comfortable position; and information support includes helping the woman to understand the labour process and assist her in communicating with the caregiver. In addition to the CBD intervention, women receive the usual intrapartum care as provided at their chosen hospital of birth. After the birth, the CBD meets with the woman to follow up on any questions or concerns the woman has regarding the birth and postpartum period. Each woman is offered a maximum of 25 hours of CBD time, which includes the two meetings prior to the birth, support during the active phase of labour and birth and a follow-up visit postpartum. If a caesarean is necessary, the CBD accompanies the woman to the operating theatre if possible and supports her throughout the surgery. Women who are in extra need of support during additional consultations, such as regarding female genital mutilation or severe psychological trauma, may be offered additional hours with the CBD.

Randomisation and blinding

Women are randomly allocated to the intervention or control group, using a computerised randomisation schedule. The randomisation ratio is 1:1, CBD support to usual care, with block sizes of 4 or 6 randomly distributed, and with block size chosen to prevent predictability of allocation without jeopardising equal final numbers in trial groups. Allocations are prepared in sealed, opaque envelopes in a central location accessible for the bilingual research assistants. The bilingual researcher and research assistants are responsible for opening the envelopes with the allocation of participants after the baseline questionnaire has been completed by the women. Participants and data collectors cannot be blinded in this study; however, analysis will be undertaken blinded to group allocation.

Primary outcomes

Women’s overall rating of labour care is measured by a single item question taken from the Migrant Friendly Maternity Care Questionnaire (MFMCQ, see under Development of questionnaires)27: In general, were you happy with the healthcare you received? with response alternatives: Yes, very happy; quite happy; not very happy; no, not happy at all. These will be dichotomised as: ‘yes, very happy with care’ and all other alternatives merged into ‘less than very happy’.

Women’s emotional well-being is measured using the EPDS.28 29 The EPDS is a 10-item self-report scale widely used in research and screening for depressive symptoms postpartum. Each item is scored on a 4-point scale (0–3 points for each question) with a total score ranging from 0 to 30. Mean values on the EPDS will be compared between the groups.29 The EPDS is widely used in all the languages included in this study, except for Somali. In Sweden, Somali women are offered screening for depressive symptoms using a Swedish version of the EPDS, which is translated on site by an interpreter. Early in the data collection we experienced that the Somali women found some of the EPDS questions difficult to understand. Therefore, parallel to this RCT, we also conduct think-aloud interviews with a different group of Somali women in order to understand how the EPDS questions are interpreted and how women come up with a response.30 31 The results from the think-aloud interviews will contribute to the interpretation of the results for the Somali subsample.

Secondary outcomes

Overall experience of birth is measured by a single-item question on a five-point scale widely used in research32 33: What was your overall experience of giving birth? with a five-response scale from Very positive to very negative. Overall satisfaction with labour support is measured by Overall, what do you think about all the support you received during childbirth? with a five-response scale from very happy to not so happy. More specific assessments of intrapartum care and support are measured by a selection of relevant questions from the MFMCQ. Other secondary outcomes are: epidural analgesia, length of labour from admission and mode of birth (spontaneous vaginal, instrumental vaginal or caesarean section). Data will be retrieved from hospital patient records to enable assessment of birth outcomes.

Development of questionnaires

The study questionnaires were developed in English, translated into Swedish and the five study languages by bilingual professionals or researchers in the team, and then independently back-translated into each respective language to ensure accuracy of the translation. Survey questions, sources and response alternatives at baseline and follow-up are described in table 3. Some questions were retrieved from the MFMCQ,27 which is a collection of questions created for studies on migrant women’s experience of maternity care by members of the Reproductive Outcome and Migration collaboration. From the MFMCQ, we selected questions on background characteristics and more detailed questions on perceptions of care. Other questions were developed and tested by team members ES and RS for the Hooyo project (Group Antenatal Care for Somali women in Sweden)34 35 and used when appropriate also in this study. The baseline and follow-up questionnaires were read and evaluated by researchers and research assistants, who could speak the relevant languages, to ensure the clarity of questions and piloted with two women. Several questions were slightly modified in response to advice from language speakers or from piloting to avoid any risk of misunderstanding.

Data collection

In order to address some of the known challenges of recruiting and retaining migrant women in clinical trials, trained bilingual research assistants collect data on women’s experiences of care in their language of choice, by telephone or in face-to-face interviews as appropriate, in gestational week 25–36. Every effort is made to make the respondent feel comfortable and at ease. The same research assistant contacts the woman again by telephone at 2 months after the birth and initially asks her how she and her baby are going, before the interview starts. If women are identified with severe prenatal or postnatal physical or emotional problems, the research assistant checks that the woman is receiving appropriate care, and if not, assist her to access care.

The questionnaires are administered by the bilingual research assistants at recruitment (Q1) and 2 months after the birth (Q2). Q1 asks about sociodemographic characteristics (parity, language, knowledge of English/Swedish, family situation, reason for migration, length of residency in Sweden, education, occupation), expectations about the upcoming birth (planned support person, feelings and concerns about the upcoming birth), emotional well-being (EPDS) and health (physical health problems, medications, self-rated health). Q2 asks about self-rated health, emotional well-being, support received during labour and birth, information and cultural aspects of care. Specific questions about doula support were stated to the women in the intervention group only.

High quality prospectively collected information on pregnancy and birth outcomes will be extracted from patient records: obstetric history (parity, previous mode of birth, maternal and neonatal mortality and morbidity), pregnancy (complications, anaemia, chronic diseases and mental health problems), labour and birth (gestational week, induction/spontaneous start of labour, pain relief, mode of birth, maternal morbidity such as haemorrhage or severe injuries) and infant data (Apgar score, need of resuscitation, transferral to neonatal intensive care unit (NICU), birth weighte expansion of ‘NICU, stillbirth or intrapartum death).

Process measures

A process evaluation is being performed continuously, informed by a framework suggested by Moore et al.36 Different aspects of the study context, implementation of the intervention, feasibility, acceptability, fidelity and mechanisms of impact of the intervention as well as of the trial (such as the ability to recruit and retain women in the study) are assessed by means of questionnaires to the women, field notes and logbooks kept by the researchers and research assistants, as well as through qualitative studies. eCBDs, midwives and obstetricians will participate in qualitative interviews with the researchers and bilingual research assistants approximately halfway through and at the end of the trial. Questions to the CBDs will cover positive and negative experiences while working as a doula, experiences of communication and cooperation with women/partners/caregivers, as well as questions about doula education, and their motivations and expectations about becoming a doula. Question areas for the midwives and obstetricians include their own experiences working with CBDs and how they perceive the CBD role during childbirth. Key stakeholder interviews will also be held with labour ward managers and with key members of Mira. Interviews with women and questionnaire data in the intervention arm will explore their experiences of CBD support, such as the length and type of support received, what was most and least helpful and if the doula had helped her to understand what was happening during labour and birth and express her wishes. Questionnaire data from the women as well as monitoring data from the trial, including information about length of doula support for each woman, will also inform the process evaluation. To investigate the experiences of working as a doula, or alongside a doula, a thematic analysis will be conducted according to Braun and Clarke.37 To investigate whether the hypothesised mechanisms of impact are apparent or not in the interviews, a deductive content analysis based on the hypothesised mechanisms of the intervention (table 1) and according to Graneheim and Lundman38 will be performed.

Sample size

To detect an increase in women’s ratings of intrapartum care from an expected 30% saying they are ‘very happy’ among those receiving usual care (based on estimates from studies of migrant women not fluent in the host country language39) to 53% in those receiving CBD support (equal to Swedish-speaking women in a national population based study40) with 80% power and an alpha of 0.05, 69 women in each group are needed. To have similar power to detect differences in mean scores on the EPDS (a hypothesised reduction from a mean of 8.0 in the comparison group—similar to that found in studies of migrant women, to 6.0 in the intervention group—similar to that found in Swedish population-based studies,41 63 women are required in each arm. Allowing for 20% lost to follow-up at the time of data collection with women 2 months post partum, we aim to recruit 174 women.

Statistical analysis

The first step of analysis will be a visual check of the comparability of the two groups. Then, the intervention group will be compared with the control group testing trial hypotheses and using intention to treat analyses that include all participants according to their trial allocation. ORs and 95% CIs will be estimated by logistic regression analysis for categorical outcomes. Comparison of means will be undertaken using t-tests where data are normally distributed, or medians compared using Mann-Whitney U tests used if not. Considerations about imputations for missing data will be made when relevant. If there are visual baseline differences on key participant characteristics (such as age, marital status, medical conditions) between trial arms, Please multivariable logistic regression analyses will be performed to adjust for any differences found.