Clinical and cost-effectiveness of parenting intervention for mothers experiencing psychosocial stress: insights from the early closure of the Mellow Babies RCT

Notes

Permissions

Introduction

This report details the work undertaken to establish the clinical and cost-effectiveness of the Mellow Babies parenting intervention for women experiencing psychosocial stress and their 6- to 18-month-old babies. It arose from a call commissioned by the National Institute for Health and Care Research (NIHR) Public Health Research (PHR) programme ‘15/126 Group-based parenting programmes for improving psychosocial health’. Although the disruption of the COVID-19 pandemic meant the trial was closed before reaching a fully powered sample, there is still significant learning to be gained from the data gathered and the experiences of the research team.

Methods

This was a single-centre RCT comparing outcomes for mothers and babies offered the Mellow Babies parenting intervention plus usual care with mothers and babies receiving usual care only. Participants were randomised 1 : 1 with minimisation to reduce imbalance between groups in terms of maternal age (< 25 years; ≥ 25 years), deprivation (working household yes/no) and age of child (≤ 12 months; > 12 months).

Eligible participants were women aged ≥ 16 years with primary caregiving responsibility of a child aged 6–18 months at the time of randomisation, who resided in Highland Council area (Scotland) and who scored above a defined threshold (≥ 85th centile) on either the anxiety (≥ 11) or depression (≥ 7) subscales of the Hospital Anxiety and Depression Scale (HADS). 23 Women who were substance dependent (if not stable on recovery programme); were unable to speak or write in English language; who had twins or other multiple births; or who had a baby with significant learning difficulties were excluded from the trial. Women could only take part in the trial once and therefore were excluded from participation for subsequent children. Participants were identified via three core recruitment strategies: health professional referral [e.g. health visitor (HV) or general practitioner (GP)]; self-referral having received a letter of invitation via NHS Patient Identification Centre (PIC) sites; or self-referral through response to advertisements on social media or posters within the local community.

For full details of the trial protocol (v20) see https://fundingawards.nihr.ac.uk/award/15/126/05. Screening and baseline data collection were conducted according to the original protocol. Recruitment was paused, and follow-up data collection commenced in a revised form during 2020 and 2021 due to COVID-19 pandemic restrictions. While the original protocol featured home visits, these data collection appointments had to be conducted remotely. Video conferencing [Zoom (Zoom Video Communications, San Jose, CA, USA) or Teams (Microsoft Corporation, Redmond, WA, USA)] was offered to all participants, but most elected for a phone appointment. All measures for the 8-month post-randomisation assessment could be completed by phone, whereas the 30-month measures were amended because the Bayley Scales required direct interaction with and observation of the child. This was replaced by the Ages and Stages Questionnaire, Third Edition (ASQ-3)24 and the Ages and Stages Questionnaires: Social-Emotional, Second Edition (ASQ-SE2). 25 Although we were able to return to home visits in 2022, the ASQ remained in place as the first 30-month follow-ups for the new cohort took place after we were initially asked to close the trial down and it was thought best to not increase participant burden if outcome analyses would not take place. All follow-up data collection ceased at the end of March 2023 when the final close-down plan was approved by Research Ethics Committee (REC) and the funder. Note, although an add-on study regarding Mellow Babies online had been included in the final protocol, delays in research governance processes, and the subsequent early closure of the trial, meant this work did not take place.

The research team was unable to conduct outcome analysis as was originally planned due to the sample size being too small. No between-group clinical or cost-effectiveness analyses were conducted. Data collected from participants at baseline, 8 months post randomisation and when the child reached 30 months of age were analysed descriptively and subgroup comparisons were conducted based on the revised set of questions above. Note that, due to the small sample size and limited value in comparing subgroups from this sample, no correction for repeated testing nor replacement of missing values has taken place. Any descriptive data are presented at face value as a basis for further discussion. This report provides a synopsis of the findings. Full results are presented in a series of appendices grouped around themes among the research questions.

Discussion

This trial was not able to establish the clinical and cost-effectiveness of the Mellow Babies parenting intervention for women experiencing psychosocial stress and their 6- to 18-month-old babies. Due to the close-down of the trial prior to recruiting to the planned sample size, we were unable to conduct any outcome analyses due to insufficient statistical power. There has been some scope for learning, primarily in terms of recruitment and retention strategies, and the rich qualitative data collected through interviews has allowed a thorough exploration of the intervention, including group process.

The recent scoping review by Goyder and colleagues,17 ‘Parenting engagement and support interventions for high risk groups’ provides a useful overview of the current state of the field, and suggests some research questions that could be included in future studies:

1. What are the most effective strategies for identifying and engaging families at increased risk in order to offer parenting interventions?

2. How feasible, acceptable and effective are the assessment tools currently in use to identify those who benefit most from the offer of additional support?

3. What forms of support or content do parents want and need most from parenting programmes; what aspects of current programmes do they value most?

4. What factors make it easier for families to accept or sustain engagement with parenting interventions? What are the reasons that families find it difficult to accept or sustain engagement with parenting interventions?

The present report offers data in relation to each of these questions, although this is limited in the context of an incomplete trial and without having had the time to interview control participants as well as those randomised to the intervention arm.

In terms of recruitment (Q1 above), there is a need to have both direct communication with potential participants (e.g. the PIC letter system) and engagement from practitioners working with families (HVs, GPs and third-sector support staff) to optimise the approach. We are aware that HVs were less engaged with the trial than we had anticipated for a range of reasons, not least the extreme pressure that services were under in terms of reduced resources, reduced staffing capacity and increased needs in the population. We also learnt that the element of randomisation was a significant deterrent to referral by practitioners: they were simply unhappy with suggesting to women in their care that they put themselves forward for a 50 : 50 chance of receiving an intervention they would likely benefit from (in the practitioner’s perception). This aligns with recent findings from Rose et al. 27 that practitioners experience role conflict when asked to recruit to clinical trials, and raises issues around how to manage equipoise when trialling an intervention already popular in practice. One solution to this would be to conduct trials such as this on a regional/service cluster randomisation basis, so that no practitioner is facing this perceived dilemma: the intervention is either available in their area or not. Additionally, feedback from participants indicated that when engaging women, the description around improving bonding with babies may lead to negative feelings about participation, particularly where HVs had made the referral, while reassuring women about how the crèche operates and that babies can come into the group if they do not settle, may reduce anxieties in attending groups. The potentially pivotal role of clinicians/practitioners in recruiting to clinical research studies is clear, and it is important to consider the specific needs/concerns within the context of each given study. 28,29

In relation to the feasibility and acceptability of the assessment tools used in screening for trial eligibility (Q2 above), although we did not gather formal data on this, we received no negative feedback. Prospective participants understood that the trial was aimed at women who were experiencing symptoms of anxiety and/or depression. Some women chose not to go ahead with screening once they had discussed eligibility with the recruiting research nurse and better understood this criterion. There is clearly a balance to be gained where recruitment materials make it clear that the research is aimed at women in psychosocial distress, but does not serve to stigmatise or prevent women from coming forward. Recent work in Highland has shown a strong tendency for women to self-censor their requests for mental health care: there was an expressed sense, especially in a climate of low resource, that women compare themselves to those in more extreme distress and this serves to minimise their own. 30 In the case of The Mellow Babies Trial, recruitment materials were designed with patient and public involvement (PPI) input to ensure language struck an appropriate balance, and the research nurse conducting initial phone calls could then explain more fully the criteria, and use the conversation in the context of her clinical experience to help mothers decide if it was worth going ahead with screening.

In relation to which aspects of the programme were most valued by parents (Q3 above), there was a clear recognition of the need for support, which was often reported to be lacking (or experienced negatively) through established services (e.g. health visiting). The fact that other women within intervention groups were also facing challenges in new parenthood was highlighted as a key strength, with other parent and baby/toddler groups being perceived to be full of parents who were happy and knew what they were doing. The context and ethos of the Mellow Babies group seemed to act to remove the stigma and allow participants to express their challenges and feel a sense of normalisation and validation. Other key positive areas of the intervention highlighted by participants were the sharing of life stories by both participants and facilitators, which reduced barriers within the group, and the provision of the crèche, both of which may be challenging to implement in mainstream services.

We have not been able to fully explore factors that encourage sustained engagement (Q4 above) with the data we have. We did not interview mothers who dropped out of the intervention/research (although simple comments were recorded where possible, such as ‘no time to take part as maternity leave has ended’). Some of the qualitative analysis on group process indicates that group composition, including number of participants, age of children and personal characteristics, were key in the intervention groups becoming cohesive (or not) which was perceived by the participants to have encouraged their sustained participation. As it stands, we had very good retention in the face of significant mental health needs in our sample, possibly due to women having otherwise good support in their lives, as indicated by sociodemographic data. Where women did drop out, it tended to be due to mental well-being concerns or life being too overwhelming and group participation having to be de-prioritised. Outside of a trial setting, and in the context of a sustained intervention programme, there is more likely to be the opportunity for such participants to come back to the intervention (i.e. attend a subsequent group) as strict age and other eligibility criteria would be less likely (e.g. Mellow Parenting tends to be more flexible with its age criterion).

The main challenges to this trial were pragmatic, including difficulty establishing a fruitful recruitment strategy and establishing the infrastructure for intervention delivery in the early months, and the interference of the COVID-19 pandemic once systems were in place and the trial was running successfully. The challenges are outlined in greater detail in Appendix 4 and include a need to redirect our recruitment efforts away from relying solely on HVs and other practitioners towards direct communication with potential participants (PIC letters), difficulties in obtaining critical mass of participants within a reasonable time frame in a low population region, and having to establish an infrastructure for intervention delivery where none previously existed. Nevertheless, the trial was established as viable on two separate occasions (initial start-up and post-pandemic re-start): the final challenge was the need for a longer time period and therefore funds to be able to recruit to power.

There are of course some limitations to consider, which are relevant despite no primary outcome analysis. The sample, although showing significantly poor mental well-being, was otherwise relatively advantaged from a sociodemographic perspective. The applicability of these findings to a wider sociodemographic group is therefore limited. Similarly, this trial focused on mothers for pragmatic reasons (see Publications), so findings cannot be extended to fathers or other adults with parental responsibility. The pragmatic difficulties faced in implementing the intervention raise questions about its viability as a sustained programme within usual services, especially in a region like Highland with a relatively small population size and lack of relevant infrastructure. Specialist practitioner posts are often vulnerable, where there are services essentially offered by lone individuals or very small teams (e.g. infant mental health service). However, this is difficult to assess concretely within the artificial context of a RCT: our findings show, for example, that HVs would have been much more willing to refer mothers had there been the certainty of receiving an intervention. Further qualitative data collection would have been useful to help elucidate the experiences of those who left the intervention early, of those in the control group, and of HVs and other practitioners in the field. Had we been able to interview representatives from each of these groups, our understanding in relation to the secondary questions addressed in this report would have been more comprehensive.

The research team and the intervention team worked extremely hard to engage with key partners and stakeholders in the research to ensure the delivery of intended outcomes. Our connection to NHS Highland and Highland Council was critical: NHS Highland RD&I was the only local organisation able to provide an appropriate infrastructure for recruiting and retaining intervention practitioners, and we relied on the collaboration with the public health department to access the PIC letter system. Health visiting is managed by Highland Council, where key stakeholders were included at the planning stage and as co-investigators in the trial. Their input in overcoming pragmatic barriers was invaluable. Similarly, establishing and maintaining good relationships with these stakeholders allowed a smoother interaction with HVs and other practitioners in participant-facing roles. Although we did connect with colleagues in third-sector services as part of our engagement work, recent Highland-based work, in the post-pandemic, post-Brexit climate, has highlighted how central the role of the third sector is in supporting young families. 30

There are several take-home messages from this research:

1. There is scope to conduct a new trial of Mellow Babies, ideally multisite and including cluster randomisation to facilitate recruitment.

2. It is possible to recruit and retain mothers who are experiencing significant psychosocial stress into trials, provided the infrastructure is realistic and flexibility in approach can be employed.

3. Recruitment needs to be direct, with potential participants being trusted to be their own gatekeepers when putting themselves forward for research. We need to avoid potential paternalism/maternalism in parenting support research.

4. Trials need to be realistically resourced/have a robust infrastructure. Recruitment and group facilitation should ideally be carried out by staff dedicated to these roles – a larger trial would help this.

5. The need for a larger trial with better infrastructure brings the focus back to conducting trials in large urban areas and marginalising those in more remote or rural areas. We need to be more thoughtful about how to ensure all participants meeting clinical eligibility have the chance to participate in an intervention that could realistically and sustainably be delivered to them as part of usual services.

6. Solid engagement with the third sector will be critical for the above.

Patient and public involvement

Both the original Chief Investigator (CI) (PW) and the final CI (LT) had already been investigators on other similar trials [e.g. Trial of Healthy Relationship Initiatives for the Very Early years (THRIVE); Henderson et al. , 201931] which we used to inform our approaches to effective engagement, recruitment, and retention in designing our materials and methodological approach. We recruited a previous THRIVE trial participant to offer PPI input at the start of the trial: she gave feedback on our recruitment posters and methods. We were not able to recruit any further PPI group members before we had to pause study recruitment during the COVID-19 pandemic (March 2020).

We worked closely with the local family nurse partnership (FNP) practitioners who kindly recruited teenage mothers on our behalf in order for the trial manager to meet with them to discuss their thoughts and opinions on various aspects of our recruitment approach. The trial manager met with two teenage mothers identified by FNP in August 2019 to discuss material the team were preparing for Facebook, as well as discussing how best to target their peers for recruitment. The meeting was extremely helpful, and we received insightful comments. They suggested that face-to-face engagement within established community groups would be more effective than social media advertisements, as they did not tend to engage online. They suggested local parent/family groups which the trial manager made contact with/met the facilitator of. There was a need for both virtual and face-to-face approaches. Over time, the research team built a network of third-sector contacts and attended local community groups (run by NHS early years practitioners and third sector) in tandem with social media and other advertisement. Although the families we met in August are not permanent members of the trial’s PPI group, the FNP has assured us they are happy to continue to work with the trial team in order to source young mothers for any future service user meetings.

We successfully recruited four participants from the pre-pandemic period to take part in a discussion about their experiences of participating in the trial and to provide direct feedback on recruitment methods and research procedures in the post-pandemic period. We met with two of them in July 2022 and there were some suggestions for small amendments to our documentation, which could not be implemented as we were asked to stop recruitment soon after. On the whole, the feedback from these participants was positive about their experience as trial participants. The two participants who were unable to attend this session were sent a summary of the discussion via e-mail and given the opportunity to comment (no comments were received).

We plan to generate a brief version of this report to make available (online) to all study participants and any member of the public who might be interested. Although we had planned to have a public-facing dissemination event, there may be little scope for this given the changed research questions and limited resources available. While we are unable to report on the original intended outcome for the trial, there would be merit in discussing the findings in the context of other work happening locally (i.e. the development of perinatal and infant mental health services). LT is working with NHS Highland on the perinatal and infant mental health workstream at present and will liaise with them around effective dissemination.

Equality, diversity and inclusion

Participants: The study was open to any mother meeting the inclusion criteria, including via self-referral. Highland is not an ethnically diverse region, with ˂ 2% of residents not of white European extraction. 32 We considered the main barriers to participation to be stressful lives, mental health, social isolation, literacy/education and living rurally or remotely. There is evidence that parents experiencing these stressors are less likely to engage in parenting interventions. 33–36 Measures to overcome these as much as possible were incorporated into the study design, partly based on the design of the Mellow Babies intervention model. Specifically, home visits were used for obtaining consent and gathering baseline and follow-up data. This meant that all measures, although mostly questionnaires, were conducted in an interview style with the researcher taking time to fit with the participant’s schedule. The participant information sheet was designed to be accessible to those with literacy difficulties, and we sought PPI input into its design. A flexible approach to communication, including the use of whichever means was preferred by the participant and the adoption of a patient approach to reminders (i.e. no ‘three strikes and you’re out’ type rule) allowed us to work around participants’ busy and often stressful lives. Critical to this study, it also allowed mental well-being to be considered: if a participant was having a difficult day and needed to reschedule a visit, we accommodated this. Although living remotely will have been a barrier for some potential participants (i.e. too far to travel to a more populous area with enough critical mass to establish an intervention group), we were careful to target recruitment to areas closer to the urban centres. Where participants did not have access to a car, transport was always provided to allow participation in intervention groups.

We successfully enrolled 106 women with babies 6–18 months old living across the Highland Council region in Scotland. Women enrolled in the trial all scored above the threshold for anxiety and depression, 61% stated they had a diagnosed mental health condition (usually depression or anxiety), and 50% had been prescribed a mental health medication since the birth of their baby. Baseline data showed a high level of education, a high level of paid employment and a high level of home ownership among participants, and most were currently in a relationship with their baby’s father. There were also low levels of stress in terms of life events. This suggests that there is more work to be done to ensure that those experiencing adversity in sociodemographic terms have the opportunity to participate and that barriers to participation need to be better understood. However, while the sample as a whole appeared to be ‘doing well’ on these measures, some participants faced specific challenges. What we do know from those who were screened or recruited is that mothers’ non-attendance when randomised to the intervention arm was linked to life stressors, including moving house, returning to full-time work after maternity leave, not feeling able to commit the time to the group, planned absences during the group, and living too far away from where the group was being delivered. Mothers withdrew from the programme due to life stressors, mental health and child factors. Reasons included experiencing a family bereavement, finding the group made PTSD symptoms worse, and the child not being able to settle in the crèche. Lack of participation in follow-up appointments was also predominantly due to life stressors and participant mental health, including experiencing a recent miscarriage, the birth of a child, time pressures, or not ‘being in a good place’. However, the majority of participants who did not take part in a follow-up did not respond to contact from the research team, so we are unable to determine the reasons for non-participation from these mothers.

Research team: The CI for the majority of the trial was a senior, white, male, medically qualified professor. Once he retired (December 2022), this role was taken over by one of the senior co-investigators, a white female senior research fellow, for the final few months of the trial. The rest of the staff on the trial were female, all of white European ethnicity. Our team included staff from different disciplines (e.g. psychology graduates, medical doctors, nurses) and with a range of experience. The main posts for the running of the trial, the Trial Manager/Research Fellow, Research Assistant, and PhD student, were all young, female, early career researchers. We believe the trial has contributed to developing their careers and capacity within the field. We ensured all staff had the opportunity to contribute to key trial decisions and to train in new areas of skill development. The trial was presented at international conferences by the early career researcher staff members. As well as our PhD student, we had a master’s student (a young female medical student) work on the process evaluation in the first year of the project, and she presented her paper at a conference.

Impact and learning

The main impact of The Mellow Babies Trial has been to demonstrate that complex intervention trials of this nature can be viable in areas lacking a large or diverse population or significant trial infrastructure. It is possible to recruit mothers experiencing psychosocial stress, especially when approaching them directly (i.e. no dependence on a ‘gatekeeper’). It is important to be able to conduct trials away from the main population centres if we are to be confident that our findings are truly applicable to the population more widely. Trials need to be realistically planned, and funded, to be able to allow people from remote and rural areas to participate, or for smaller population centres without the level of trial infrastructure of larger cities to be able to evaluate the effectiveness of interventions in a real-world setting.

Although the participant group was representative of the local population in demographic terms, they showed a relative advantage compared to the target population for the intervention (which is likely to experience a greater disadvantage). 20,21 It is likely more mothers experiencing sociodemographic disadvantage might have been recruited if practitioners working with families had more capacity to recruit to the trial. While the PIC letter has been the most effective means of recruitment, it may be that a formal letter would have been off-putting to some potential participants. A two-pronged approach, involving a letter and direct practitioner contact, would be ideal. It could also have been useful to work more directly with third-sector organisations, who often have more direct contact with families, in recruiting participants. Although we did make contact with as many organisations as possible, in both the pre- and post-pandemic phases of the trial we were just getting to the stage of establishing procedures when we had to stop recruitment.

Regarding trial infrastructure, conducting a trial in Highland meant planning based on tenuous resources. For example, when we submitted the funding application, there was a mobile crèche in operation locally that would have been able to staff the intervention crèche facility. By the time the trial started, this business was no longer in operation and there was no alternative. We worked with local nurseries to develop a preferable alternative (hosting the intervention groups within nurseries where child care could be absorbed), but this could not be reinstated post pandemic partly due to the continued restrictions on adults entering the nursery buildings, but also due to new national legislation increasing early education provision meaning that nurseries no longer had the capacity to host the intervention group or to absorb the child care. Staffing the intervention also presented significant challenges, and ultimately the NHS was the only organisation with the size and infrastructure to accommodate staff in secure posts to allow their sustained involvement in the trial.

The Mellow Babies Trial has been able to demonstrate considerable desire for this type of intervention in the community. Almost 400 mothers expressed interest in the trial. Although only 25% of these ultimately became participants, the artificial barrier created by trial recruitment procedures is almost certainly a significant factor.

We have actively disseminated the learning from this work as much as possible and will continue to do so in the coming months to ensure learning is maximised among key stakeholders. We have presented the protocol and aspects of the development at the following conferences:

• Thompson L, Buchanan L, Christie H, Wilson P. The Mellow Babies Trial: Early Process Evaluation of a Complex Intervention. Oral presentation at the NHS Highland Research, Development, and Innovation Conference, 2019.

• Thompson L, Wilson P, The Mellow Babies Trial team. The Mellow Babies Trial Protocol. Poster presentation at the NHS Highland Research, Development, and Innovation Conference, 2019.

• Tanner J, Wilson P, Thompson L. Group Processes and Interpersonal Change Mechanisms Within a Group-based Intervention for Mothers and Their Infants. Oral presentation at the World Association for Infant Mental Health Congress, 2023.

There are no plans to publish any of the findings contained within this report separately. The interview data have been used in a separate analysis (i.e. different research questions and methods) by Jessica Tanner, a PhD student, as part of her thesis focused on group processes in parenting interventions, successfully examined on 29 April 2024. The paper titles are:

• Tanner J, Wilson P, Wight D, Thompson L. The importance of group factors in the delivery of group-based parenting programmes: a process evaluation of Mellow Babies. Front Child Adolesc Psychiatry 3:1395365. https://doi.org/10.3389/frcha.2024.1395365

• Tanner J, Wilson P, Wight D, Thompson L. The Mellow Babies parenting programme: role of group processes and interpersonal change mechanisms. Front Child Adolesc. Psychiatry 2024;3 https://doi.org/10.3389/frcha.2024.1395363

As stated in Patient and public involvement, we plan to generate a brief version of this report to make available (online) to all study participants and any member of the public who might be interested. Although we had planned to have a public-facing dissemination event, there may be little scope for this given the changed research questions and limited resources available. While we are unable to report on the original intended outcome for the trial, there would be merit in discussing the findings in the context of other work happening locally (i.e. the development of perinatal and infant mental health services). LT is working with NHS Highland on the perinatal and infant mental health workstream at present and will liaise with them around effective dissemination to key stakeholders.

Implications for decision-makers

Due to the close-down of the trial prior to recruiting to the planned sample size, we were unable to conduct any outcome analyses due to insufficient statistical power. This limits our capacity to make recommendations for decision-makers. There is, nevertheless, some scope for learning, primarily in terms of recruitment and retention strategies, and the rich qualitative data collected through interviews have allowed a thorough exploration of the intervention, including group process. As articulated in the Discussion section, there is scope to continue to conduct trials of parenting interventions, including of Mellow Babies, taking into account the learning from this trial regarding logistics and the acceptability and effectiveness of our approach to recruitment and retention. Indeed, research evidence regarding the impact of Mellow Babies continues to accumulate,19–21 which provides useful information for decision-makers, but not at the level of a definitive effectiveness trial. Our take-home messages from the Discussion form the basis of implications for decision-makers:

1. There is scope to conduct a new trial of Mellow Babies, ideally multisite and including cluster randomisation to facilitate recruitment.

2. It is possible to recruit and retain mothers who are experiencing significant psychosocial stress into trials, provided the infrastructure is realistic and flexibility in approach can be employed.

3. Recruitment needs to be direct, with potential participants being trusted to be their own gatekeepers when putting themselves forward for research. We need to avoid potential paternalism/maternalism in parenting support research.

4. Trials need to be realistically resourced/have a robust infrastructure. Recruitment and group facilitation should ideally be carried out by staff dedicated to these roles – a larger trial would help this.

5. The need for a larger trial with better infrastructure brings the focus back to conducting trials in large urban areas and marginalising those in more remote or rural areas. We need to be more thoughtful about how to ensure all participants meeting clinical eligibility have the chance to participate in an intervention that could realistically and sustainably be delivered to them as part of usual services.

6. Solid engagement with the third sector will be critical for the above.

In addition to the above specific implications, there is a need to think more generally about the context and infrastructure around trials of complex interventions such as Mellow Babies. The trial was in many ways constrained by resource availability locally. In order to ensure we reach underserved populations there need to be higher-level actions (infrastructure/funding) that take into account the added contextual barriers (e.g. outsourcing of resources at extra cost), in keeping with NIHR INCLUDE guidance37 and the ongoing NIHR Under-served Communities Programme. 38

Although recruitment was successful in this study, it involved considerably more staff time and logistical effort than anticipated. Despite any efforts made, the trial was not able to reach those in the most remote areas, which in part was due to logistics and resources, but also the constraints related to eligibility criteria (age of child) and needing a critical mass for groups within a short time frame. Rurality has been a barrier to implementation of intervention groups outwith the context of a trial in Highland before, so there is a need to learn directly from those experiences to allow appropriate trial design in future.

Those who did not complete the intervention or were lost to follow-up cited mental health and/or life stressors as the underlying reasons. There is scope to explore more thoroughly how best to retain participants in interventions which are designed to be accessible regardless of these stressors. Outwith the rigid constraints of a conventional RCT design, interventions such as Mellow Babies tend to offer as flexible an approach as possible to allow for continued participation. It may be most prudent to ensure that future trials/interventions exist as part of a wider programme where, for example, those unable to maintain participation in a 14-week whole-day programme may still be able to obtain other relevant support (i.e. an online group) that is more accessible. Having trials more embedded with usual care (such as health visits) without compromising scientific validity would seem to be beneficial not only to trial success but also to participant welfare.

Research recommendations

As this trial was unable to recruit to the planned sample size, there is still ample scope for the original question to be addressed: What is the clinical and cost-effectiveness of the Mellow Babies parenting intervention for women experiencing psychosocial stress and their 6- to 18-month-old babies? This is supported by the review by Goyder and colleagues,17 which highlights the need for higher-level evidence for a number of existing programmes, including Mellow Babies. With thoughtful design, a trial could address not only this primary question, but also several secondary questions along the lines of those proposed by Goyder et al. (see Discussion).

Beyond this specific area of enquiry, our recommendations for future research align with those proposed by Goyder et al. Specifically, the example questions posed in their review present an ideal foundation on which to build on the learning from The Mellow Babies Trial (see Discussion).

• What are the most effective strategies for identifying and engaging families at increased risk in order to offer parenting interventions?

• How feasible, acceptable and effective are the assessment tools currently in use to identify those who benefit most from the offer of additional support?

• What forms of support or content do parents want and need most from parenting programmes; what aspects of current programmes do they value most?

• What factors make it easier for families to accept or sustain engagement with parenting interventions? What are the reasons that families find it difficult to accept or sustain engagement with parenting interventions?

The learning from The Mellow Babies Trial, detailed in previous sections, serves to reinforce the statement made by Goyder et al. , with reference to Hackworth et al. :33 ‘Enabling parents to engage with the offered support requires an understanding of their most immediate support needs and an understanding of the practical, social, economic and cultural barriers that may make it more difficult for parents to accept support or engage with programmes’.

Conclusions

This trial was not able to establish the clinical and cost-effectiveness of the Mellow Babies parenting intervention for women experiencing psychosocial stress and their 6- to 18-month-old babies. Due to the close-down of the trial prior to recruiting to the planned sample size, we were unable to conduct any outcome analyses due to insufficient statistical power. There has been some scope for learning, primarily in terms of recruitment and retention strategies, and the rich qualitative data collected through interviews have allowed a thorough exploration of the intervention, including the group process. Key learning from interviews with intervention participants included the importance of the social cohesion/community the groups offered, having a safe shared space to focus on themselves, as well as the need to be mindful of not only group size but group composition and their potential impact on the intervention experience. Barriers to participation included a sense of stigma, not wanting to deprive those more in need of support, reluctance to use the crèche, and concern about lack of anonymity and speaking out in a small group. Group facilitators need to be carefully selected and trained, as well as provided with ongoing support from each other and professional supervision. Pragmatic issues related to delivering an intervention which is not embedded in services were discussed, including having to work part-time around other commitments and not always being able to be as flexible as the intervention/participants’ needs might require.

There is no doubt that more trials of parenting support interventions are needed, especially where public services are already investing in well-liked programmes lacking a robust evidence base. Any future trial must take on board the recommendations of Goyder et al. ,17 including the need for proper consideration of contextual barriers to participation. This should include consideration of the need to ensure that those in underserved populations, such as those in remote and rural areas, have the chance to meaningfully participate in research.

Additional information

List of abbreviations

ADHD

attention-deficit/hyperactivity disorder

ASQ

Ages and Stages Questionnaire

ASQ-3

Ages and Stages Questionnaire, Third Edition

ASQ-SE2

Ages and Stages Questionnaires: Social-Emotional, Second Edition

CI

chief investigator

EQ-5D

EuroQol-5 Dimensions

EQ-5D-5L

EuroQol-5 Dimensions, five-level version

FNP

family nurse partnership

FU1

follow-up 1

FU2

follow-up 2

GP

general practitioner

HADS

Hospital Anxiety and Depression Scale

HADS-A

Hospital Anxiety and Depression Scale – Anxiety subscale

HADS-D

Hospital Anxiety and Depression Scale – Depression subscale

HV

health visitor

NICE

National Institute for Health and Care Excellence

NIHR

National Institute for Health and Care Research

PCQ

Participant Cost Questionnaire

PHR

Public Health Research

PIC

Patient Identification Centre

PPI

patient and public involvement

PTSD

post-traumatic stress disorder

QALY

quality-adjusted life-year

RCT

randomised controlled trial

RD&I

research, development and innovation

REC

Research Ethics Committee

SDQ

Strengths and Difficulties Questionnaire

SIMD

Scottish Index of Multiple Deprivation

SSLM

Sure Start Language Measure

THRIVE

Trial of Healthy Relationship Initiatives for the Very Early years

VAS

visual analogue scale

References

1. Moffitt TE, Caspi A, Harrington H, Milne BJ. Males on the life-course-persistent and adolescence-limited antisocial pathways: follow-up at age 26 years. Dev Psychopathol 2002;14:179-207.
2. Jokela M, Ferrie J, Kivimӓki M. Childhood problem behaviors and death by midlife: the British National Child Development Study. J Am Acad Child Adolesc Psychiatry 2009;48:19-24.
3. Wilens TE, Biederman J, Mick E, Faraone SV, Spencer T. Attention deficit hyperactivity disorder (ADHD) is associated with early onset substance use disorders. J Nerv Ment Dis 1997;185:475-82.
4. Milberger S, Biederman J, Faraone SV, Chen L, Jones J. ADHD is associated with early initiation of cigarette smoking in children and adolescents. J Am Acad Child Adolesc Psychiatry 1997;36:37-44.
5. Miniscalco C, Nygren G, Hagberg B, Kadesjö B, Gillberg C. Neuropsychiatric and neurodevelopmental outcome of children at age 6 and 7 years who screened positive for language problems at 30 months. Dev Med Child Neurol 2006;48:361-6.
6. Law J, Plunkett C, Taylor J, Gunning M. Developing policy in the provision of parenting programmes: integrating a review of reviews with the perspectives of both parents and professionals. Child Care Health Dev 2009;35:302-12.
7. Eisenberg N, Fabes RA, Shepard SA, Guthrie IK, Murphy BC, Reiser M. Parental reactions to children’s negative emotions: longitudinal relations to quality of children’s social functioning. Child Dev 1999;70:513-34.
8. Tough SC, Siever JE, Leew S, Johnston DW, Benzies K, Clark D. Maternal mental health predicts risk of developmental problems at 3 years of age: follow up of a community based trial. BMC Pregnancy Childbirth 2008;8:1-11.
9. Wilson P, Bradshaw P, Tipping S, Henderson M, Der G, Minnis H. What predicts persistent early conduct problems? Evidence from the Growing Up in Scotland cohort. J Epidemiol Community Health 2013;67:76-80.
10. Scott S, Lewsey J, Thompson L, Wilson P. Early parental physical punishment and emotional and behavioural outcomes in preschool children. Child Care Health Dev 2014;40:337-45. https://doi.org/10.1111/cch.12061.
11. Allely CS, Johnson PC, Marwick H, Lidstone E, Kočovská E, Puckering C, et al. Prediction of 7-year psychopathology from mother-infant joint attention behaviours: a nested case-control study. BMC Pediatr 2013;13.
12. Marwick H, Doolin O, Allely CS, McConnachie A, Johnson P, Puckering C, et al. Predictors of diagnosis of child psychiatric disorder in adult-infant social-communicative interaction at 12 months. Res Dev Disabil 2013;34:562-72.
13. Allely CS, Purves D, McConnachie A, Marwick H, Johnson P, Doolin O, et al. Parent-infant vocalisations at 12 months predict psychopathology at 7 years. Res Dev Disabil 2013;34:985-93.
14. Puckering C, Allely CS, Doolin O, Purves D, McConnachie A, Johnson PC, et al. Association between parent-infant interactions in infancy and disruptive behaviour disorders at age seven: a nested, case-control ALSPAC study. BMC Pediatr 2014;14.
15. Thomson R, Allely CS, Purves D, Puckering C, McConnachie A, Johnson PC, et al. Predictors of positive and negative parenting behaviours: evidence from the ALSPAC cohort. BMC Pediatr 2014;14.
16. Poobalan AS, Aucott LS, Ross L, Smith WC, Helms PJ, Williams JH. Effects of treating postnatal depression on mother-infant interaction and child development: systematic review. Br J Psychiatry 2007;191:378-86.
17. Goyder E. Parenting Engagement and Support Interventions for High Risk Groups: Short Report on Topic Scoping by the Public Health Research Review Team. University of Sheffield/NIHR; 2022.
18. MacBeth A, Law J, McGowan I, Norrie J, Thompson L, Wilson P. Mellow Parenting: systematic review and meta-analysis of an intervention to promote sensitive parenting. Dev Med Child Neurol 2015;57:1119-28. https://doi.org/10.1111/dmcn.12864.
19. Levi D, Ibrahim R, Malcolm R, MacBeth A. Mellow Babies and Mellow Toddlers: effects on maternal mental health of a group-based parenting intervention for at-risk families with young children. J Affect Disord 2019;246:820-7. https://doi.org/10.1016/j.jad.2018.12.120.
20. Raouna A, Malcolm R, Ibrahim R, MacBeth A. Promoting sensitive parenting in ‘at-risk’ mothers and fathers: a UK outcome study of Mellow Babies, a group-based early intervention program for parents and their babies. PLOS ONE 2021;16. https://doi.org/10.1371/journal.pone.0245226.
21. Davidson C, Raouna A, Malcolm R, Ibrahim R, MacBeth A. ‘There’s more love between us’: the parental experience of attending Mellow Babies, a targeted, early intervention program for parents and their babies. Infant Ment Health J 2023;44:100-16. https://doi.org/10.1002/imhj.22029.
22. Orkin AM, Gill PJ, Ghersi D, Campbell L, Sugarman J, Emsley R, et al. CONSERVE Group . Guidelines for reporting trial protocols and completed trials modified due to the COVID-19 pandemic and other extenuating circumstances: the CONSERVE 2021 statement. JAMA 2021;326:257-65. https://doi.org/10.1001/jama.2021.9941.
23. Crawford JR, Henry JD, Crombie C, Taylor EP. Normative data for the HADS from a large non-clinical sample. Br J Clin Psychol 2001;40:429-34. https://doi.org/10.1348/014466501163904.
24. Squires J, Bricker D. Ages and Stages Questionnaires® (ASQ‐3™): A Parent‐Completed Child Monitoring System. Baltimore, MD: Paul H. Brookes; 2009.
25. Squires J, Bricker D, Twombly E. Ages and Stages Questionnaires®: Social-Emotional (ASQ:SE-2™). Baltimore, MD: Paul H. Brookes; 2015.
26. Braun V, Clarke V. Using thematic analysis in psychology. Qual Res Psychol 2006;3:77-101. https://doi.org/10.1191/1478088706qp063oa.
27. Rose J, Lynn K, Akister J, Maxton F, Redsell SA. Community midwives’ and health visitors’ experiences of research recruitment: a qualitative exploration using the Theoretical Domains Framework. Prim Health Care Res Dev 2021;22. https://doi.org/10.1017/S1463423621000050.
28. Preston NJ, Farquhar MC, Walshe CE, Stevinson C, Ewing G, Calman LA, et al. Strategies designed to help healthcare professionals to recruit participants to research studies. Cochrane Database Syst Rev 2016;2. https://doi.org/10.1002/14651858.MR000036.pub2.
29. Rendell JM, Merritt RD, Geddes JR. Incentives and disincentives to participation by clinicians in randomised controlled trials. Cochrane Database Syst Rev 2007;2007. https://doi.org/10.1002/14651858.MR000021.pub3.
30. Thompson L. NHS Highland Perinatal and Infant Mental Health Workstream Evaluation. Dublin: World Association of Infant Mental Health Congress; 2023.
31. Henderson M, Wittkowski A, McIntosh E, McConnachie A, Buston K, Wilson P, et al. THRIVE Trial Research Team . Trial of healthy relationship initiatives for the very early years (THRIVE), evaluating Enhanced Triple P for Baby and Mellow Bumps additional social and care needs during pregnancy and their infants who are at higher risk of maltreatment: study protocol for a randomised controlled trial. Trials 2019;20. https://doi.org/10.1186/s13063-019-3571-5.
32. Scottish Government 2011. www.scotlandscensus.gov.uk/search-the-census#/explore/snapshot (accessed 29 March 2023).
33. Hackworth NJ, Matthews J, Westrupp EM, Nguyen C, Phan T, Scicluna A, et al. What influences parental engagement in early intervention? Parent, program and community predictors of enrolment, retention and involvement. Prevent Sci 2018;19:880-93. https://doi.org/10.1007/s11121-018-0897-2.
34. Hine RH, Maybery D, Goodyear MJ. Challenges of connectedness in personal recovery for rural mothers with mental illness. Int J Ment Health Nurs 2018;27:672-82. https://doi.org/10.1111/inm.12353.
35. Owens JS, Richerson L, Murphy CE, Jagelewski A, Rossi L. The parent perspective: informing the cultural sensitivity of parenting programs in rural communities. Child Youth Care Forum 2007;36:179-94. https://doi.org/10.1007/s10566-007-9041-3.
36. Smokowski P, Corona R, Bacallao M, Fortson BL, Marshall KJ, Yaros A. Addressing barriers to recruitment and retention in the implementation of parenting programs: lessons learned for effective program delivery in rural and urban areas. J Child Fam Stud 2018;27:2925-42. https://doi.org/10.1007/s10826-018-1139-8.
37. National Institute for Health and Care Research . Improving Inclusion of Under-Served Groups in Clinical Research: Guidance from INCLUDE Project 2020. www.nihr.ac.uk/documents/improving-inclusion-of-under-served-groups-in-clinical-research-guidance-from-include-project/25435 (accessed 30 August 2023).
38. National Institute for Health and Care Research . Under-Served Communities 2022. www.nihr.ac.uk/about-us/our-key-priorities/under-served-communities.htm (accessed 30 August 2023).
39. Scottish Government . Scotland’s Census: Education 2021. www.scotlandscensus.gov.uk/census-results/at-a-glance/education/ (accessed 28 August 2023).
40. Brugha T, Bebbington P, Tennant C, Hurry J. The list of threatening experiences: a subset of 12 life event categories with considerable long-term contextual threat. Psychol Med 1985;15:189-94. https://doi.org/10.1017/s003329170002105x.
41. Youthinmind . Scoring the Strengths and Difficulties Questionnaire for 2–4 Year Olds 2015. www.sdqinfo.org/py/sdqinfo/c0.py (accessed 4 July 2023).
42. Van Hout B, Janssen M, Feng Y, Kohlmann T, Busschbach J, Golicki D, et al. Interim scoring for the EQ 5D 5L: mapping the EQ 5D 5L to EQ 5D 3L value sets. Value Health 2012;15:708-15.
43. Hunter RM, Baio G, Butt T, Morris S, Round J, Freemantle N. An educational review of the statistical issues in analysing utility data for cost-utility analysis. PharmacoEconomics 2015;33:355-66.

Appendix 1 Trial recruitment, retention and participant characteristics: full results

This section will include data to address the following questions from the close-down protocol (https://njl-admin.nihr.ac.uk/document/download/2042261; accessed 20 November 2024):

1. Which eligible mothers agree and which decline to participate in the intervention, and what reasons do they give?

2. Which recruitment methods were most effective and was there was any variation in participant characteristics with means of recruitment?

3. ======

4. How do sociodemographic characteristics and maternal mental health at baseline relate to:

   1. changes in parenting behaviours?

   2. maternal mental state at 8 months post recruitment?

   3. child development at 30 months, using ASQ rather than Bayley?

5. ======

6. Are there family characteristics associated with greater adherence to Mellow Babies?

7. Are there family characteristics associated with greater retention to follow-up?

Recruitment

Figure 1 (Consolidated Standards of Reporting Trials diagram) gives details of numbers of participants expressing interest/screened/eligible/recruited to the trial. Of 390 expressions of interest, 179 consented to take part in screening, and 139 of these were eligible for participation. The most common reason for ineligibility was not meeting the HADS threshold criterion (n = 36).

FIGURE 1.

Consolidated Standards of Reporting Trials diagram of recruitment to The Mellow Babies Trial.

Of those screening negative on the HADS, the median scores were 3 [interquartile range (IQR) 2–5] for Depression and 6 (IQR 3–8) for Anxiety. Scores within this group were fairly evenly spread, with five participants (14%) scoring just below the threshold for Depression and three participants (9%) just below the threshold for Anxiety (Figure 2).

FIGURE 2.

Boxplot of HADS Anxiety and Depression scores among those not meeting screening criteria.

Of those who screened positive, the most effective recruitment strategy by far was PIC letters, resulting in 75 (57%) of eligible participants. While we had anticipated HV referral might be our most successful method, only 26 (20%) of those eligible came from this source. Anecdotally, we understand that HVs were reluctant to refer to a trial where participants had a 50 : 50 chance of receiving an intervention. There were only two referrals from GPs. All other referrals came from social media (n = 19; 14%), the study website (n = 1; 1%) or other forms of self-referral (n = 9; 7%). (Referral source data were not available for 7/139 eligible.) Table 2 shows the proportion eligible per recruitment source and associated median HADS scores. Collapsing the categories and comparing the ‘professional referral’ method (HV or GP) to the others (all ‘self-referral’) made no impact on whether screened as eligible: 90% of ineligible participants were self-referrals versus 79% of those eligible.

TABLE 2 - Eligibility and HADS scores by recruitment source

Frequencies were too low to allow IQR on most categories.

	Eligibility criteria met?		HADS anxiety	HADS depression
Recruitment source	Y, n (%)	N, n (%)	Median (rangea)	Median (rangea)
Study website	1 (33)	2 (67)	9 (5–15)	3 (3–7)
Facebook/Instagram	19 (91)	2 (9)	12.3 (6–19)	9 (1–15)
Baby group	1 (100)	0 (0)	11	5
Other self-referral	8 (100)	0 (0)	13 (12–20)	10 (3–16)
PIC letter	75 (70)	32 (30)	11 (0–20)	7 (0–19)
HV referral	27 (87)	4 (13)	12 (2–18)	9 (1–18)
GP referral	2 (100)	0 (0)	16 (16–16)	8.5 (7–10)

Of those who screened positive but ultimately were not recruited to the trial, the most common reasons were changing their mind (n = 13), no longer being able to commit the time (often due to returning to work after maternity leave) (n = 7), or due to trial recruitment being suspended in March 2020 (n = 7). Only two were uncontactable at this stage, one had left the area, and one chose not to go ahead to baseline visit as there was not going to be an intervention group in her area in the timescale needed. Finally, two participants who had been screened late in 2022 and remained keen to participate ultimately could not due to trial closure.

Demographic characteristics at baseline

Full baseline characteristics for the final sample of 106 are provided in Appendix 5. Key highlights include little ethnic diversity (representative of the region), a high level of education [50 (47%) educated to university level; Scottish national average 26%39], a high level of paid employment (or on maternity leave and planning to return to paid employment) (n = 68, 64%) and a high level of home ownership (n = 55, 52%). Participants ranged from 20 to 46 years of age {mean [standard deviation (SD)] = 32 (5.9)}, and their babies ranged from 5 to 17 months [mean (SD) = 9 (3.7)] at point of screening. About two-thirds of the sample were from relatively affluent areas (i.e. Scottish Index of Multiple Deprivation (SIMD) quintile 3–5), and only 14 (13%) lived in areas classed as remote or very remote rural. See Appendix 5, Table 21 for a full breakdown.

Mental health at baseline

All participants scored above the 85th centile23 on one HADS subscale at screening, and mean scores were 11.7 (SD 4.2) for Hospital Anxiety and Depression Scale – Anxiety subscale (HADS-A) and 7.8 (SD 4.0) for Hospital Anxiety and Depression Scale – Depression subscale (HADS-D). Although 61% stated that they had been diagnosed with a mental illness, much higher proportions (86%/76%/96%) reported experiencing mental ill health before/during/or after pregnancy, respectively. Rates of prescribed medication for mental ill health were 64%/23%/50% before/during/after pregnancy, respectively. The lower rate during pregnancy will likely be due to mothers choosing to cease medication so as not to harm their unborn baby. It is interesting to note that prescription rates did not increase in line with the increased experience of mental ill health postnatally. In a separate question focused on pregnancy, 84% of participants stated they had experienced anxiety or depression during this period (higher than the diagnosed mental illness rate of 61%). See Appendix 5, Table 22 for a full breakdown of responses.

Stressful lives at baseline

Using the Recent Life Events scale,40 with two items added pertaining to child protection (Has a child living in your household been placed on the child protection register or been taken into care?) and homelessness (Have you been homeless?), this sample endorsed relatively few items: the median number of events endorsed was three (IQR = 1–5). Of those who had endorsed life events, most (n = 62, 81%) stated that at least one of these still affected them (median = 2, IQR = 1–4). Regarding the added items, seven participants had been made homeless, with five stating it still affected them; three had experienced a child protection situation within their household, with one stating this still affected them. See Appendix 5, Table 22 for a full breakdown of responses.

Social isolation at baseline

Most participants (84%) were in a relationship at baseline, with all but one stating their partner was the father of their baby. Participants reported good practical and emotional support from partners, although 21 (23%) felt they were less close with their partner since having their baby, and 14 (16%) reporting recent tension, irritability, etc. in their relationship. Reported relationship quality was generally very good, with 81 (91%) stating they enjoyed spending time with their partner. Of some concern was the small number (7) who reported that conflict with their partner had often ‘come close to coming to blows, or worse’, although physical violence was not mentioned in free text comments. With regard to support or advice, family or partner (usually baby’s father) were the most popular sources, with HVs and GPs both being endorsed as an accessible professional source for most. There were very few negative responses to four questions about availability of practical support in the preceding 6 months, with all participants being able to endorse at least one of the four items. See Appendix 5, Table 23 for a full break down of responses.

Child development outcomes at follow-up 2 (child reaches 30 months of age) in relation to sociodemographic characteristics and mental health at baseline

Tables 8 and 9 show outcomes on the main developmental measures – the ASQ-3, ASQ-SE2 and Strengths and Difficulties Questionnaire – taken at FU2 (when the child reached 30 months). There were no statistically significant differences between the subgroups on any of these measures, most likely due to the small sample size involved. Of note is that all participants scored above threshold on the ASQ-SE2, indicating that further assessment with a professional may be needed. This is surprising given how few scored above threshold on the ASQ-3, including on the personal-social subscale, and may indicate a specific domain of challenges being faced by this small group.

TABLE 8 - Ages and Stages Questionnaire 3 and ASQ-SE2 at FU2 in relation to sociodemographic characteristics and mental health at baseline

ASQ-C, communication; ASQ-GM, gross motor; ASQ-FM, fine motor; ASQ-PrSo, problem solving; ASQ-PeSo, personal social.

> 85 for 30-month version, > 105 for 36-month version, as relevant.

	ASQ-C				ASQ-GM				ASQ-FM				ASQ-PrSo				ASQ-PeSo				ASQ-SE2
	Mean score (SD)		> 22.77, n (%)		Mean score (SD)		> 41.84, n (%)		Mean score (SD)		> 30.16, n (%)		Mean score (SD)		> 24.62, n (%)		Mean score (SD)		> 33.71, n (%)		Mean score (SD)		‘Refer’a, n (%)
Whole sample	52.6	(11.9)	1	(3)	49.6	(8.9)	7	(21)	44.9	(14.8)	6	(20)	46.8	(11.6)	1	(3)	50.0	(9.0)	1	(3)	131.7	(24.9)	33	(100)
Study arm
Intervention	51.2	(14.4)	1	(6)	49.7	(8.9)	4	(24)	46.5	(14.2)	3	(20)	46.7	(14.4)	1	(7)	50.9	(10.3)	1	(6)	132.1	(22.2)	17	(100)
Usual care	54.1	(8.6)	0	(0)	49.4	(9.1)	3	(19)	43.3	(15.8)	3	(20)	46.9	(8.8)	0	(0)	49.1	(7.5)	0	(0)	131.3	(28.3)	16	(100)
Mother’s age band
< 25 years	42.0	(23.1)	1	(20)	51.0	(8.2)	0	(0)	44.8	(15.7)	1	(20)	37.4	(15.8)	1	(20)	47.2	(8.8)	0	(0)	147.0	(28.2)	5	(100)
≥ 25 years	54.5	(8.0)	0	(0)	49.3	(9.1)	7	(25)	45.0	(15.0)	5	(20)	48.6	(10.1)	0	(0)	50.5	(9.1)	1	(4)	128.9	(23.8)	28	(100)
Child’s age band
≤ 365 days	52.4	(14.0)	1	(5)	50.2	(9.3)	5	(24)	45.4	(15.1)	4	(22)	47.9	(12.7)	1	(5)	49.8	(8.9)	1	(5)	128.1	(25.9)	21	(100)
> 365 days	52.9	(7.2)	0	(0)	48.3	(8.3)	2	(17)	44.3	(15.0)	2	(17)	44.9	(10.0)	0	(0)	50.5	(9.5)	0	(0)	137.9	(22.9)	12	(100)
SIMD
1 (most deprived)	53.0	(10.4)	0	(0)	44.0	(2.2)	1	(20)	41.0	(22.3)	1	(25)	47.5	(6.5)	0	(0)	49.0	(5.5)	0	(0)	139.0	(19.5)	5	(100)
2	56.7	(4.1)	0	(0)	53.3	(5.2)	0	(0)	45.5	(12.3)	1	(17)	47.0	(11.9)	0	(0)	48.5	(10.1)	0	(0)	137.5	(32.7)	6	(100)
3	55.0	(7.1)	0	(0)	53.8	(12.5)	1	(25)	51.7	(7.6)	0	(0)	53.8	(4.8)	0	(0)	50.0	(10.8)	0	(0)	132.5	(11.9)	4	(100)
4	48.2	(15.9)	1	(7)	48.9	(9.0)	4	(29)	44.5	(15.2)	3	(23)	44.4	(15.1)	1	(8)	50.0	(10.6)	1	(7)	133.9	(26.4)	14	(100)
5 (least deprived)	58.3	(2.9)	0	(0)	55.0	(5.0)	0	(0)	53.3	(7.6)	0	(0)	45.3	(8.1)	0	(0)	51.7	(2.9)	0	(0)	108.3	(2.9)	3	(100)
Rurality
More urban	51.9	(13.5)	1	(4)	50.2	(8.4)	4	(17)	46.7	(14.2)	3	(14)	44.8	(12.2)	1	(5)	49.2	(9.2)	1	(4)	134.6	(27.0)	24	(100)
More rural	54.4	(5.8)	0	(0)	47.8	(10.3)	3	(33)	40.9	(16.3)	3	(33)	51.7	(9.0)	0	(0)	52.2	(8.3)	0	(0)	123.9	(17.3)	9	(100)
Education
Degree level	54.7	(7.2)	0	(0)	46.3	(8.1)	5	(33)	45.4	(15.4)	3	(21)	47.0	(10.5)	0	(0)	51.3	(6.4)	0	(0)	130.3	(24.4)	15	(100)
Below degree level	50.8	(14.7)	1	(6)	52.2	(8.8)	2	(11)	44.5	(14.8)	3	(19)	46.6	(12.8)	1	(6)	48.9	(10.7)	1	(4)	132.8	(26.0)	18	(100)
Employment
Currently employed/mat leave	52.8	(12.8)	1	(5)	49.3	(9.1)	5	(25)	46.1	(13.7)	3	(18)	48.1	(13.6)	1	(6)	48.8	(9.2)	1	(5)	131.5	(24.9)	20	(100)
Not employed	52.3	(10.7)	0	(0)	50.0	(8.9)	2	(15)	43.4	(16.6)	3	(23)	44.9	(8.4)	0	(0)	52.0	(8.6)	0	(0)	131.9	(25.9)	13	(100)
Home ownership
Yes	55.3	(6.7)	0	(0)	48.3	(8.7)	5	(28)	47.8	(12.4)	2	(12)	48.4	(10.1)	0	(0)	51.1	(6.1)	0	(0)	127.8	(23.9)	18	(100)
No	49.3	(15.7)	1	(7)	51.0	(9.1)	2	(13)	41.2	(17.3)	4	(31)	44.8	(13.3)	1	(7)	48.7	(11.6)	1	(7)	136.3	(26.2)	15	(100)
Current relationship
Yes	52.3	(12.2)	1	(4)	49.1	(8.9)	6	(21)	44.0	(15.0)	6	(22)	45.9	(11.8)	1	(4)	51.4	(6.7)	0	(0)	129.5	(24.8)	28	(100)
No	54.0	(10.8)	0	(0)	52.0	(9.1)	1	(20)	53.3	(11.6)	0	(0)	53.0	(8.7)	0	(0)	42.2	(15.9)	1	(20)	144.0	(24.3)	5	(100)
Help in past 6 months
4 items	54.1	(8.9)	0	(0)	49.8	(9.3)	7	(24)	46.3	(12.9)	4	(15)	48.0	(9.7)	0	(0)	50.9	(8.9)	1	(3)	130.0	(25.0)	29	(100)
5–7 items	41.3	(23.9)	1	(25)	47.5	(5.0)	0	(0)	36.0	(24.7)	2	(50)	38.8	(21.0)	1	(25)	43.8	(7.5)	0	(0)	143.8	(24.3)	4	(100)
Diagnosed mental illness (baseline)
Yes	54.3	(12.2)	1	(5)	48.5	(9.6)	5	(25)	47.2	(12.0)	2	(11)	47.0	(12.9)	1	(6)	49.5	(10.1)	1	(5)	127.8	(24.4)	20	(100)
No	50.0	(11.4)	0	(0)	51.2	(7.7)	2	(15)	41.5	(18.3)	4	(33)	46.5	(10.1)	0	(0)	50.9	(7.1)	0	(0)	137.7	(25.5)	13	(100)
Anxiety or depression (baseline)
Yes	53.2	(12.3)	1	(4)	49.6	(8.8)	5	(18)	44.5	(14.8)	5	(20)	47.3	(11.9)	1	(4)	50.2	(9.1)	1	(4)	129.1	(24.0)	28	(100)
No	49.0	(9.6)	0	(0)	49.0	(10.2)	2	(40)	47.0	(16.4)	1	(20)	44.0	(10.8)	0	(0)	49.0	(8.9)	0	(0)	146.0	(27.7)	5	(100)

TABLE 9 - Strength and Difficulties Questionnaire at FU2 in relation to sociodemographic characteristics and mental health at baseline

SDQ, Strengths and Difficulties Questionnaire.

*p < 0.05. **p < 0.001.

	SDQ-emotion				SDQ-conduct				SDQ-hyperactivity				SDQ-peer relationships				SDQ-prosocial				SDQ-total difficulties
	Mean score (SD)		> 2, n (%)		Mean score (SD)		> 3, n (%)		Mean score (SD)		> 5, n (%)		Mean score (SD)		> 2, n (%)		Mean score (SD)		< 7, n (%)		Mean score (SD)		> 12, n (%)
Whole sample	1.69	(1.5)	6	(17)	2.8	(1.6)	12	(34)	4.3	(2.6)	10	(29)	2.2	(1.8)	14	(40)	7.3	(1.8)	11	(31)	11.0	(5.3)	14	(40)
Study arm
Intervention	1.8	(1.6)	3	(17)	2.7	(2.0)	7	(39)	4.1	(2.9)	5	(28)	2.2	(1.5)	8	(44)	6.9	(1.9)	7	(39)	10.8	(6.1)	8	(44)
Usual care	1.6	(1.4)	3	(18)	2.9	(1.2)	5	(29)	4.5	(2.2)	5	(29)	2.2	(2.0)	6	(35)	7.8	(1.7)	4	(24)	11.2	(4.5)	6	(35)
Mother’s age band
< 25 years	1.8	(0.8)	1	(20)	3.6	(1.8)	2	(40)	6.6	(2.7)	4	(80)	3.2	(1.9)	4	(80)	6.4	(2.6)	3	(60)	15.2	(5.8)	4	(80)
≥ 25 years	1.7	(1.6)	5	(17)	2.7	(1.6)	10	(33)	3.9	(2.4)*	6	(20)*	2.0	(1.7)	10	(33)	7.5	(1.7)	8	(27)	10.3	(4.9)	10	(33)
Child’s age band
≤ 365 days	1.8	(1.8)	5	(22)	2.7	(1.8)	9	(39)	4.9	(2.5)	8	(35)	2.2	(1.8)	10	(44)	7.2	(2.0)	8	(35)	11.7	(5.6)	10	(44)
> 365 days	1.4	(0.9)	1	(8)	2.9	(1.2)	3	(25)	3.2	(2.3)*	2	(17)	2.1	(1.7)	4	(33)	7.6	(1.4)	3	(25)	9.6	(4.6)	4	(33)
SIMD
1 (most deprived)	2.3	(2.4)	3	(43)	2.6	(1.3)	1	(14)	6.0	(1.9)	4	(57)	2.4	(1.6)	3	(43)	8.0	(1.7)	1	(14)	13.3	(4.6)	4	(57)
2	1.2	(0.8)	0	(0)	2.7	(1.8)	3	(50)	3.3	(1.6)	0	(0)	1.8	(2.2)	2	(33)	8.0	(1.8)	1	(17)	9.0	(5.1)	2	(33)
3	1.0	(0)	0	(0)	2.0	(1.0)	0	(0)	4.2	(2.5)	1	(20)	1.6	(1.1)	1	(20)	7.6	(1.5)	1	(20)	8.8	(2.6)	1	(20)
4	1.6	(1.0)	2	(14)	3.1	(2.0)	6	(43)	4.2	(3.0)	5	(36)	2.5	(1.9)	7	(50)	6.7	(2.0)	6	(43)	11.4	(6.2)	6	(43)
5 (least deprived)	1.5	(0.7)	0	(0)	3.5	(0.7)	1	(50)	2.5	(3.5)	0	(0)	0.5	(0.7)	0	(0)	6.0	(0)	2	(100)	8.0	(4.3)	0	(0)
Rurality
More urban	1.6	(1.5)	5	(19)	3.0	(1.6)	10	(39)	4.6	(2.7)	9	(35)	2.0	(1.8)	10	(39)	7.4	(1.8)	9	(35)	11.2	(5.4)	11	(42)
More rural	1.9	(1.7)	1	(11)	2.1	(1.5)	2	(22)	3.6	(2.2)	1	(11)	2.7	(1.7)	4	(44)	7.3	(2.0)	2	(22)	10.2	(5.1)	3	(33)
Education
Degree level	2.2	(1.9)	3	(20)	2.8	(1.6)	6	(40)	4.5	(2.6)	4	(27)	2.5	(1.6)	6	(40)	7.3	(1.5)	4	(27)	12.1	(4.8)	6	(40)
Below degree level	1.3	(1.0)	3	(15)	2.8	(1.7)	6	(30)	4.2	(2.6)	6	(30)	1.9	(1.9)	8	(40)	7.4	(2.1)	7	(35)	10.2	(5.6)	8	(40)
Employment
Currently employed/mat leave	1.6	(1.6)	4	(18)	2.6	(1.9)	8	(36)	4.4	(2.8)	7	(32)	1.9	(1.6)	7	(32)	7.2	(2.1)	8	(36)	10.6	(5.7)	8	(36)
Not employed	1.9	(1.4)	2	(15)	3.1	(1.0)	4	(31)	4.2	(2.2)	3	(23)	2.5	(2.1)	7	(54)	7.5	(1.3)	3	(23)	11.7	(4.7)	6	(46)
Home ownership
Yes	1.7	(1.6)	3	(16)	2.3	(1.6)	5	(26)	3.7	(2.7)	4	(21)	2.1	(1.6)	6	(32)	7.2	(1.8)	6	(32)	9.8	(5.2)	5	(26)
No	1.7	(1.5)	3	(19)	3.4	(1.4)*	7	(44)	5.0	(2.3)	6	(38)	2.3	(2.0)	8	(50)	7.5	(1.9)	5	(31)	12.4	(5.2)	9	(56)
Current relationship
Yes	1.8	(1.5)	5	(17)	2.7	(1.6)	10	(35)	4.3	(2.7)	9	(31)	2.5	(1.7)	14	(48)	7.3	(1.9)	9	(31)	11.2	(5.4)	12	(41)
No	1.3	(1.9)	1	(17)	3.5	(1.8)	2	(33)	4.2	(1.7)	1	(17)	0.7	(0.8)**	0	(0)	7.3	(1.6)	2	(33)	9.7	(4.9)	2	(33)
Help in past 6 months
4 items	1.7	(1.5)	4	(13)	2.8	(1.6)	11	(37)	4.0	(2.6)	7	(23)	2.1	(1.7)	11	(37)	7.2	(1.6)	10	(33)	10.6	(5.0)	11	(37)
5–7 items	1.4	(1.5)	2	(40)	2.8	(1.9)	1	(20)	6.0	(2.2)	3	(60)	2.8	(2.2)	3	(60)	8.0	(2.8)	1	(20)	13.0	(7.1)	3	(60)
Diagnosed mental illness (baseline)
Yes	1.8	(1.8)	5	(23)	2.8	(1.8)	9	(41)	4.6	(2.6)	7	(32)	1.9	(1.5)	7	(32)	7.5	(2.1)	7	(32)	11.1	(5.4)	9	(41)
No	1.5	(0.8)	1	(8)	2.9	(1.3)	3	(23)	3.8	(2.6)	3	(23)	2.6	(2.2)	7	(54)	7.1	(1.3)	4	(31)	10.8	(5.2)	5	(39)
Anxiety or depression (baseline)
Yes	1.8	(1.6)	6	(20)	2.8	(1.7)	11	(37)	4.2	(2.6)	8	(27)	2.1	(1.6)	12	(40)	7.4	(1.8)	9	(30)	10.8	(5.4)	12	(40)
No	1.0	(0.7)	0	(0)	3.0	(0.7)	1	(20)	5.0	(2.8)	2	(40)	2.8	(2.6)	2	(40)	7.2	(1.9)	2	(40)	11.8	(4.8)	2	(40)

Strengths and Difficulties Questionnaire subscale and total mean scores were relatively low, with 40% scoring above threshold (i.e. above the ‘close to average’ category41). There were few significant differences by the various subgroups: children of mothers aged 25 or older at baseline had lower scores on hyperactivity/inattention, and were less likely to score above threshold. Significantly lower hyperactivity/inattention scores were also observed where children were at least a year old at baseline, but they were not less likely to score above threshold. Mothers in a relationship at baseline had children with significantly higher scores on peer relationship problems at FU2. As with the previous section, the lack of significant subgroup differences is most likely due to the small sample size for this data collection point.

Finally, Sure Start Language Measure (SSLM) data are shown in Table 10 and Figure 3. Note that these were based on a smaller range of words than usual: the first 10 words were omitted in error from the data capturing system for those pre COVID; these were captured only for a small subset of participants (those recruited post COVID), so they were omitted from analysis: the highest possible total was 40 words. Not surprisingly, the number of words being spoken by children was much lower at FU1, when children were younger, than at FU2, when all were around 2.5 years old. There were no significant associations between SSLM scores and any of the sociodemographic characteristics or mental health measures featured in the other subgroup descriptive analysis, with the exception of maternal education, suggesting a trend towards those children with mothers having education at degree level or above speaking more words at FU2 [mean (SD) < degree = 28 (13.6); degree + = 34.8 (8.1) (t = 31.8; p = 0.04)]. Most children at FU2 were producing two-word utterances.

TABLE 10 - Sure Start Language Measure descriptive data at FU1 and FU2

	Number of words		Two-word utterances
	Mean (SD)		n (%)
FU1	9.3	(10.2)	Often	16	(24)
			Sometimes	15	(22)
			Not at all	37	(54)
			Missing	2
FU2	31.0	(11.9)	Often	32	(89)
			Sometimes	3	(8)
			Not at all	1	(3)
			Missing	0

FIGURE 3.

Boxplot showing distribution of SSLM scores at FU1 and FU2.

Appendix 2 Characterising use of other services during trial participation

This section will include data to address the following questions from the close-down protocol (https://njl-admin.nihr.ac.uk/document/download/2042261; accessed 20 November 2024):

4. What is the nature of usual care offered to participants?

It will also describe baseline and follow-up sample characteristics in terms of quality of life and service use.

Appendix 3 Intervention participants: characteristics, cohesion and process: full results

This section will include data to address the following questions from the close-down protocol (https://njl-admin.nihr.ac.uk/document/download/2042261; accessed 20 November 2024):

1. 3. How do sociodemographic characteristics and maternal mental health at baseline relate to:

   1. level of participation in Mellow Babies?

   2. group composition?

2. =====

3. 5. How do participants describe their experience of participating in Mellow Babies, which elements of the intervention are considered most influential, and is participation stigmatising?

4. =====

5. 8. How are the features (in terms of process and outcomes of care) of Mellow Babies valued by mothers?

6. 9. What contextual factors facilitate or hinder delivery of, and engagement with, Mellow Babies?

Findings from process evaluation interviews

Process evaluation interviews explored the facilitators and barriers for recruitment, delivery and participation, along with participants’ perceptions of the benefits of the programme, and the influential components. Four themes and 30 subthemes were identified, depicted in Figure 4.

FIGURE 4.

Themes arising from process evaluation interviews with intervention participants and group facilitators.

Appendix 4 The impact of the COVID-19 pandemic on The Mellow Babies Trial: full results

This section will include data to address the final question from the close-down protocol (https://njl-admin.nihr.ac.uk/document/download/2042261; accessed 20 November 2024):

1. 4. How do sociodemographic characteristics and maternal mental health at baseline relate to:

   1. whether recruited pre-pandemic (to March 2020) or post (since November 2021)?

2. =====

3. 10. What specific impact did the COVID-19 pandemic have on the ability to run intervention groups and the trial?

The COVID-19 pandemic had a significant influence on the trial. When lockdown restrictions were first imposed in March 2020, we had to pause recruitment. We had recently m)et the criteria for the stop/go point (i.e. internal feasibility pilot) of recruiting 40 participants and having 2 intervention groups established. A third intervention group had also been recruited and was due to start in late March. Although we were behind our initial recruitment target, we were aware of the barriers to recruitment, had developed sound strategies to overcome these, and the trial had established momentum. In March 2020 intervention group 2 was interrupted and ultimately could only complete 7/14 sessions. Group 3 could not begin at all. This resulted in 14 participants randomised to the intervention pre pandemic not being able to receive the full intervention. Although we proposed that these participants be offered the online version of Mellow Babies, this was not approved by the funder. Thus, our Data Monitoring Committee advised we exclude those participants and associated controls from any outcome analysis, reducing our sample size by 28 and meaning we would have to over-recruit by that number to be able to meet our target sample size. Trial recruitment re-started in November 2021 and continued until September 2022.

Appendix 5 Baseline tables